Tech Brew Ride Home - (TWTR SPC) Stock Options; DALL·E; Internet Explorer

Episode Date: June 18, 2022

Some thoughts on what startup workers can do with their stock options (in this economy!?). DALL·E blows our collective minds. Considering the legacy of Internet Explorer. Thanks to @adam_keesling and @jnack. What Should You Do With Your Options During a Downturn? manual.withcompound.com New podcast: DALL•E & You & Me Sponsors: LinkedIn.com/techmeme Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Wow, I'm very excited. This is the first time I might be recording this without an echo in my ear, of myself. So I might actually sound intelligent.

Just as you said that, you started to pixelate or whatever the word is for that. Oh, wow. Yeah, I'm going to have to work on my internet connection now. It's ironic. You know, it's one thing or another in my new shed. Yeah, I, and they're like, hey, there's a problem with your internet. We want to come out and fix it for you. And I'm like, all right.

And they did. And the guy was awesome. He went like, he followed the line from, you know, the backyard into the house, place by place by place until he figured out where. And so far so good. That's awesome. I actually ran an Ethernet cable 100 feet from my fiber, my sonic fiber hookup all the way to the shed. It's a whole thing. It's a whole story.

I'm getting into like the whole house improvement thing. And anyways, but the Wi-Fi router that I had that was actually at the back of the house that was feeding. my shed is now in a in a skinner box basically it's like in a box out front to serve as the router and so now I have very poor Wi-Fi reception outback so we'll see we'll see how this goes this time shout out to Spectrum for actually coming and fixing my stuff but um god if I could have what what's your your one called again out there we have Sonic Sonic Sonic Sonic I've always heard Sonic's amazing yeah yeah I mean I will say um you know and this is probably not going on the pod but the guy that came out to set up my internet, so this is the day before I closed on buying the house,

we actually got the internet set up first. It turned out that the house had Ethernet like set up in all the ports, but the end of the Ethernet cables hadn't been configured with whatever the little plastic nubs are that actually connected the computer or something. Yeah, yeah. And so the guy went around and actually hooked up like eight or nine of these outputs, these jacks, you know, totally unprompted. I had no idea that Sonic would do this. It was amazing. So now I actually have Ethernet through the house, which I was just like through the roof. By the way, keep this in.

Keep this in if you do be the one editing this because my biggest fear, I think I've said on the show before, is I buy a house and find out that eventually the internet is just crap or like for some reason. Like that would make me sell a house. Like I mean, honestly, when I found out that there was Ethernet, I was like through the moon because I was like, it's going to be, you know, Wi-Fi is good, but it's not, you know, as reliable. It's not, you know, you're paying for fiber, but you're actually getting like, you know, you're actually getting like 300 megs. I don't know. It's like not a big deal. But when you're on video all the time, it kind of matters. So yeah, I was, yeah, I'm totally with you. If the internet sucked wherever I lived, I would be very disappointed. Um, okay. Uh, we should be respectful of Adam's time. Yes. So go ahead. kick it off. Welcome everybody to the tech meme ride home experience for June 17th, 2022. Today we have a two part conversation. Um, the first part is with Adam. from compound. We're going to talk about what to do with your employee equity stock options, which, you know, he has this amazing post that goes into all these scenarios and details. And we're

going to unpack some of that, especially now, given where the market's at. And then the second half is going to be in a totally different direction. We're going to talk about Dolly, which is this new kind of cheap BT3 style image composer, editor, creativity tool. Anyways, it's super interesting, super exciting. And by the way, if we have time, we might do a third little segment about pouring one out for Internet Explorer, which I figure you would have thoughts on that as well. I do have thoughts. I do have thoughts. I do have thoughts. So maybe they'll leave on this third round. First, let's kick it off with Adam. Adam, welcome. You're at Compound, which is a company that helps people in the tech industry deal with their finances and whatnot. If I'm being too

simplistic about it, but tell us real quick about what compound does. Compound is a, what we say is like a full stack financial management platform for tech founders and tech employees in particular. So a lot of questions come up around exercising equity, how do I manage my money,

do I need a financial advisor, how do I file my taxes, what are my tax strategies, all of those things we can help with. And again, we're like soccer forward and particularly help with this question of like, what should I do with my liquid equity? Well, and this is what I discovered. I shared it just earlier today in the weekend Long Reads.

You have a big piece out that is specifically talking about what to do right now. We should probably say up front that, by the way, don't take any money advice from this show. You know, talk to your accounting professionals. I'm going to second and third that my compliance officers could be very happy. you said that. Yes, yes, 100%. We're going to talk in broad strokes. Don't take advice from us, talk to a professional before you do anything. But so the piece that you wrote, it was for our friends at Evry. But as I said on the Long Reads today, you sort of go through the scenarios that people might be facing right now. And maybe what we should do is we should start out

thinking about people working at private startups and then maybe a few, if we have time we can talk about public companies because people there are in a slightly different situation. So let's just start from the very basics percentage of the equity for the employee pool, for the options pool, for your workers at your startup.

And then can you just in broad strokes outline the exercising process and how that works and like when how generally you're supply those options at the price that they're offered to you. Yeah, yeah, no, exactly right. So how this typically works for employees is you want to join a tech startup, right?

And so you go through the interview process, you get an offer. And then like the CEO or the, you know, head of HR, whoever sits you down is like, okay, here's your salary and cash compensation. I get at the startup, which is the equity compensations. And in that piece, typically what you get is options to buy a stock. You don't actually get shares, right? And so what that means is that typically you're granted a number of shares and a strike price. And that's how you can exercise your options.

And exercising your options means using that right to exercise actually purchase your shares. So this is kind of particularly a unique situation as startups because, you know, in addition to kind of investing your time into your career, which, you know, pretty much everybody does. everybody has to make a living. In startups, you also have this opportunity to invest money into your career, which is, you know, a little bit more atypical. So that's maybe like setting the stage. Yeah, and to try to put some like tangible numbers on this, like let's say you're join a startup, you're offered, I don't know, let's let's use round numbers to be easy. A hundred thousand shares at a certain price. And that's what you mentioned, the, the, the strike price.

Let's, let's call it your strike price when you offered them years down the road, your company has done multiple rounds at hopefully an increasing valuation. And so the price of those shares could be $10, right? And so you have the option, if you exercise the option, to purchase those 100,000 shares at the dollar. That's your strike price. And thus, in theory, on paper, you're instantly, it's a little complicated because of dilution and things like that.

But you own that equity. In theory, you've paid a dollar to make 10, right? Exactly right. Yeah, and so that's the kind of nice part about this, is that when you're issued options at a strike price, that's typically the fair market value, also called a 4-9-8 at the time that they're granted to you. And, yeah, hopefully the idea is that the company gets more valuable

and the share price increases, and so that these shares will be worth more later. And maybe one of the really sticky situations that certain employees get into, that is, again, something we're hopefully trying to solve here in the long term at a compound is you actually pay taxes on that difference in the share price, right? So if you have the strike price at $1 when you are granted the options and then it goes up to 10, that's really good that that value is increased and you'll only pay at the $1 price with the extra percent, you know,

get tax on the difference between $1 and $10, which would be $9 a share. And that can be extremely significant, and that's a very large cash expense that you have to pay today. So that could be, you know, five, six, seven figures worth of expense. And again, the hairy situation is, you know, even a company at Series D, Series E, not necessarily 100% guaranteed to go public and get liquidity. So you could, you know, pay all this money, not only to exercise your shares, but also to pay the taxes. And you're kind of in this kind of weird situation where it's like, yeah, I own something.

I'm really excited about the company, but it's like a really large cash expense and it's not guarantee. Because again, keeping with these round numbers, I've bought the shares at $1,000, so I've spent $100,000 to do it, which, by the way, you're buying it, so you have to have that cash to exercise the option. And then you're essentially, now on paper, you have a value of a million dollars, because again, 100,000 shares, but you have to pay the taxes on that, despite the fact that we're not necessarily talking about a publicly traded stock or something like that. So, these aren't the most shared tax implications and things like that. Exactly right. Yeah. And in that case, it's called the kind of bargain element is what the

lawyers call it, that difference between a million and 100,000. And so that 900,000 would be basically treated, again, depending on the exact kind of options you have as income. And so you're paying when you exercise, right, so you could be paying up to, again, a couple hundred thousand depending on your exact situation. So the reason we're talking about this now is because for the last decade at least, not for every company, but in general, for tech workers. What's the crypto phrase, Chris, line goes up or whatever. Well, yeah, yeah, always goes up, right, exactly. Yeah, so it's sort of been that way, but now we're in a situation where, depending on, well, and depending on when you got hired.

because think about it, you know, if you were like employee number five, even if your company does a downround, you could still exercise your options and they'd be above water. But if you got hired right at that last, you know, series C raise, and now they're going to raise a D as a down round, your options are completely underwater. And so this is sort of what your piece talks about. you give like several different scenarios and how in what cases what options people have in which cases they should maybe exercise the options or maybe not and one of the key things about the piece and then I'm sort of teeing you up to like give us some scenarios is in essence you say that you know there's an opportunity cost here and the opportunity cost is time you know your life and the amount of productive years that you have and things like that. So, you know, if you're working for a startup and all of a sudden your options are underwater, maybe you should jump ship. Maybe you could go to another company that

maybe isn't as far underwater, et cetera, et cetera. So walk me through some of these sort of things that workers might need to be considering right now. Totally, yeah. So let's walk through this kind of first situation just to kind of set the stage. So one of the things that could have happened is you joined a company maybe a couple years ago. They've raised a couple rounds. and so the value of your options to increase. And now they're just kind of coming down and in the down around, right? So your options, like you said, are still positive, but they're less valuable than they were, call it a year ago.

So there's a couple of key things you can kind of think about in that situation. So like one of them would be actually now might be sort of a uniquely great time to exercise your options. And the reason is, again, the, you know, if you believe in the long-term prospects of the business, then the share price is just going to continue to increase and increase. And if you owe taxes when you exercise, right, when you actually buy your options, then this would be the lowest spread, the lowest amount of taxes that you'll pay. So again, this is not a guarantee, but that's one thing, is that now, because the share price is lower and you'll pay less taxes,

it could be an interesting time to purchase your options. Another thing that's kind of interesting about that scenario, that many, many people also don't realize is, that if you do leave your company, right, you were talking about opportunity cost of time. If you do leave your company, your options actually expire. So this is, again, it's something that it depends on exactly the types of options. But the industry standard is that if you have options that are vested, so you're capable of exercising them, but you haven't exercised them, it's typically 90 days after you leave the company that they will essentially like disappear. there will be, you know, back to the company.

And so this is like a really tough situation, again, for people, right? It's like, if you worked up for this company for a couple of years, your share price is increased, but you haven't purchased your options. Now there's kind of this golden handcuff situation where you would have to pay, again, depending on how many options, hundreds of thousands, potentially of dollars, you know, out of pocket. And, you know, and if you don't do that, then your options will disappear. And then, yeah, and then finally, like, as I mentioned before,

this is not, you know, startups are inherently, you know, very risky, and you can kind of dive into some of the numbers there where early stage, you know, a lot of people are aware that most startups die. But even, again, like one of the stats I pulled is even a series D company. Yeah, I love that stat. Yeah, it's only about a 30% chance of getting to an exit. And, you know, and that's any kind of exit, that's not necessarily like a $5 billion outcome that changes your life, right? So there's a kind of, you know, yeah. What you're saying is that you could be at a at a unicorn, right? And you feel like, oh, we've been going great guns.

We're at the late stage or whatever. And so we're getting close to the finish line here, but you're saying that even then you have a one and three chance of essentially, you know, the payday coming good. Absolutely, exactly. So one of the things that you did say that people should start to think about is whether or not you believe, If you're in a situation like this where maybe your company's facing a down round or maybe things are looking dicey, it's time to evaluate your faith in that company because on the one hand, this could be a uniquely great time to exercise your options.

Explain why that might be. Yeah. So, again, there's, you may, you know, if you're working at a startup, you probably, you know, you're around people. you know, you kind of have some of the information maybe about how the company is doing. So maybe you have a kind of a hunch. Maybe you're like either confident or hesitant maybe is the two, you know, sides of this. And if you're confident about the prospects of the business, right, you believe in the founder, you believe in that your product market fit, you have a real business model.

Yeah, now might be actually, again, counterintuitively a uniquely great time to exercise your options. And again, the reason is because going back to the tax situation is you'll pay, taxes on the difference between, again, the strike price and the current 401A, the fair market value, right? And so if you were issued your options at a current strike price, again, let's say, going back to our simplified example of $1 per share, and it went up to $10 a share, and then now there's a down-round and it's at $750 a share, right? Now that difference, instead of being the difference between 1 and 10 of 9, it's the difference between 1 and 7.5, so it's 6.5. So you'll actually be paying less taxes because

that spread between the strike price and the current price is smaller. So again, this is, again, multi-dimensional, which is why this whole problem is not so easy, but that's one of the things there where you can actually pay less taxes because they repriced those, the company lower. Re-up to not take this as advice and talk to a professional. 100%. 100%. All right.

So give me the scenario where maybe I just walk away. Maybe I walk away because I feel like there's a better opportunity out there or maybe I walk away because it's like, I mean, am I going to really spend three, five years to figure out if this company is going to dig out of this hole? So when might I consider walking from these options? And by the way, with the caveat being that it depends maybe on if I've exercised some, exercise already, or exercise not at all. Absolutely, yeah.

So if you're a little bit more hesitant about the company, company. First of all, I mean, the equity is one piece of this, but again, you shouldn't be valuing your time as much as anybody. And so if you're kind of like hesitant about the situation, maybe an interesting time to think about getting a good job anyway. But specifically on the kind of options and equity piece, yeah, one of the things that, again, not many people realize is that your options will expire if you haven't exercised them. So, you know, if you're hesitant about the company, not a big deal, right? If you, the numbers show out that most, you know, most startups don't work out. And if you don't think that this is one of the really exceptional ones, then you're not really missing out, right? Like that actually might be a good thing. You're going to save your cash. You can go put that into, like, public market investments or if you have angel investments or crypto or something else. And so that's what you can do with that. And where was I going with this?

Yeah. And then you can, there's also the time factor around all of this, right? So this is something that kind of every company and every employee is thinking about right now is even in a down round, right? Or even in a scenario, another scenario, which would be like your company doesn't raise a downrun because they have two to three years of runway of cash. Any of these situations, it's like, are there more interesting opportunities where the equity isn't inflated? Or are there more interesting opportunities where you believe in the company and the founder and the people you're working with? And maybe she do you again, go look at kind of doing something like that instead. It's almost identical to holding a stock that has gone down 70%.

If you believe it's going to go back up to its highs, great. But if it's going to take seven years to do that, you could have that money invested in something else that could be making money all this time. Yeah. And just on that point, another interesting kind of framework to have for your equity is to think about this as an investor. And I think getting at your point is like this is all kind of a time we needed to return. Right. So like, you know, sure, you could have invested in something and, you know, if it doubles in value over 100 years versus doubles in value over one year, it's like a massively different return profile, right? So that's another thing you have to think about. I'd also say particularly on your point around the public markets, I think it's you're correct in the time-based factor. But another kind of interesting nuance of the private markets is that they are illiquid. And so you can't like, you know, exit a certain points or like, you know, there's some other considerations there. In addition, to, again, most startups, even quote-unquote mature startups end up failing. Can you tell me, let's switch to the other scenario now.

Let's say I'm an employee at Amazon, right? And so, and I got hired a year ago, you know, and the stock is down 30% since then. I mean, we're seeing lots of big public tech companies, you know, raise salaries, increased bonuses, even reprice options and things like that. But if I'm at a big publicly traded company, what should I be considering and thinking about if my options are in trouble because the stock has been down a lot?

Yeah, so I'm a little bit less familiar around this kind of situation, so I'll put that as a caveat. But I think it's a lot of similar kind of points, right? you know, maybe there's some other considerations because, like, one, typically people at larger companies have higher cash salaries. They have a little bit more of a cushion, perhaps more liquidity to, like, think about these sorts of things. But there's also kind of the situation of, you know, maybe thinking more about diversification in this kind of market, right? And, and, you know, if you're highly concentrated in something like a big tech company and you have, you know, and it's gone down a lot, obviously, hopefully your portfolio is, you know, kind of around a number of different assets and so that you can kind of leverage that below. But another thing that kind of generally, again, not investment advice, but just something that that I kind of think about is that, you know, it's, it's the data actually points out that timing the market almost never works out. So one of the things you can, again, you can always think about is like, yeah, the, you know, a lot of these stocks have gone down quite a bit and there may be a buying opportunity.

But generally, again, for sure, not investment advice, but generally, you know, something more like dollar cost averaging, sticking to a long-term view with kind of this, you know, the money you've sat aside for investing for capital raising. And again, just not maybe not worrying about kind of the short term quite as much is generally what I talk. One more time, not investment, not financial advice. But let's end with this one. I'm going to jump back to I'm working at a private startup. up. You do point out that one thing that is possible is to ask for option repricing or like a refresh in the option program or something like that. What's your advice in terms of approaching your boss and floating that boat?

Yeah. So this is something we're seeing all over from founders or employees, just kind of everybody, is like, how do we take care of our employees or how do employees make sure that they, you know, you know, if you're underwater, all these kinds of things, it's like a really scary situation. So, again, generally, what I would say is, like, similar to many negotiations, if you can prove kind of your value, then sometimes you can, you know, have a better outcome. So one of the things, for example, you can do is say, like, over the past, you know, six months or whatever the time period is, here are the accomplishments that I particularly, you know, helped push forward. and these are the business outcomes they have.

And if you kind of present this, I think, in a particularly thoughtful way to kind of of whoever is your manager or the CEO or whatever, you can be like, I really want to stay at this company and I really believe in the potential of it. But I just think that, you know, kind of repracing the options so they're underwater or having a refresh, you know, can be something that would be really motivating to me. And again, typically, you know, a lot of times companies are, you know, are, you know, receptive to this in certain situations, especially because, again, think about it from their perspective. If all of your, you know, in terms of like a refresh, for example, if all

of your options were already vested and you exercised all of them and owned all of them, you certainly have some incentive to stay around, but the company would actually like a percentage of that to be, you know, in the future and to have this, again, kind of incentive for you to stick around longer. So there's also ways that, you know, it's not, you know, obviously you would be receiving more value in a repricing or refresher, but there's also good cases for the company to do it too, so it's not completely sort of this one way asked. He did get in trouble for this. Oh, yeah, sorry, go ahead. Just practical question about that. You said that you could ask the company, but can you be a little more specific about who in the company you might ask about that repricing option? Yeah, and I,

I'm actually not going to say too much about that just because I am not kind of an expert on this particular situation. But generally, yeah, and I would say generally that this would be something I would assume you ask anybody

you would talk about a promotion or compensation questions. So typically that is either your manager or sometimes it's handled by like the general council,

head of people, someone like that. The snarky thing I was going to say is you're in good company if you asked to reprice your options because Steve Jobs did that.

He got in trouble for it. Apple got in trouble for it. So we'll just leave that there. worked out though. Yeah. Yeah. So by the way, thank you so much for telling us all about that. But I discovered today by, you know, preparing to talk to you that compound has these sort of manuals. They call the manuals. Like if you're an early employee, if you're an employee approaching a liquidity event, early stage founder, just joined a unicorn. There's really interesting data and and like resources here.

It's at manual. Dot with compound.com. Yes. So check that out. But Adam, is there anything else you want us to know about compound or you want to plug yourself personally? Whatever. Yeah.

No, I think going to the manual, manual. That withcompound.com is a great resource. We've gotten a lot of good feedback on that. It's, again, like you said,

like you said, a good educational content around these topics. I will say. It's such a well-designed resource. Like, Once I found it a couple months ago, I was like, oh, my God, this is gorgeous. I am completely in all. Our design team is top notch.

I'm constantly impressed by them. But, but yeah, and if you, so head over to the manual, I think that's a great kind of early introduction and also spend the time just to like think about this and kind of get it right for yourself because it'll be such an important and kind of, you know, high value use of your time. Additionally, we also have software tools to help you manage your employee equity, model a lot of different scenarios, kind of similar to what I did in your article. And you can sign up for that at withcompound.com.

We currently have a wait list, and we're kind of being really intentional about making sure everybody has a good experience and kind of, you know, letting people off that intentionally. But that would be another thing that I think I've also had some, you know, behind the scenes peaks into that. It's a very exciting tool that we're really excited to show people too. So manual.compan.com and then withcompon.com. But yeah, thanks so much for having me on. Yeah, well, thanks for sharing all that with us.

hopefully it helps people, or it gets people thinking, gets people thinking. Yeah, I mean, I just want to, like, jump in. Like, I mean, I think this is so useful from a resource perspective. These are all questions that I didn't even know how to formulate when I was going through my own startup journey and career. I sort of embarrassingly, embarrassingly, once was an advisor to Visco, the camera photo sharing app, and they'd given me some, like an equity grant very, very early on. and I totally spaced it on like the 409 valuation and like locking in the price and like all this

other stuff. And I have no idea like how much it's worth or whatever. But it's one of those things where had I known that this was even something like a set of considerations to make, it would have definitely, you know, made a big difference in probably at least to some degree my economic security at this point. So, you know, for anybody who is not thinking about these things or just like, whatever, just knowing that there are these choices and options available to you. And especially as you said, like not freaking out right now given the sort of macro market conditions and thinking about your long-term goals and plans and then thinking about how these different scenarios, Adam, that you've outlined can align with them. I think it's just so useful and valuable.

And I'm really stoked that you put this resource out there. Perfect. Yeah. Thanks, guys. Completely agree. And that's another thing we often find is that people actually don't think about these things until they've had that situation where it's like, oh, shoot, like, I should have done something else. And so we're really trying to change that and kind of get the word out through some of these resources. But But yeah, I appreciate you guys having me on. Awesome. Well, thanks, Adam. You're welcome, of course, to stick around.

But now we're going to talk about something completely different. And actually, let me change the title of this space here. We're going to actually, first I think we'll do an intro, and then we're going to bring up a new front of the show. His name is John Nack, who actually works at Adobe and has worked on Google image tools and things like this in the past. He was a Photoshop PM. I'll actually have him do a little bit of an introduction himself.

But before we get there, Brian, do you want to just like kind of give us your interest in this? Because this is something that has been coming up lately in a number of different, I guess, context categories. Casey Newton was writing about it. And it just feels like it's kind of exploded a bit. But you seem to be particularly interested in this. Just because it's been a while since sort of technology has wowed me. And I think that's what everyone's reaction has been, where it's like, oh, you can do this.

Do you know what I mean? This is maybe related or a poor analogy, but I remember early on in the smartphone days, I took my wife to Paris to ask her to marry me. And that was like the first time that you could go with a smartphone and point it at, you know, signs and stuff in French and have it on screen translate it for you. And I was like, oh, my God.

That's amazing. I remember that a couple years ago. and now it's like built in and you take it for granted. Right, right. You don't even, yeah, you don't need an internet condition. I remember download, I had to download like a module so that it could do French or Italian or whatever. Anyway. Yep, I remember that too.

It is sort of like that in the sense that if you're not familiar, essentially Dolly is this program where you can literally say, I want to see a picture of a Victorian child wearing a VR headset, I think is one that Casey Newton did. Yeah, this is like not even a joke. It's like literally that phrase was the seed phrase. Or you can say, and I wanted in the style of, I don't know, the Duesbury cartoon. Yeah, or whatever. Or Dali himself. Yeah. Right. Exactly. And it'll give you like 10 different options and they're all good. See, that's the, first of all, it's impressive to me that this is possible. But they're good. Like, they're not. They're really believeable. I mean, I will say so they're like that we should probably talk a little bit about what this technique is and what this technology is. You know, there's a lot of deep learning,

um, AI and machine learning techniques that are going on that are kind of these enormous kind of branching trees with different types of logic that run. Yeah, that's what I want to find out about. Yeah. Like current processors, you know, is able to happen so fast. And they're able to take influences from all these different sort of parts of the tree and then recombine them together into something that is passing as an image, you know, based in all the training that they've gone through. And I think what's so incredible about it is that it really is this kind of uncanny valley moving into the Canny Valley where you look at it and you're like, that could actually pass for, you know, insert X here.

And in fact, I've been talking, I guess following this guy, Ken is actually where I found John. Ken works at Humane. He previously worked at Apple. And he's been doing a bunch of, I, I guess just experiments with Dali. And some of them relate to kind of taking Da Vinci style images and then applying them to things, of course, that DaVinci would never have been able to design or draw.

For example, UFOs or, you know, televisions. And so what I find so fascinating about this, because it's so believable and it's in this, like, old style that you can imagine people, you know, unaware that this technology exists, suddenly discover these things on some Reddit forum and actually believe that they're real. And somehow that spirals up in some new, I mean, like, if you think about the effectiveness of QAnon now combined with Dolly, like, it's going to be just mind melting and mind blowing and et cetera. Anyways, I won't talk anymore. John, do you want to come up and say hello, introduce yourself and kind of tell us a little bit more about your experience and background maybe in computer graphics and then how you came to Dolly? Yeah, thanks.

Thanks for having me. Wow, where to start. So my family is actively teasing me now because I am so like Dalit Pilled. It's crazy. Reading your blog. I'm going to pin out a link to your blog. It's been amazing to see all this stuff that you've been finding, but keep going. Yeah, yeah, thanks.

Gosh, so yeah, kind of speed run for me. I was a kid, you know, using Apple II computers. And then as I was tweeting in response to Ken today, I remember. being at my friend's birthday party, I was nine years old, I think, when the Mac arrived. And one of the big treats in my career has been getting to meet Bill Atkinson, who was the creator of Mac Paint, and telling him just, you know, what a sort of salient moment in my life that was. I mean, I literally, it was like it was yesterday, because I had been the kind of kid who'd love to draw. I remember, you know, being three years old and figuring out how to draw a car on a cocktail napkin at the Algonquin Hotel, you know. And so this moment where I could draw with a computer in the way I could draw on paper,

yet I could produce things that were totally impossible on paper, was so revelatory. I remember I drew a jail just because I wanted to use the floodfill with a brick pattern. And it was just, it's so trivial now, but it's the kind of thing that was this step change in what you could do expressively. And throughout the 80s, I begged my parents for a Mac. and, you know, they were super expensive. And so it wasn't until college, you know, some years later. And I ended up joining a startup in the late 90s, you know, doing like design and animation.

And every time Adobe folks or macromedia, whoever would come by, I was very nice, but I was always very, really, like, I know you're good at building them, but, but like, why does Flash work this way? Why does Photoshop work this way? So I ended up getting into the whole PM jam really just to try to make my friends happy and try to make the tools I wanted to use, they wanted to use. And, you know, I won't belabor all this, but basically spent 14 years at Adobe.

I spent 10 of those on Photoshop, got to work on, you know, content aware fill, things like that, which were, you know, another sort of, like, I think that's a very important thing to just call out because content aware fill, when I saw it, you know, and I'm someone who is, you know, like a naive kind of Photoshop user. I mean, it was like a game changer, right? Because so much of a photographer's journey in,

you know, editing photos and so on is kind of using like the, if not the blend tool, the patch tool and kind of copying over, you know, background stuff and, you know, trying to replace weird people who showed up in your photo in the background

or something trying to make it look believable. And here Photoshop was just, as you said, sort of like, it was like that powerful moment where not just like, you know, like you said with MacPaint, like creating a rectangle,

than wanting to fill it in with like the fill tool just because it's like so cool. But now you're doing that with like reality. You're like filling a box with reality based on the things around it in a believable way. And I think that was one of the first moments for me when it was like, oh, wow, this is like a total game changer. Like I know I know I don't want to get to like Dolly, but I'm just, I'm curious given that you were at Adobe. If you can unpack a little bit of that journey from where it was kind of maybe more of like a manual artistry kind of sense where the photographers or the editors were doing work kind of themselves and there was like an art to it versus realizing that the software could actually kind of become a creative collaborator, if not kind of execute some of these

things and what that experience or conversation was like within Adobe. I mean, was there resistance or was it kind of like, let's, yeah, let's do this. Let's make it happen. Yeah, gosh, that's a great, it's a great question. So it's interesting. Kevin Connor was my boss on Photoshop for many years and he used to talk about how tools were sort of, okay, so, so he was a VP of imaging and he he would sometimes point out that

there was this progression from you know very very simple tools like in principle you could kind of do anything in Photoshop 1.0 right like you literally could take a one pixel brush tool and you could make stuff right so it's all possible

but obviously impractical and you know when I first joined Photoshop this was God 2002 they had just introduced the healing brush and the patch tool, which were mind-blowing, you know, as you may remember back then. And Kevin's point was that, you know, the clone stamp tool had been around for many years by that point, and it would just sort of naively let you sample one area, copy it to a different area.

And it's still incredibly powerful. It's a fantastic tool, but it's very naive. It's just kind of, you know, with the exception of maybe a blending mode change or opacity, it just kind of dump stuff from point to point. And he brought up that, you know, the healing brush is really that tool. it's just more opinionated and specialized for certain kinds of things, like retouching or object removal. And, you know, I think of that a lot because I see it happen again and again,

where, you know, content or fill was, you know, much more sophisticated and could look at different patches and synthesized new textures. And now when we jump ahead to something like Dali, I've been calling it content or fill cubed. You know, I don't know what the power is, but it's like, it's now hallucinating things from whole cloth. In fact, I just retweeted a little experiment I did with a picture of Russell Brown. People who know Adobe may remember Russell. Basically, Atari laid him off in the early 80s. He didn't know what he was going to do. And he's like, well, what's this

like a printer startup? I don't know. Adobe, I guess. And so he ended up going there. And he ended up being one of the four names on the Photoshop One-O splash screen. Amazingly, he's still around. His son is now my designer, which freaks me out because time goes by. But I mentioned it because I just took this headshot of Russell. I selected his face, and then I had Dully inpaint the rest. It's off the chain, right? I just pin that tweet. And yeah, it's amazing. Okay. Yeah. So, but back to your question about the, the genesis of content or fill integration. It feels really similar to me in the sense that I had a great relationship with the Adobe research team and they in turn work with tons of academic folks, you know, much like the Google

team does and meta and everything. And some of the most fun I ever had would be to go to off-sites and they would, you know, show, hey, I've got this thing and kind of cooking. Like, what do you think? Like, there might be something there. They might not be. And I would bring perspective from what I've been, you know, trying myself and talk to the customers. And at one point, there was this paper called Patch Match. And it's really, it's worth Googling. It's still a very interesting paper from, I think, 2009. And it kind of promised the sun and the moon. It was like, you could select this, you know, gazebo in a, you know, a yard and move it and then, you know, clone it or switch it or you could like reconfigure a building. And I looked and I thought, wow, that, it's incredible if it works.

I have no idea if it does. It seems a little ambitious. But what if we just took one little piece of that puzzle. What if we just took the part where, for some reason, there's not a hole in the yard or the sky? And, you know, we got to work. And a couple of the engineers, a guy, Dan Goldman, who's now at Google, Avoncarra Ballundry, he's at Apple, it's a very small world, right? They took the patch match paper and started pranking on it. And we thought, okay, here's one piece of the puzzle. It was, you know, incredibly slow relative to what we shipped. It was, you know, God, probably 10 times, 20 times slower. and they really did all the hard work to make the research code become practical. And then once we had that, it was like, okay, the natural touchpoints would be, you know, you could just hit delete and you could do a fill, or you could make this method within healing, for example.

And the reason I'm mentioning it in the context of Dali is, I think sometimes you look at this whole huge, like, well, it could synthesize stuff, it could, you know, it could do a million things. I think one of the most exciting parts is if we can kind of break. down the constituent pieces, whether it's like auto generation of layouts or style transfer or, you know, object removal. I think what we'll see is these general methods of this, you know, diffusion-based, text-driven synthesis will end up rippling out and touching lots and lots of tools in maybe unexpected ways, maybe in really narrow applications. It's not to say that the, you know, the big top level thing like people are seeing from Delhi won't be hugely popular. I think it will, as long as there's enough GPUs on the planet to run it, which there aren't,

apparently, relative to the demand. I mean, I think that's the weird part is like there's, you know, I think an enormous backlog of people who would like to have access. It's just, it's physically constrained right now by the amount of compute required and scale in that. I almost did a story today about how GPU prices might suddenly crash because all of a sudden, Ethereum is moving to proof of stake. But neither here nor there. Continue your story. Well, no, I think it's funny.

You know, I listen to a podcast where, gosh, it's Ben McKenzie, is that his name? The guy who was on the show, the OC. He's now written a book about crypto and talked about walking into some, you know, hangar-sized data center in Texas where all these machines are just whirring all day, you know, doing crypto mining. And I was thinking, gosh, maybe if, you know, that praise creators a little bit, who knows, maybe the capacity will free up to let people do more artwork creation, which, in my view, has more intrinsic value to it.

But who knows? And, yeah, the other funny thing I discovered is the PM for Dali is named Joanne, and we got in touch. and she said, oh, you're my brand mentor. And I was like, what? It's, yeah, I'm your brand mente. Right? I know, right? So she was at Google, I was at Google.

We have a friend in common in between. And I didn't know that I should turn into dust or walk into the ocean. But I told her, you know, I've been doing this at some combo of Adobe and Google for 22 years. And I honestly, my wife and I are, you know, same age. And we both kind of looking at our career like, well, do you want to do this forever? Do we want to go teach high school? Do we want to open a coffee shop? And then, you know, Dali decloaked.

And right afterwards, Google decloaked theirs, which is called Imagine. There's also Mid Journey. There's disco diffusion. Like, suddenly it's, you know, this zero to N production of these models. And I was saying to Joanne and others, like, it has flipped the script for me where I'm like, thank God I'm only 46 in quotes. Right. Whereas previously I'd been like, well, maybe I've come and seen and done everything I can do. I went to Google with this stated mission of teaching Google Photoshop,

meaning like, how can I move from serving the sort of 1% of people who know and love tools like Photoshop? And they've got, you know, the energy and ambition and need to go and master those, which was, it's a super fun market to serve. And that's where I come from. But at Google, it's like, hey, could you reach 10 times more people, 20 times more people? Did you work on SnapSeed or Google Photos? Oh, boy. I did.

I did. There are stories to be told. Well, sorry, sorry, disclosure. I worked on Google Plus. And so I was also with the photos team. And so I'm very aware of, you know, some of the aspects of Google photos and just the, I mean, the crawl and the index and the inventory of content, you know, that was publicly available to the service that could be used for training models. So that's why I ask. No, no.

It's a great point. And sort of my origin story was, you know, after Instagram, got purchased in 2012. I wasn't at Google, but my understanding is that he was running thing, his conclusion, though, is, well, Google needs to own the means of production when it comes to photography and started getting really serious, it seemed. They bought up Nick Software, which was, you know, 120 people in Germany.

I knew them super well because they were Photoshop developers. Like, super nice guys, did great quality work. And I'd always wanted to buy them at Adobe. I just, you know, I was afraid Apple would get them and put it into Aperture, you know, which thanked me this for, us they did. But then I would joke because I was like, oh my God, at least Apple would have charged money for it. Now Google's going to give it away. They did. And so the next year, I pinged my friend Josh Haftel, who was there. He had come in with the acquisition and said,

right. I work with Josh. Okay. Yep. Memorable dude, you know. And, you know, just, yeah, a colorful, literally colorful guy coming back to Burning Man. And he was like, oh, he's like, bro, I'm on a plane with Aung Sang Suu Kyi. I'm piecing out to Burma. You know, I'm going to go walk the earth after, you know, he'd investing at Google.

Exactly. Right. My journey was that I wanted to take these tools I really love, like SnapSeed, put them at the disposal of the sort of Google automation staff and then make a world where

people like me who really care about crafting particular looks, kind of like Visco, right? You know, that, you know, Lightroom creators. It's funny, Visco came out of, you know, like three guys taking a little utility we made for Lightroom and creating some really great looks and then selling them as like Triax and Kodak and, you know, Delvia. Anyway, so I wanted to say, hey, if we have the tool that a small but passionate subset of people,

small being tens of millions, you know, but small by Google standards, would want to you to bring that to everybody right exactly and do it in the way where for example my wife and son are on a road trip right now they're taking photos um it it will always okay okay but wait so because because like i i'm so excited to like hear kind of about the the connection into dolly right so yeah yeah totally fine is is like wanting to deliver these tools to the masses and there you go what brian was saying right was that you know now we may have something extra GPUs available because crypto is, you know, in the shitter. And now if we can actually make a tool like Dolly and it's, you know, sort of brethren

available to a much broader audience, I think the thing that I'm curious about your, your perspective to bring it into Dolly, you know, is. So one, of course, we can talk about the mechanics of how it works, which I think it's one of those cases where, you know, anything that is, what is it like, I don't know, the whole magic quote, like, it's indistinguishable for magic or something. Like, in this case, you see these images and you literally like type in a prompt and the computer responds with here's what I think this should look like based on looking at billions and billions of images and people seem to associate these words with this look and feel of a thing. And so when I put those things

together, this is the output thing you get. When it comes to creativity, I guess like what would be kind of interesting to hear from you is the way in which you've seen creativity change as a result of Like, for example, you know, another sort of thing that I'm trying to tee up for you is if I think about introducing the idea for the hashtag in 2007 and then seeing the various ways in which people interpreted and adopted that idea over time and used it for both good and not so good output. From your perspective, you know, from from Adobe, from Google, from Content Aware Phil with Dali, where do you see this kind of going in terms of adoption and usage and application? Yeah. Yeah, no, thank you. Thanks for the redirect. I get, I get, we'll go ahead. What does this do for creativity, right? Because there's this question of replacement of human talent and ingenuity. And on the other hand, complementariness, which is kind of like the content

aware of Phil thing. Yeah, yeah. So I forget, maybe you've seen this article from some years ago. Somebody wrote a post called shit work. And it was like basically don't get people shit work. Right. Like a lot of digital creativity has been that class of stuff, you know, part Part of why I joined Adobe was like, to get a 20-layer PSD into Flash was 168 steps. And I know that because I counted, because I used to do that nights and weekends. It was just like a bizarre misallocation of time and talent. And coming to Adobe, I'm like, great, we'll make it two steps. And three if you want layers, right?

Bang, you're done. And it was like, oh, great. So now, much like the GPUs, my brain is freed up to do something interesting. And really, I think the promise of Dolly-type tools is that. I think it's obviously disconcerting, and I don't want to be Polyana here. I do think there is

some class of work for which people are paid now, which will go away, and it doesn't matter if we like it or not, it's just what will happen. I think the analogy that a lot of creators have been making is to photography in the 19th century, where, you know, similarly, there was

a huge rebuke and revulsion from traditional artists. They're like, how dare you? you know, I've spent my life with what was then the cutting-edge technology, oil paint, making that, you know, reproduce reality as closely as a human could. And suddenly, more or less anybody can reproduce reality in that same way. And I think what it freed up was once you don't have to be a mimic in that manner, you can do other things. You can go and become an impressionist or a cubist.

This is so much what I wanted to ask you. Okay, okay. So very much. The photography analogy is beautiful in the sense that, you know, you're an artist in the 19th century, 18th century, and you spend your whole life essentially trying to replicate reality, right? Right, right. Now, there's also, you know, aesthetics, personal tastes and things like that, and talent that goes into it, right? Am I right in thinking that essentially the way that, essentially the way that, you know,

the AI works in this, it is still pattern matching. Yeah, yeah. So here's the interesting thing for me. You know, I came back to Adobe to work on what we were calling, and still are calling AI-first creative tools. And it's like, what does that mean? It's what nobody knows what it means, but that's what's fun to figure it out. And Open AI, which makes Stully, has this technology called Clip, which, you know, my layman's understanding is it basically what you just said. It's this mass association between visual, and textual descriptions.

So it's trained on hundreds of millions of image caption pairs. And that's how it learns, you know, what is a dog? What is a Renaissance painting? What is a Renaissance painting of a dog? That kind of thing. And my team at Adobe made a paper called StyleClip. So basically they took Clip, the language model, and paired it with StyleGAN, which is a generative adversarial network from Invydia to make StyleClip.

So the upshot of this is you can use text to edit a face. First, you can create a face, but you can also edit a face. And so it's amazing and uncanny. I can type an old Irish woman with a page boy haircut and like my mom and her sisters pop out. And it's like, oh my God, this is very weird. Now, what you instantly discover, of course, like people are doing with Dolly Mini, which is not at all Dolly, but it's like an open source version that folks can access is your inner beavis comes out instantly.

It's like this id magnet. And you start wanting to type in inappropriate, problematic stuff. And guess what? It can do it because it has seen the internet. And so a lot of, I mean, here's an example, is like beautiful, right? What if you type in the word beautiful or there's a beauty slider? Well, naively, if you don't correct for the biases in the dataset, it will make people younger, whiter, and more female.

Because that is what it has learned. That's the bias. Exactly. It's looked at the corpus of advertising imagery across the globe, and those are the trends that are present. Similarly, if you make somebody older, they also look more male because there's a disproportionate number of older men and younger women in the data set. So it kind of conflates these terms. And this gets to a lot of the issues people raise with something like Dali, where if you type in CEO, overwhelming you get white and Asian men. If you type in nurse or health care worker, you overwhelmingly get women and women.

of color. And this is a much longer interesting debate. We have a different time. But like what should be produced? You know, what do we as like ethical creators want to see? How much should our opinion influence

what the machine does? And so this is the stuff I've been wrestling with my whole time back at Adobe. I mean, it's brilliant tech, but it's also you know, super risky. And the last thing any of us want to do is make something, you know, use to either create fake news or something hurtful or problematic.

So it's, again, you know, if it is patterns, it depends on what patterns it's trained on and what patterns it executes on. So this was the question I was trying to lead towards is what it's been making me think about is to what degree, and maybe none of us here are qualified to even opine on this. That's what creativity is, is like patterns and, what your talent is, you know, if you look at different cartoon artists, you can tell, you know, a Bill Waterson sketch drawing from who's the guy that did, you know, Garfield or Peanuts or whatever. You know what I mean? Because those are patterns that you learned in the same way that, you know, Hemingway established a pattern for his words and use of words and things like that.

It's been making me think, like, I know that what, when you say, okay, I want you to do a, uh, uh, a cityscape of the Metaverse, but in the style of, you know, Van Gogh, it is doing those patterns. But to what degree is human creativity the evolved patterns that are ingrained in our brains? And also, maybe I'm getting way too, we're smoking weed in the dorm room here, but like that's sort of what I'm thinking of. Not yet. No, it's early on the day, remember. Right. But any thoughts on that in terms of, are we revealing things about how creativity works and how maybe not talent works? But it is sort of pattern matching that then is a learned sort of expressive behavior.

Yeah, totally. I mean, I think of that famous quote from our story about Picasso where a woman asked him, hey, will you draw me a picture of a bowl or something? And he thinks for a second, he does this flourish with his hand. He gives her this drawing and says, you know, that'll be $10,000. And she's aghast and says, you know, how dare you? That it only took you a moment. He said, no, ma'am, it took my whole life. Right.

Like he had spent his life. Like, have you thought of all the GPUs I've had to run to get to this point? Right. But it's like the GPU, I mean, machine learning is trying to approximate what humans do. But to do it at a superhuman scale. Wait, wait, okay, I want to ask you something about that. Like, because one of the things that I've been noticing is about human perception. And when I notice or take a look at like these, actually, the dolly ones tend to be quite realistic in the way that they are portrayed.

But the hugging face kind of dolly light or mini approach probably is trained on fewer images. And so the perception that it produces is much less kind of accurate and it's more impressionistic. So it kind of takes shortcuts, which I presume are based on. human perception. So my question kind of is, like the thing that that I find so interesting, right, and we recently had this, there was a Google researcher, I guess, that, you know, was kind of believing that the AI that they'd created is now somehow sentient or something. We start to have these, it's not just that we're anthropomorphizing kind of artistry or creativity in the AI, but that ultimately creativity is a form of curated probabilities being

kind of discarded or accepted. That's a better way to put it. Yeah. Right. Because when I was thinking about like the AI researcher at Google having this conversation, this like Elizabot style conversation and getting realistic responses back, it's also based on what our anticipated expectations of the response would be. Right. So one of the things that Ken did recently was was kind of Batman doing different things, Batman drinking coffee, Batman going to McDonald's like in the Batmobile. And Dali, of course, produces passable images that certainly at a glance, you believe that that's what you're seeing. But on closer inspection, you know, it's sort of like I saw the everything all at once or

happening all at once moving or something. It's like hot dog fingers, you know, like where, you're like, that's not quite right. So my question is, is about the artistry that's kind of achieved here, the believability of it and the precision that humans might either demand or expect, you know, like based on like content aware, Phil, right? Like that to me, is a tool that only is so magical because when you do it compared to what you would have done on your own as, you know, putting human effort and talent and skill into it, you're like, actually, this is just about as good as I would do. Whereas Dolly is producing something that's a little bit different. So how does that change? Right. Like, I'm just, I guess what I wonder is do we start to let go of a certain set of human learned skills? Because the computer can just do a good enough job that we no longer need to learn watercolor or oil painting or things along those lines.

It's the photography analogy again, yeah. Right. Yeah. So I've been talking to a lot of artists and scientists about this. And, yeah, in a way it kind of, depending on, you know, like importing a PSD file. Or, you know, could be something people tactively physically enjoy, like doing watercolor. Right. A lot of people don't want that automated way.

But it does give most of us superpowers. I mean, I know, I know just like I go to CrossFit, you know, and I suck at it. And it's like, all right, but you'll grind away. I put on 10 pounds or 20 pounds, you know, and I got a little better. And suddenly, if somebody shows up with a bunch of mecksuits, and it's like, guess what, everybody, everybody can lift 10,000 pounds. Oh, cool. And now I incrementally, I'm still trying to add my 10 or 20, but it doesn't look so impressive.

I will say, you know, at Google, one of my ambitions was to make Google photos kind of be like Anna Winterer at both. Like my mental model is she's hauling ass down the hallway. where assistant pops up and is, you know, man, we've got three covers of the March issue. Do you like this one, this one or this one? She's like, this one's crap. I like the middle one, but lose the sunglasses. Give me the scarf number three. Right. And then off goes the lackey, and they sit in indesigned Photoshop, whatever, and they make it happen. So she still applied her creativity and, you know, taste and life experience, but didn't have to get mired in the execution.

And tools like Dolly promise that a lot more of us can kind of function at that level. I think what's interesting is it is always lossy, though. I think of like when the web evolved from using Flash to HTML. And to this day, there's things I could do in Flash 20 years ago. I still can't do on the web. It'll probably always be that way. It's like MacOS 9 to MacOS 10. We lost some stuff.

In aggregate, it was a gain. But I feel like because people don't have to do the mythical 10,000 hours of, you know, using shock, using French curves, watercolor, whatever it might be, or watching all the movies or, you know, do it. the things that make you this perceptual being, which can then execute in different media, yeah, to some extent, it cheapens the coin, right? It's like you didn't do the work. You know, you're just in the mex suit, whether it's, hey, you lifted 10,000 and I lifted 10,000. So the thing that the way that's, it's interesting is if you end up inventing a new sport,

let's say, using those mex suits. So what I'm curious, you know, again, if you're thinking about AI first tools at Adobe, you know, given the move to the metaverse, if you think about content-aware fill for the Metaverse, there are certain, maybe I'll describe this concept. Maybe I've been hanging out with Josh Heptel too many times. But as I've been told about taking acid, there are moments in your journey where maybe you're starting to go down a path that is not so positive. That's the start of what might be a bad trip. I can imagine being in sort of a virtual world or metaverse kind of reality, a place in the metaverse where I'm starting to like not having a good time. Like it's kind of boring or it's just like an interesting or something. And this content

aware field concept for the Metaverse could be super interesting where I just sort of decide to, with a few prompts, decide the type of direction that I want my experience to go in. And then through tools like Dolly or other types of, you know, kind of again, content aware in the most general sense, right? Like content could be music. Content could be visuals. Content could be objects or game like, you know, activities or narrative, stories, whatever. You can almost invent. kind of like directions for the MetaVosch to go in. And so then my question back to you is about being, you know, a tool builder and creator for people.

What like how do you maintain a sense of agency in that world where like you said, you know, it's not just that now you have like these mech, you know, armor. And of course you can lift everything. You don't need muscles anymore. But you're actually engaged in the need to kind of overcome or to master something. Like what does mastery of Dolly look like for an artist that's maybe starting out their career today?

Yeah, fantastic question. It's interesting now that there's this emergent genre of prompt. Exciting time because it's still only a very small number of people who can access it. Unfortunately, it's not by design. It's just, you know, constraint. But I think prompt... Wait, are you saying, by prompt, you're saying like almost the Zen cones that people, like, poetry are suggesting that Dolly do? Is that where you're saying? Wow. Okay. Yeah, yeah. So, exactly. So there's this, you know, subset of ML prompt engineering. How do you figure out terms that produce desired results and do you control that? So Mario Klingman, Quasimondo, is his handle. He's been doing amazing, you know, essentially there could be a Shakespeare someday that has the most beautiful way of prompting the AI to create the most beautiful image. Oh man, that's amazing. Right, right. It's a good pat.

Yeah, so the point I'm just trying to drive is he's tried to come up with a term for what is that art. And then we was asking GPT3, which took from OpenAI, for suggestions for like, how would you characterize the art of? It's unbelievable. And then you talk to people working on these systems are like, oh, yeah, it wrote poetry. And, you know, I mean, I really think everyone's continuously being surprised. I always think of it as that March on Lunch thing where, like, every day I think I'm going to be. ready for it. I'll kind of be over it. I get it. And it's like, no, it just runs through your face over and over and over. So, but anyway, the prompt thing, yeah, I think actually this, it'll be

be interesting to see if this is a long-term thing or whether this is just sort of a brief waypoint until we can make visual tools that then sit atop the prompt engineering where, sure, you can dial in language and we've been doing that with the tools, you know, my team's been building. But you can also just, you know, make a slider, make a preset. If it's a, a, a, uh, if it's a, latent space, you can use things like grids, you know, to sort of traverse all of these possibilities. I mean, you could start with Minecraft, right? Like, it doesn't even have to be like that complex, but it can be, again, suggestive. It can persuade you of being in this immersive environment that's totally, right? I mean, your incantations, you know, can be like these

prompts and dolly. Yeah. Right, right, right. Exactly. And, but it's interesting because, you know, like you were saying about anticipation, it's that feeling of, of not only is maybe one's work less special because everybody can do something similar, maybe oneself is less special. There's this thing called the Dictionary of Obscure Sorrows, and there's a term they made up called VemmoDalin, which basically it's the realization that the photograph you're taking is one that a million other people have taken. And people, of course, can now scrape Instagram and they're like, oh, look, it's the legs looking at the beach in the water on vacation shot, right? Or it's the girlfriend pulling you along by your hand shot

or whatever. And you're like, oh my God, like I am in the Matrix. I am a dust particle. And I think of the Louis CK bit where, you know, everything's amazing and no one's happy. Right. And he made this joke, you know, in the era of people just getting cell phones. And it's like, it takes two seconds and yet everybody's pissed because it's like, it's going to space. He's like, give in a minute to get back from space. But I tell you, even four years ago, I went to visit in my friend's class at high school class. He's a graphic design teacher. I was showing them the AR stuff we were building at Google.

And to your point, if you've ever tried to do any of this stuff by hand, like if you've gone on after effects and try to select somebody in rotoscope them, you'll just want to. Oh, yeah. It's terrible. Shit jobs. It's much better now since you have like ML driven tools. But historically, it completes shit work, right? And you could be doing something massively more interesting if you weren't sitting there.

And hey, I got my start doing that stuff as an intern. And thank God I did. But thank God it's gone. But I tell you, I showed all these 16-year-old guys. So I was basically showing a real-time body segmentation, which is now in, you know, Snapchat and Instagram and TikTok. And they're like, yeah. And so we had our own version. It wasn't in those tools yet.

And I was like, I was fascinated and amazed. Oh, my God, guys, it was such a breakthrough. And these guys were graphic design students. So if anybody should appreciate how hard it was and how much better this is, it would be them. And yet they were like, cool. And I'm like, no, no, no, but wait, did you not get it? And I was, no, I get it. Yeah, it's cool.

And so. That's always the story I tell about I was the last generation to learn how to splice physical selm when I was in film school, you know, like, which was it, which, which, which, which was a craft, an art, but is meaningless. And so like every tool in the same way that we don't have abacuses, abacchis, abacai anymore. We have spreadsheets and things like that. Right. Like you're saying, this is getting rid of the shit work or whatever the term. Right. Well, I have to, since you mentioned Advocas, I'll tell you a funny example. So there's this artist named August Camp. And before I had Dolly Access, I had posted this old onion article, which says, dolphins develop opposable thumbs. Oh, crap says humanity. And in the story, all the, like, dolphin technology starts washing up on shore. And it's like today, it's like, you know, some cruise. you know, coral acts.

And tomorrow, it's like an abacus made of seashells. And so August went off and typed in Dolphin with opposable thumbs, Coral Abacus, and tweets it back at me. Right? And I'm just completely

freaking out. Again, so I'm like, I can't handle it. Just runs to your face again and again. And so it's like, now I'm so Dali-Pilled. I'll see somebody say something online. And I'll respond to it by, you know, I take 60 seconds. I swipe over to Dolly and Chrome.

I type it a prompt. I wait, you know, 20 seconds. I get back a bunch of results. And I will then. It's going to be the most annoying thing. I do this all the time with GIFs or GIFs, as Brian likes to say. Sure. You know, constantly.

And I got to imagine it's so annoying. You could only imagine if I were able to turn, like, reproduce like jips in retorts that are actually dolly powered. Oh my God. People would be so annoying. Exactly. Well, I mean, you know, this is why my family teases me, because I just get hyper-focused on things.

But it does become, again, it literally changes the way you see the world. Like, to me, one of the most charming aspects is Dali, it's really good at reproducing, like, the layout and visual characteristics of something like type. Like I typed in, you know, Zizi Top concert poster with, you know, 1930s Hot Rod, whatever. And it gave me fantastic, like 10 amazing-looking posters, all of which were complete gibberish. but that's super delightful like my son took some of those terms and I was like writing a short story for class, you know, using this, these insane terms. But, you know, and I'm sure they will fix that at some point, you know, fix in quotes because it'll be like less charming. But it's, it's like watching a kid try to master something, you know, and it's like, what is, what is, what does, where do these come from? Is it like, is that how it thinks about the world? Like when I see, you know, my dog, Like, somebody actually has a really fascinating thread about this, which I think was like maybe partly debunked, but like figuring out, oh, you know, you type in birds and terms, you get out these weird little chunks of language.

But if you then feed them back into the system, it, like, makes you birds again, you know. So not only are people trying to, you know, compose the poetry of good prompts, they're trying to figure out, well, okay, I look at this like hand-lettered, beautiful graffiti. How would the computer describe that in words? And then how would I turn those words back around to make more art? Yeah, one of the things that Ken was doing, and I just pinned it to the space, is producing money from like the Roman era in the style that might have been used back then, right, with crosshatching and those types of styles. And it's quite like, you know, believable. I mean, again, you look at it and you're like hot dog fingers, but nonetheless, it's still believable to the degree where, I mean, it allows you to imagine kind of unreal things that are. plausible. Yeah. Well, let me give you one specific funny example. So there's a guy named

Guido Corroney who works at Adobe. I knew him from Pixar. He was there like 25 years. So he's actually the voice of Guido the Forklift in the Cars movies. And before he knew about Dali, I forwarded him a tweet where somebody had posted images of a Lego Tesla in Manhattan. And I didn't really give him enough context. I mean, I thought I did. And he looked at it. He's like, are great renderings. And I'm like, they're not renderings. They're not models that, like, they don't physically exist. This is something that somebody just told the computer to hallucinate. It did it. And I'm sure, of course, if he looked more closely, he would have noticed something was off. Because, like, you know, the little studs on the Legos are a little

work. But come on, that's like the last, again, details, right? Perception. It's like, what did you expect to see? And you expect to see good-looking Legos? Yeah, literally yesterday, I accidentally fooled a friend because he has a pet, a hedgehog. My son loves Legos. And somebody had posted. They'd actually made like a life-sized Lego Hedgehog had it in an airport. So again, now I want to run everything and see, well, what would Dali make of that? So I go in and type Lego Hedgehog in an airport, which I can tweet out the results, you know, in a minute.

But, and then it was so interesting to see what is its take on that. And instead of making something that approximated an actual Hedgehog just made of Lego, which is what the human had done in real life, it made a bunch of Lego mini-figs, but with hedgehog faces. And then it was trying to give them like roly suitcases. but the suitcases were also hedgehogs slightly. And so you're just like, what is going on down there? Like with the dolphins and the thumps.

Like, what are they doing? And that is it for us, monkeys. Like, we've all seen the writing on the wall. We know it's coming. You know, I just went a Devo interview yesterday. And it's like, just accept the devolution, man. We are getting lapped by our own creations. And you can at least enjoy it, hopefully.

So, yeah. Well, John, I really appreciate you. you coming on. I mean, this is totally last minute, and you've been your joy to bring in the pod. You've been amazing, John, by the way, absolutely amazing. Chris and I were backstage talking, you're like our dream guest. Yes. So, yes, Chris, go ahead, ask him for plugs and things. Yeah, well, I mean, you know, obviously, John, we'd love to have you back anytime as of course the stuff develops or if you guys launch stuff, you know, at Adobe or whatever. Yeah, sure of your research. Is there, you know, where can people find out more about you?

maybe, you know, check out your blog, your website, Twitter, et cetera. Well, thank you, guys. This is, I was, it was so serendipitous. I was literally watching or walking my dog. I saw, you know, Ken's tweets about Batman, and then I tweeted some of the response, and you jumped in and super, super perturitous. So, yeah, it's a lot of fun. And I think this will be a great ongoing conversation because I really think we're just seeing

the start of things. So, yeah, just on Twitter, JNAC, JNAC is my handle. There's a link in my profile. I'll take you to the blog, jenac.com, slash blog. there's a whole category on there of AIML stuff. And so just if you click that, it's funny, people kind of tease me like, oh, you still blog, that's so quaint.

I know. I've been doing it for 17 years. But the nice thing is then you get this, at least searchable archive of interesting stuff. And so, yeah, I mean, check out what, like, VDIA is doing with GANS. Check out, you know, what Google's doing with Imagine. obviously I hope we'll have really exciting stuff to share soon.

And, you know, I love to see what other folks are exploring. And, you know, I keep gently knocking on my grand mentee's door to try to get more folks access. I think she's probably furiously trying to get more machines stood up so they can get more people in. And yeah, I would love to keep the conversation going. Amazing. Well, John, once again, thank you very much. We are going to be putting this recording out tomorrow on the TechMeme podcast. So if you want to share it then,

it'll be available. We are going to actually talk now about another, I don't know, generation defining and also ending of an era. And I suppose I'll tee this up just by sort of reflecting on my own career in technology. And by the way, it's so great to have someone like John come on who's had, you know, a deep career in technology has been through, you know, many evolutions, you know, from the very, very early days when we were building just like the primitives to where this stuff is gone now when you apply kind of, you know, deep learning and AI and all these futuristic technologies to solve human problems and I guess to, you know, get rid of some of the shit work.

And it just, you know, I reflect back to my start in Silicon Valley in 2004, you know, the first thing that I worked on coming out here was the launch of Firefox. And I believe deeply in the mission of what Mozilla was trying to build. I believed in the mission of the web. I believed in removing gatekeepers. I believed in Chris, this is an encomium for. Internet Explorer. And you're only talking about the competitor to Internet Explorer. Well, I am because I'm setting it up as why Internet Explorer was such a big part of the web

and also a big problem for the web and why the demise of Internet Explorer, which is basically what we're talking about, is such a big deal personally for me, right? Oh, this is interesting. So you've come here to bury Internet Explorer, not to praise. Oh, I always, I was, I was coming here to, like, put the history hat on and be like, listen, I'm going to tell you. I'm putting a pin in it. It is done. Like, I am putting up the tombstone. Well, listen, as all, as anyone that's ever tried to design a website for, what, 25 years, we're all happy Internet Explorer is dead.

Okay. Can I interrupt you? Sure. And tell you, can I give you why it's important? And it's almost important. And you know what? Maybe you're making me think of it differently. because it's almost important to talk about the counterfactuals.

Sure, right. Okay, so, you know, I did the story today about how, you know, or was it yesterday? I can never remember. You know, Facebook being more TikTok like or whatever. Look, the beginning of the modern era of, oh, my God, someone's going to eat our lunch. We're going to, we're going to turn all the ships, and we're going to completely pivot to something that was Bill Gates, Microsoft's 1994. Netscape is coming along.

There's a template there of this is how we get religion on this new thing. We're applying all of our resources to it. That's sort of a cadence for a startup, for a tech company. I would say, you know, this is the first modern version of that. Think of a counterfactual of,

let's say, Netscape doesn't get killed by Internet Explorer. Is Mark A. Andresa not one of the biggest VCs in the world. He's the CEO of maybe one of the most important tech companies in the world. Also, think about, I did a whole bunch of episodes on the Internet History podcast with people on the IE team, by the way. But again, bringing the idea of, I think they hired like a thousand people in a month to throw on the IE. Sure. That sounds reasonable.

And they could, and they did. And then I continue to make the argument. Everyone always pushes back on this. But if Microsoft hadn't been so aggressive against I.E., now listen, for a decade, they had been doing anti-competitive stuff, but it was the Internet Explorer versus Netscape thing that caused the antitrust stuff to happen. It was the bundling. They had a monopoly on the operating system with Windows.

Right. Right. And they essentially bundled in the internet browser, which for which there was a competitor in the marketplace. And now they use their dominance in operating systems to distribute a web browser. Now, of course, there was a moment there where the web browser was actually built into Windows Explorer. And so you can type in C colon slash slash to get to your hard drive or you could type

in HTTP and it was just a matter of protocol. And so the web and the desktop were actually merged at one point. And it was almost like the big mistake was splitting out the internet into its own application that caused in some ways

so much problems, so many problems for Microsoft and for interest. And I'm going to bring in tweets from our friend Alex Cantorwitz who can't be with us today because he's down in South America doing some crazy shit.

let's say the antitrust thing doesn't happen does Microsoft buy Amazon does it buy Google does it buy whatever I still maintain that and people always push back on me on that but I still maintain that the whole web 2.0 era doesn't happen if Microsoft is not constrained it doesn't have its hands tied and can't acquire people so this is not really giving i.e. its We should really put this into Dolly and see what it comes up with. Okay, real quick, and then I'll let you end with your thoughts on IE and especially how shitty it was, because it was shitty. But actually, so this is Alex's tweets that I'm reading about. Like, he says, so Microsoft kept Internet Explorer slow and Google and response built Chrome,

why would Microsoft have harmed its own browser? Simply simple, because if people could run programs on a Facebook, fast browser they wouldn't have to use Windows machines. They could do the same stuff on Apple computer, blah, blah, blah, blah, blah. And it's the whole Steve Bomber era of the bots where it's Microsoft's lost decade, protecting the Windows Monopoly, Innovators Dilemma, blah, blah, blah. But this is also, so they put Internet Explorer to a skeleton crew, Villa and Firefox, which I'm going to let you jump in there in a second.

But one more thing that Alex points out is Microsoft only recovered after it invested heavily in the cloud leading to Azure, and the person who led that effort was a middle manager named Sachin Adela. The person who led Google's Chrome project was a product manager named Sundar Pichai. So for all of these sort of like, you know, sliding doors moments of history.

That's why I think Internet Explorer is interesting. But also, all right, I praised it, or I didn't really praise it, bury it. It was a shitty product, right? Well, you certainly contextualized it. And in fact, I mean, Internet Explorer, in the early days, I mean, it was actually like pretty good. At least as far as browsers went, you know, compared to Netscape. Netscape was pretty primitive. Yes.

Listen to my episodes with Ben Slivka and I can't remember all the names, but whatever. It really was the better product. Those thousand people made a better product. At first period, it was a much better product. And even Internet Explorer 5 for the Mac was a great browser. My friend Tom DeCellic worked on that and was one of the leading adopters of web standards. So there were these interesting factions. And the lore of like web technology and web history, I think is super interesting and really

relevant, especially for all the conversations happening in Web 3 today, where standardization is happening through very different mechanisms than kind of this like kind of NATO style or I should say maybe UN style standards body where everyone comes together. They kind of, you know, decide on a set of things that they want in the standard. Everyone agrees to adopt them. They go off and they do their reference implementations and then they kind of like, you know, roughly conform. And it's funny. there was a recent new effort to create a re-standardization or re-implementation of fitting to the standards more aggressively by all the different browser companies. But setting aside, like the leapfrogging that was happening back then and the competition that was happening back there, I think was quite healthy. And then suddenly, you know, whether it was the road ahead, which Bill Gates wrote, you know, about the Internet Super Highway and all the rest, or just seeing that, as you pointed out, like the Internet was going to be disruptive to Microsoft's business.

And especially an internet and a web that was not built in Microsoft technologies, i.e. ActiveX. And Flash was also rising at the time. All these companies wanted to own not just like the means of production, but the mediums of production. And they wanted to essentially define the substrate on which all applications would be built. This is why you see so many different, you know, large tech companies, you know, Apple, Google, etc. kind of inventing their own programming languages in order to advantage their own infrastructure and known platforms and systems. Now at least there is, you know, several large companies that are competing with one another. But back then, you know, Microsoft moved aggressively to stomp out, you know, the threat of Netscape. And that actually imperiled the fate of the open and free web. So the reason why I'm like product aside, it was the strategy that Microsoft was taking on to squelch the open and free internet that, or at least the open and free web, that I have the most problem with. The fact that I, as a high school student, was able to learn HTML and write my first web page and publish it to the web without getting anyone's permission is what made me so dangerous

and also effective. And also, we're having the same conversations right now when it comes to free speech and who should have the privilege to be able to speak freely. And the web kind of removed that question and said, anybody should be able to publish. And then, you know, obviously since then web 2 is kind of more about distributing reach and audience and access. But it was that primary sin that I, that sort of turned me against Microsoft for at least a generation. I do think it's interesting when you consider the people who are behind kind of the overall strategies and who's at the top of Google and Microsoft now. But when I think about the underlying fundamental technologies of enablement of giving other people the permission to build and create and to publish

and to share, Microsoft seemed to be very much about owning everything and everything reporting back to Redmond as opposed to being kind of part of a global community that was seeking more people being able to contribute. What they wanted is, if you believe the conspiracy theories about Web 3, what they wanted was they wanted to take a Vig on everything. And the original said to them of the internet was they couldn't take a percentage of everything that happened. I mean, the fact that like Office, you know, kind of is Microsoft's Cash Cow is because it's sort of necessary in order to open their files. And if you remember Open Office was meant to be a hedge against that tax being deployed on all, you know, upcoming nations and countries and people

who couldn't afford the Microsoft big. And, you know, Facebook has gone the same route. Like, I mean, this is a very common pattern in software development, but Microsoft was kind of like the first, you know, evil empire of the, you know, internet era, I guess, of the software era. Yeah. Well, before them, it was IBM, as Steve Jobs and Bill Gates both would tell you. And I don't know. Before that was... I mean, it goes all the way back, right? Like, I mean, every hacker, you know, kind of like sees themselves in this, like, heroic stance.

And then, like, they become, you know, the thing that they sought to replace. It's just, you know, struddles all the way down. What was the version of I.E. that even when, you know, we were designing websites for our companies in 2003, 2005, that we had to... You had to have a separate version of your website for. was it i.e4 or i.e6 or something like that? I was going to say I think I was the real, that was the one that probably came with Windows 98. We would literally do the website. You have all these weird CSS kind of like, you know, hiccups in terms of how they like implemented it.

Oh, man. And there was so much malware. And, you know, it was just a whole era that people who grew up on the mobile internet will never have to deal with. The fact that IE would crash all the time. And like you said, it was slow. It would deliver malware. Yeah, but listen. You might have, there was a time when you created a separate mobile version and a separate,

but we had to create a separate version for just one version of one web browser because it was too important. Because it's so much. And the crazy thing is, and I think this will be like the last point on this. You know, you did a story about this that, you know, apparently there are still, what is it, like? In Japan. Yeah, financial services that in Japan that still require Internet Explorer.

in order to access those services. Like, I can't even imagine or understand or comprehend what technologies that brow, it's like COBOL. It's like, what technologies is that browser offering that those banking services need to use that they can't migrate to some modern web browser? You know what I mean? Like it breaks my brain.

Like all the other security issues that are rife in that browser, but yet they still have to use it for this other purpose. I just don't understand it. The internet history. It shows how deeply embedded it is into, you know, people's livelihood still. A lot of this did not make it into my book. But if you're interested in the history of Internet Explorer,

I recommend the Hottie Partovi episode of the Internet History podcast. Chris, let's go. We've been going an hour and a half. No, this is great. You know, we covered a lot of stuff all over the place. Yeah. I love it. So anyways, go ahead. If you want to do your closing remark.

Oh, oh, is this my new thing? This is my catchphrase. I love everybody. Or it said, I love you all. What is it? Which do you prefer. I love everybody. It kind of has a nicer, you know, inclusivity to it. Yeah, but if...

I love you all. I think that's good. Yeah, go with that. I love you all because that means I love the people listening. I love you all. Later, everybody.