Big Technology Podcast - Google’s Deceptive Gemini Release, More Tech Layoffs, George Santos Saves Cameo

Episode Date: December 8, 2023

Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover: 1) Google releases its Gemini large language model 2) Why Gemini is a win for Google 3) Google’s deceptive Gemini marketing campaign 4) Why Google felt compelled to exaggerate 5) Which data sets add value to LLMs after the commodify 6) Spotify lays off 1,500 6) The economy is still on shaky footing, especially for companies with prolonged covid hangovers 7) George Santos may just save Cameo ---- Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. For weekly updates on the show, sign up for the pod newsletter on LinkedIn: https://www.linkedin.com/newsletters/6901970121829801984/ Questions? Feedback? Write to: [e-mail]

Google releases its long-awaited Gemini large-language model, but it used some deceptive marketing in the introduction. Where's the real value once these models are commoditized? Spotify lays off 1,500, and George Santos may just save Cameo. All that and more coming up right after this. Welcome to Big Technology Podcast Friday edition where we break down the news in our traditional cool-headed and nuanced format. We have a big week of news this week.

Touching on Gemini, Gemini and plenty of other interesting stories. We're joined again by Ron John Roy of Margins. Ron John, welcome back. Maybe George Santos and Cameo is more interesting than Gemini, but I think there's a lot to talk about Gemini as well. That's right. Okay, so let's start with Gemini. The first thing we should just do is talk about what this model is.

I think that you have some thoughts on it. You contributed to the panel on big technology.com this week. Some really interesting thoughts. But basically it's Google's answer to GPT4. It's a large language model. It's multimodal, which means that it can understand text and images. And we thought video and voice, but okay, we have some questions about that, which we'll get to in our next segment. And it also seems to outperform GPT's OpenAI's chat GPT model in a number of different cases.

What's your general read on the introduction of Gemini? I guess, like, my question about it is, you know, has Google caught up to GPT4? but I've gotten some pushback on that this week. What do you think? All right. So we will definitely get into this introduction video. But I think this is a big moment for Google. This is what I've written about for the panel newsletter. I believe strongly that Gemini is their first really core offering that's clear to customers that this is the main model and this is what you can use to do the business type operations or build on in the same way you can easily have built on GPT 3, 3.5, and 4.

And I think Google this actually, they have an advantage where four businesses already operating on Google Cloud to be able to build on a kind of foundation model like this, build easily, connect to their existing data infrastructure, it presents a big advantage. I think already OpenAI has that relationship with Microsoft and every Azure client. So now any Google Cloud client can build on this. I do think one thing as they release this that I was thinking about is we all can remember the name Gemini. Do you know the names of their previous iterations of these large language models? Yeah, I think you wrote that in your blurb. It's like what, like fourplex or some

stupid name like that? So in classic Google fashion, they have vertex AI is like the overall ecosystem, but then Palm 2 was the actual large language model equivalent. Then under that, they had these things like text gecko, text, but they had all of these different names that were very confusing. And then Vertex AI would live differently, whether it's in Google Cloud or Google Workspace. Like, everything was incredibly confusing. I hope they have it together where now Gemini is the model, Gemini Ultra, Gemini Pro, Gemini Nano. And they don't G-chat this whole thing and turn it into Hangouts, chat, meet, whatever else. And just make a clear offering for customers who want to use it to be.

be able to use it. Why don't they just call it Bard? Like, why don't they call everything Bard? Because Bard is a terrible name. I think they should name barred Gemini then. Yeah, yeah. Actually, I, if I was Google right now and if they, I actually thought that exactly same thing, Gemini is kind of a cool name. It's a great name. I would be happy to Gemini something or ask Gemini, but I don't want to barred anything. Exactly. Yeah. Okay. If they're listening, change Bard to Gemini. It's not too late. It'll make a huge difference. Everyone will be happy. Let me ask you this, though. So you say the main advantage here is that they have like a GPT4 like model that they can use for Google Cloud customers. But Google Cloud is number

three. And if you look at what Open AI is doing with Azure, which is number two, and what Anthropic might be doing with Amazon, bring the own model type of approach, number one, like you are actually maybe at a severe disadvantage if you're Google, because you now have a preference in terms of cloud platform with Gemini and you're effectively like, I don't know, seeding or you're allowing other models to run away with it on bigger cloud platforms. No, but I believe we're entering a new era for cloud services and this is why it could be a key for Google Cloud. In one of these companies is going to build the best overall cohesive offering. One of these companies, and it doesn't exist yet because like open AI still is relatively separate from Azure. It's somewhat integrated

and anthropic is even more disconnected from Amazon. One of these companies is going to build the best overall offering where customers feel secure, can build on it easily, can integrate it into other parts of their business where these cloud offerings live. Like one of these companies is going to figure it out. So I actually think even though Google's number three, if the fact that that it's in-house, it gives them an advantage here. The fact that they're building it, it's their model, it's not a partnership. I think the way a lot of this stuff's going to go is it's less about power and it's more about ease. Now, Runjan, I'm sure you've read the coverage of this and it's been kind of interesting, you know, going back to our Bard or Gemini points,

because half of the press is like this is its answer to chat GPT and then half of the press is like now it has a large language model it can offer similar to GPT4. And, of course, like, this is obviously more aimed at the enterprise to build on top of than it is necessarily with the chatbot. And it kind of puts an interesting question about Google into the field here, which is that, like, they have to do something on the enterprise side, but also are they accelerating the death of their consumer business in search? What do you think? Yeah, that's it. That's the right question. And I think they're going to have to do something.

And Bard slash Gemini chatbot is going to be part of that. experience search generative experience SGE, which we've talked about plenty of times where now you start to see generative results in your regular Google searches, it's going to change the entire way we search. So I think on the consumer side, they do have to do something, and I think this is it. And they said that Gemini is going to power both Bard and the SGE search generative experience. So it's clear that they're working on it and they're approaching it that way. But yeah, I think overall that they are, They have opportunity in both the consumer and the enterprise, and if they roll this out properly, it puts them on not necessarily an even footing just yet, but at least puts them back in the game in a way they have not been for a while. So there's one more thing I want to talk to you about, which is the timing of this. It was very interesting because the information earlier in the week had a report that Google had scheduled a launch event for Gemini and then decided to push it to next year. and then all of a sudden there was another report that said, hey, actually Gemini might be out soon. And it almost seemed like Google was trying to capitalize on the chaos at OpenAI and then decided, okay, let's wait until this is ready because there are parts of it, including Ultra, which is their most powerful model that aren't going to be ready until next year. And then for whatever reason said, F it, let's just put our foot on the gas and release it.

I mean, what do you make of like the back and forth here? Yeah, the rollout, and we can definitely get into that video, and we will. It's suspicious. I agree that already in this entire cycle over the last year, Google has repeatedly made big demos and flashy demos and then not had a product to back it up. Whereas Open AI, every time they announce something, it's there. It's ready. And that's part of been what's made them feel more powerful that they talk about something and then people can actually use it. So it does feel a bit rushed in the sense that. that they debuted all this stuff.

The most powerful model still isn't available other than to maybe a very limited set of developers. So, you know, getting ahead of that year-end, I hadn't thought about the capitalizing on the open AI chaos, but that actually could be, now that I'm thinking about it, that could be the right answer here because it's true that, again, on the enterprise sales side, people at big clients have to be feeling a little uncertain

ahead of year-end renewal talks and contract talks, which for the most part are probably done, but there still could be ones ongoing, or as those talks would start at the beginning of next year for Q1 to look at the year ahead, positioning themselves a little bit stronger. Because, again, no one talked about text bison, but we're all talking about Gemini. So something's working. Yeah, there was something weird about it because I almost missed the introduction of this. So what they usually do is they have these big release events where they gather people in, you know, together or they at least give you like a good amount of notice if they're going to do something online. And for this, I think I got noticed the day before that they wanted to do some sort of embargoed briefing.

And I was like, oh, like I actually missed the first briefing because I was like, I don't know what this could be. And I was like, maybe that was Gemini. And I wrote back and they were doing one at 830 at night in New York time. And I was like, all right, like, I'll watch this. And I was like, oh, this is Gemini. And it was like, huh, that is interesting. Like it does seem like they rusted out the door. And then let me tell you a little bit about what I saw and what the world saw in terms of the way that they were introducing this.

So they had this video that's now, you know, been infamously called the duck video where they effectively like did all these different really interesting interactions basically in live video asking, you know, this mysterious Gemini like what they're drawing and which car was more aerodynamic. and how a roller coaster would work and all these and to take two different balls of yarn and give ideas of it. And it was all in one shot of a video. It was jaw dropping stuff. I mean, definitely like I was like, holy crap, this is advancing the, you know, the state of the art. And then the next day it comes out that this marketing video was quite deceptive. So first of all, it showed a user speaking with Gemini, but the whole presentation, I mean, the whole interaction was entirely a text conversation that they sort of dramatized as someone speaking to it. Then it showed a moving video, but it was actually just feeding Gemini steel images. It showed fast responses,

but those were actually sped up. It showed tight model outputs, basically the answers from Gemini were quite tight and precise, but it had shortened those for brevity. And it also showed a very brief prompts that were eliciting terrific answers. But then it listed longer prompts in a blog post depicting what was actually happening in this video. Now, look, later, it clarified that the shorter prompts would, you know, would actually work. But I also asked them, like, then, why did you list the longer prompts in the video? And I, and in the blog post, and I still haven't gotten an answer on that front. So overall, like, this video backfired on them.

It was viewed as extremely deceptive. It made the rounds on Y Combinator, like, basically was, like, a front page of hacker news the entire day. And I wrote about this, some big technology and included, like, just one of the more benign quotes from one of the developers on hacker news who said, I was fooled. It's one thing to release a hype video with what ifs and quite another to claim that your new multimodal model is the king of the hill and then game all the benchmarks and fake all the demos. Now, it wasn't entirely fake, but it's certainly like rubbed a lot of developers the wrong way and it's actually become the story, the fact that like the marketing was deceptive as opposed to the actual

model that they released. So I'm curious what you make of this. I mean, to me, it seems like this is Google kind of operating from a place of weakness where they feel like they have to produce something revolutionary, even though they only had something evolutionary. They portrayed it as if they had that revolutionary advance and just totally got busted on it. And I'm curious what your perspective is. It infuriated me. This launch video infuriated me because, again, and you had written this as well, Google really didn't have to exaggerate. Like, they didn't. I'm so confused, again, as Alex was describing, like, there's one part of the video,

it looked really cool. There's, you lay out a world map, and then you say, make a game for me, Gemini. And then it's like, point your finger to a country that has won the most World Cups. And, like, you know, there's really cool ideas about imagine all the things you can do. But none of that matters right now. The multimodality, it's not something that. those use cases have even been invented at scale yet. Like, there's so many just straightforward business problems that customers want to pay for

and know they have a reliable service for. There's just text alone, text to image, all these things could sell and be great. It literally was just, it's pretty much as good as GPT4. Like, that's all they would have to say, but it's from us. It's easier to access. it's free, all these kind of things. So I was so confused as to why they felt the need to try to do something that flashy.

The one thing I will give them that maybe is hopeful for them is I did see friends who are not that into technology sharing that video and being like, this looks pretty cool. So whether that conversation about what was deceptive or not makes it over to the normie world, I think will be interesting

or whether the Gemini brand still look as thought of as cool to the average consumer. But I was completely confused as to why they would try to do something like this when they absolutely do not need to. No, I love when you look at marketing and you try to make your best guess as to what the heck is going on inside the company behind some of this stuff. So why don't we put the conspiracy hat or the speculation hat on? And Ranjan, from your perspective, what the heck has to be happening there? My guess is that want for virality and kind of like consumers all talking about your product was at the core of it.

And I don't think it's necessarily the wrong thing to want because as we've talked about a lot, because of the chat GPT consumer familiarity and like that because it's almost the verb to look up. something via generative text. They want something like that. They feel they need to. That stuff is going to translate both on the consumer and the enterprise side. So it's not unreasonable to me, but you can totally see how they're like, we need something viral, really cool. And then they figured let's stretch things a little. The prompts are a little too long. So let's shorten it. Let's dramatize it a bit. And here we are. Yeah. What do you think? It doesn't. It doesn't see. It doesn't seem like people are buying their explanation. I mean, they had one of their vice president's Oriole vignales go out and talk about this.

And this person said all the user prompts and outputs in the video are real. Shortened for brevity. And like, everybody was like, wait, what? You can't be both. It is interesting that they had, because he's like the deep mind vice president or something like that. Yeah, like, it's funny that I wonder if it was more engineering-driven than marketing-driven or more marketing-driven than engineering-driven that they had him speaking. But, yeah, actually, that would be, do you think this was, do you think that demo was pushed

by the engineering side of the organization or the marketing side? Because my initial gut feeling would have been the marketing side, but the more I think about it, maybe the engineers want a bit of credibility and some people talking about, how powerful their models are. So I think it was both. And my perspective on this is that Google's pressing a little bit. And sorry, I'm going to have to use a sports analogy here. But, you know, the Jets, who I watch a lot, unfortunately.

We don't talk about this often, Jets fan, Patriots fan, but we managed to get along. But we're playing the same game right now. The Jets can't score points on offense. And they have this great running back, Brise Hall, who's trying to score a touchdown down every time he gets the ball and he ends up getting tackled in the backfield because he thinks he has to do so much in order to, you know, make the thing run. And that's kind of what I think

Google is doing here, which is that they feel they're behind. They feel they have to go for those home run hits as opposed to get, you know, five or six yards and put themselves in good position to score. And that's kind of what happened here. Just an obvious case of pressing from someone who's behind, from a company that's behind. Actually, but if you think about it, I always go back to, do you remember the Google Duplex demo when Sundar called a restaurant and they made a reservation all over the phone? So this isn't their first kind of playing from behind, rushing things out a bit, forcing things a little too hard. So maybe it's more ingrained. But I agree, though. They are playing from behind right now. So I agree. You can feel it a little bit. I don't

want to say desperation, but at least additional energy, let's call it. And I had someone basically put it to me as this, like everybody thinks that they're behind. And so to exaggerate in their marketing is the reason why it's bad. If they were open AI and they did something that was a little bit embellishing, people would be like, all right, but their products are cool. But in this way, this is really just egg on the face. Like the actual, the cream filling has to measure up, you know, we don't have the cream filling yet.

No, but they don't even need it right now. I wish, just get that, find that confidence, Google. Find that confidence. So right now, with the release of Gemini, what we're seeing is that a lot of these models are starting to look the same. You have GPT4, you have Gemini, you have Anthropic, and it's moving towards what we've been talking about for a while here, which is that these models are going to start to commoditize and look the same. And they will start to differentiate based off of things that you said, ease of use. and then also probably the data sets that are ingrained within them.

So this is a tweet that you and I have been kind of pushing back and forth and chatting about. But Chimath Polyopatia talked about how basically he says this is a race to own data that is valuable in fine-tuning and reinforcement learning with human feedback. You should expect foundational models to be largely commoditized and largely free, largely soon. And the non-obvious data sets are going to be in huge value in the end state. So he lists a number of them. One is Tesla's real world environmental data. Two is met as real people preference and behavioral data. Three is Google's search intent, YouTube behavior, and Gmail. Four is Amazon's purchasing behavior. I'd add a fifth, and that's real-time information

and news. Basically, that if all these models start looking the same, those that can actually take and exploit proprietary information are going to be the ones that have a leg up. I'm curious if you think that's really the next step here. Is that where we're heading? Yeah. And we were definitely talking about this because it's rarely that I agree with Chimoth, but I think he's totally right here that. And we've been talking about this for a while ourselves, that foundational models are going to be largely commoditized and free. That's, again, even not necessarily always free, but integrated into existing ecosystems, but the actual power of the foundation model is not as what's as relevant. But I do think the difference here is

foundation models to be used by a large group of potential clients or use cases, or foundation models used by the companies themselves to release new products. And that's where Tesla could have a self-driving advantage because of their data set. Meta certainly and probably already is using their data advantage to build out their advertising ecosystem. I mean, their advertising is back and then some. So Amazon obviously reinforcing their own business, like, even for me, like I still, I'm still convinced Reddit has one of the most valuable content data sets in the world, and they actually cut off access to the API because they're clearly going to be doing something interesting. So I think each company looking at their own data set and thinking about how do we use

this in a generative way is going to be a very, very big thing. But again, it doesn't translate to like sales of the model. It translates to new products and experiences of your own business. Absolutely. All right. Let's take a quick break. We have two more stories we want to cover, the layoffs at Spotify, broader layoffs in the media, and then the George Santos, cameo story, which might be my favorite story of the year. We'll be back right after this. Hey, everyone, let me tell you about The Hustle Daily Show, a podcast filled with business, tech news, and original stories to keep you in the loop on what's trending. More than two million professionals read The Hustle's daily email for its irreverent and informative

takes on business and tech news. Now, they have a daily podcast called The Hustle Daily show, where their team of writers break down the biggest business headlines in 15 minutes or lists and explain why you should care about them. So search for the Hustle Daily Show and your favorite podcast app like the one you're using right now. And we're back here on big technology podcast with Ranjan Roy of margins. Ron John, it's been a really rough couple of weeks in the media business. Spotify in particular is about to lay off or is in the process of laying off 1,500 people, but we also have layoffs at Wired and Vox and the Washington Post seems to be laying off. I mean, And it's wild how many companies are cutting now.

And the thing that really caught me by surprise about with Spotify was the fact that the business is doing great. It's profitable. It is on pace to add more than 100 million users this year, which is going to be its best year yet. And but it's apparently still spending too much. And the CEO, Daniel Lek, is talking about an economic slowdown and the rising cost of capital. Are these just like the push down the road impacts of what happened when rates were raised? I mean, why is, why does it seem like the economy is going well? Like we were talking about a potential of a soft landing and yet all these companies are still cutting.

It's still the COVID hangover. And it's a combination of rates are no longer a zero interest world. And I do love that even the Spotify CEO now cites the rising costs of capital, i.e., no more preserve in his own layoff announcements. I never thought that day would come. But then that combined with Spotify doubled its headcount during the pandemic. And now it's still well, well above where it was pre-pandemic. So this goes back to, and you've written about this plenty, the extrapolation of COVID trends and that not being realized. And we're seeing it. And honestly, there's probably room to go in a lot of these companies that just when you double your headcount in the space of

a year or two and then realize that you're not going to grow in the way you thought you were there's probably a lot of people that are not going to be there for much longer and do you think that this is going to be constrained to our corner of the world like the media and tech world or is this going to be broader in the economy i still i think we're seeing more now and again there's plenty of indication and talk around a soft landing in the economy overall that that this is going to be still constrained to companies that grew too much during the pandemic and overinflated their own expectations around where they would go and what they would do. Because again, streaming, they had a profitable quarter.

They're not necessarily doing amazing. They finally saw a bit of profit. But again, you can imagine a business like Spotify, you always do one. like how much operational overhead really is there to run a software product like that, obviously licensing and marketing and all these other things. But overall, you know, how do you double your headcount? What are you going to do with that? So I think in technology overall, if like, if there's definitely the more aggressive CEOs become and also the more positively capital markets respond. And so,

other CEOs see that, I think we'll probably see more of this among companies like this. But I guess, like, my question is, wasn't this supposed to be over? Like, we are, we have a story in the, in our, you know, doc that we use to, like, talk about what stories we find interesting this week about how the economy just added, 190,000 jobs in November, the employment, unemployment rate fell. It looks like we're going to leave this year without a recession. It seems like, I mean, You said more to come, but like it also seems like shouldn't we just be done with this already? I mean, but I think, again, that's why it's even more because the jobs being added. I think I saw services jobs continue to be like at the core of a lot of our job growth versus Spotify and software engineer or product marketer or consumer marketer at Spotify.

So I think these businesses realize like they if you're Spotify right now, you're back to looking at. let's say a three to five year plan, and you're thinking about that strategically, and you're trying to figure out who do I need to get where I want to go and what's the world going to look like? And you realize that the company you built over the last two years isn't that company. I think there's going to be a lot of companies like that. And it's pretty independent from the overall macro environment, I think. I think that's both comforting and also extremely scary because so many companies built up during the pandemic. Well, one company that's in the middle of a run-up is George Santos, Inc.

If you're not in the U.S. and you're listening, we had this guy, he's a congressman from my home island of Long Island, who lied his way into Congress, made up a ton of stuff about himself, used his campaign funding to do things like buy handbags and Botox and all this stuff. And he was finally expelled from Congress. He's like the six guy, six Congress, member of Congress to be expelled in its history. And like the other reasons are like treason and other terrible things. And he got the boot for buying Botox.

but he's making this amazing second career now as a cameo performer and I think that what he said and has been confirmed by some reports is he's made more money in one day of cameo than he made like in the year as a member of Congress which is just unbelieving how many thousands of cameos has he done but my thesis here is that George Santos is going to save cameo and make sure that this crucial part of tech and public culture infrastructure remains intact. I know we've had cameos hit some hard times, but I'd imagine this is something that you and I are going to agree on, right, Ron John? I think George Santos will save cameo. I will say, hold on. Let me let me just play

a clip here. Hey, he, George Santos here. I'm so proud of you for coming out as a furry and I just wanted to tell you that your friends and family all accept you. And they're all excited about your persona, which is awesome to be a bevipus, a beaver and a platterpuss. So let me tell you, they all love you, bevapus. Don't you ever get your head down. And don't you ever, ever let anybody tell you what you can and can't be. I'm so proud that the corporate folks at Arby's gave you to go ahead to go to work in your persona. So if you could just, you know, live it up and be as perfect as you want. Just keep doing you and yif, if, yif, yif, bye.

Okay, Ron John, yes. Go ahead. Okay. So listening to that, there's many feelings. around the state of our country and democracy and future and the movie idiocry because this is the state of our politics what what was even more traumatizing is it was like senator John and sitting US senator bought one of the cameos and tweeted it and then played it against his you know George Menendez like the fact that our in senior political establishment is playing along with this game just for the impression

and the tweets and, like, is a bit terrifying, but it's good content. George Santos makes great content. I cannot ever disagree with that. I absolutely love these cameos. They're terrific. And, you know, this is clearly American political, the political life hit rock bottom. But maybe this is what gets us out of it. And for Cameo, it couldn't have been better timed.

We were just talking, I think in July, about how this company, which by the the way, allows you to buy video messages from celebrities, largely B&C list celebrities. It was valued up more than a billion in 2021. This is from TechCrunch, but it just laid off 80 workers and it's currently operating with fewer than 50 employees, a shift from the nearly 400 it had when business was booming. I mean, I think we have, you know, going back to our theme, maybe there's two, you know, three eras of the modern company. It's ZERP. Post Zerp and George Santos. And cameo is clearly the George Santos.

George Santos is creating jobs. George Santos is an American hero. He is going to create jobs and be the backbone of the entire technology industry to take us into this third act. I was about to say that who says Congress can't be a job creator. But it is a little bit ironic that once Santos leaves is when he starts having his greatest impact. Wait, have you ever ordered a cameo? Yeah, I think I did. I did once.

Who did you do? I ordered a cameo from like, you know, one of like the back of the roster in New York Giants players to say something to my dad or something like that. I have the most rent. Do you know the band Guar? No. It's like a metal art band. They all like dressing these crazy costumes.

My brother-in-law is a big metal head and I'd gotten him a Guar guitarist birthday message when shortly after cameo launched and I'd read about it. and he enjoyed it. It's a good... It's a cool product. We needed George Santos to come in and save it. That's what I'm saying. This is an important service.

If you think about it, though, to get into what cameo represents about the modern economy, it really is. It makes sense. It's a great product. Great idea. Cela celebrities, get some extra cash. And it totally was. The pandemic made us think that everyone is going to be a content creator selling their time and quick videos for a little bit of to cash. And that's not the case. We're back to expelled members of Congress, being the only one

among C-list Giants players, probably Patriots quarterbacks after Tom Brady. But yeah, they represent. It's a good business, raise too much money, and now it's right-sizing and finding its way thanks to George Santos. So thank you, George. Amen. And let's leave it at that. Ron General, great speaking with you as always. Thanks for being here. Good to see you, man. Good to see you, too. All right, everybody, thank you for listening. We'll be back on Wednesday with an interview with Intel CEO, Pat Gelsinger, about how the chip manufacturer is attempting to ride the AI wave. I'm about to record it.

I'm stoked for it. It's going to be a fun interview. So make sure to tune in next Wednesday. And then we'll be back here next Friday talking about the week's news. So thanks for listening. And we'll see you next time on Big Technology Podcast. Don't know.