Tech Brew Ride Home - Thu. 03/16 – We’re A Long Way From Clippy

Episode Date: March 16, 2023

For the first time ever, I had to delay producing this show so I could watch and cover and AI product announcement. Let me tell you about how Microsoft is bringing GPT-4, to basically all their most famous apps with Copilot. Looks like the potential for a TikTok ban is getting really real. And why folks are suddenly turning on OpenAI. Sponsors: Overtired Podcast Links: Introducing Microsoft 365 Copilot — your copilot for work (Microsoft) Microsoft 365 gets a host of new AI-powered features (TechCrunch) U.S. Threatens Ban if TikTok’s Chinese Owners Don’t Sell Stakes (WSJ) OpenAI co-founder on company’s past approach to openly sharing research: ‘We were wrong’ (TheVerge) Payments giant Stripe raises $6.5 billion at a $50 billion valuation (Axios) 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. Welcome to the TechMeme ride home for Thursday, March 16th, 2020. I'm Brian McCullough today. For the first time ever, I had to delay producing this show so I could watch and cover an AI product announcement.

Let me tell you about how Microsoft is bringing GPT4 to basically all their most famous apps with co-pilot. Also, looks like the potential for a TikTok ban is getting really real and why folks are suddenly turning on OpenAI. Here's what you miss today in the world of tech. If you've been skeptical about all of this AI hype because you're not quite sure how it could potentially affect your life, I encourage you to watch some of the video of Microsoft's announcement today because it'll give you some really tangible use cases. Link is at the top of the show notes. Like literally, there were use cases for basically anybody. Imagine going through your email inbox to triage emails by having an assistant organize the most important ones for you to

respond to, and more crucially, quickly crafting responses on your behalf with the click of a mouse, like superhuman on steroids, or imagine taking an Excel spreadsheet and turning it into graphs with the click of a mouse, or imagine preparing for a meeting by having co-pilot search your emails and files and then giving you a bullet point list of what the agenda is and even what you might want to contribute to even say in the meeting. There were tons of examples of this new co-pilot tool being like a personal assistant, like for being a thought starter or for making smart summaries of stuff. They even showed a live video meeting with like a team on Microsoft teams of like six people having a discussion about something. And in the side panel, you could be like,

co-pilot, what have we talked about so far? And it would tell you even co-pilot, what is the mood of the team? Hmm, the mood of the team seems tentative and unsure, Brian. We're a long way from clipy here, people. Let me give you a quick summary of what was actually announced, and then I'll come back with some thoughts of my own, quoting TechCrunch. Currently in testing with select commercial customers, co-pilot combines the power of AI models, including OpenAIs recently announced GPT4, with business data and Microsoft 365 apps like Word, Excel, PowerPoint, Outlook, and Teams. Today marks the next major step in the evolution of how we interact with computing, which will fundamentally change the way we work and unlock a new wave of productivity growth, Microsoft's

CEO Sachin Adela said in a statement. With our new copilot for work, we're giving people more agency and making technology more accessible through the most universal interface, natural language, end quote. Co-pilot handles different tasks depending on the app in which it's used. For example, in Word, copilot writes, edits, summarizes, and generates texts, while in PowerPoint and Excel, co-pilot turns natural language commands into design presentations and data visualizations. The PowerPoint capabilities are particularly nifty. Using copilot users can create a presentation based on a Word document, complete with a slick deck and table of contents. They can then refine that deck by asking Copilot to, for instance, add animations to this slide

or apply a modern style to the presentation. In Outlook, Copilot can help synthesize and manage inboxes. Meanwhile, in Teams, Copilot provides real-time summaries and action items in the context of conversations. One of the more intriguing elements of Copilot is business chat, which brings together data from across documents, presentations, email, calendar, notes and contacts to help summarize chats, write emails, find key dates, or even write a plan based on other project files. With prompts like, tell my team how we updated the product strategy, business chat will generate a status update based on the morning's meetings, emails, and chat threads. In a blog post, Microsoft stressed that the models driving co-pilot aren't trained

on customer content or individual prompts. Specifics on pricing and licensing will be shared soon, it said, end quote. Yes, and they also took great pains to stress, over and over again. For all these features, they wanted you to know you're in control. This thing saves you time and effort, yes, and yes, it can make a whole slide deck for you all at once if you give it the right information, but you can always go in and correct errors or edit slides, edit bullet points, edit individual words. Again, they're stressing this as personal assistant computing. And this is why I think people are so excited about all this. A lot of people are coming to realize this might be an entirely new era of computing, sort of like Sacha just said.

This is the computing we were always promised, the computer from Star Trek the next generation. You know, Picard saying, computer, enhance, computer, plot a course. There once was a time when to get a computer to do anything, you had to basically program it yourself in real time. Then that was abstracted away by icons and clicking and dragging and dropping and all that good stuff. But as they made the point in this presentation, you still basically have to do everything by hand. And the tools to do that are your hand right now, literally moving the mouse around, to draw graphs, hunting through the menu system to find the tools you need. Sure, Photoshop was a revolution. It allowed you to create a picture of a tree in ways that

you couldn't before, but you still had to know how to work Photoshop. You still had to learn to use the tools, the menu system, everything. Microsoft said in this presentation that people generally only use 10% of the features available in a complex app like PowerPoint or Excel. Now, with this sort of stuff, the potential is you can just say, computer, give me a picture of a tree. No, that's not good enough. Try again. No, more like a mighty oak. Okay, that's good. Now more sunlight. And you don't have to know any of the underlying manipulations you need to get it to be exactly the way you want. you can just focus on the output itself. It's computer enhance, computer enhance, computer enhance all the way down. If I were superhuman, I'd be worried. If I were Zoom, I'd be worried. If I were Salesforce, I'd be worried. If I were Google, I'd be very worried, even more so than the threat to my search engine.

Shoot, Apple should be worried too. Watch that video and see all the cool things potentially you can do just with email. Again, we'll have to wait and see if this pans out in real life. But what if you could only get all of the best of that if you're on Outlook. Potentially, if these systems work tightly in a closed loop, might these bells and whistles be enough to entice you away from Gmail, from your Mac, even? Sure, I'm sure Google will have an answer for this, but it's like that Varietals theory that I mentioned from a few days ago. What if you like the Microsoft AI Variable better? Heck, what if the vaunted privacy and security and lockdown nature of iOS meant that iPhone users were suddenly locked out of this sort of cool new compute revolution.

Not saying that any of this will happen, not sure that any of this is even likely or possible, but just think of those last few examples of the way computing is going and how it could upend the game for everyone across the tech landscape, and you'll see why everyone is suddenly so very excited and so very scared by all of this at the exact same time. I told you I've been hearing rumblings,

but now this sounds like we're getting close to it being real, real. The Wall Street Journal is reporting that the Biden administration and Sipheus are demanding that TikTok's Chinese owners sell their stakes in the company or face a possible ban of the app in the United States. Quote, the Committee on Foreign Investment in the U.S. or Sipheus, a multi-agency federal task force that oversees national security risks in cross-border investments, made the sale demand recently, the people said. TikTok executives have said that 60% of ByteDance shares are owned by global investors, 20% by employees, and 20% by its founders,

though the founder's shares carry outsized voting rights, as is common with tech companies. The company was founded in Beijing in 2012 by Zhang Yiming, Bight Dance Chief Executive Liang Rubo, and others. TikTok said Wednesday that a forced sale wouldn't address the perceived security risk. It has pledged to spend $1.5 billion on a program to safeguard U.S. user data and content from Chinese government access or influence. If protecting national security is the objective, divestment doesn't solve the problem. A change in ownership would not impose any new restrictions on data flows or access. TikTok spokeswoman Brooke Oberwetter said in a statement, the best way to address concerns about national security is with the transparent U.S.-based

protection of U.S. user data and systems with robust third-party monitoring, vetting, and verification, which we are already implementing, Ms. Oberwetter said. The negotiations with Sifias over a way to secure TikTok's data have been going on for more than two years and have been at a stalemate for months. The Wall Street Journal has reported previously with Pentagon and Justice Department representatives on the panel, among those, supporting a forced sale. Deputy Attorney General Lisa Monaco and other senior U.S. officials have repeatedly cited China's national security law, which requires companies there to turn over customer data if requested as fueling their concerns. Quote, our intelligence community has been very clear about China's efforts and intention to more. mold the use of this technology using data in a worldview that is completely inconsistent with our own,

Ms. Monaco said in an interview last month in response to a question about TikTok, end quote. But of course, we've got to go back to the AI stuff. Among the very interesting things about all this AI stuff is how quickly it's moving all the sudden, including how quickly, suddenly, open AI has begun to be seen as the 800-pound gorilla or maybe the death star in this whole situation. All across the internet yesterday, there was criticism of Open AI for not disclosing GPT4's training data or methods used to develop the new system. In fact, OpenAI's co-founder says its past approach to openly share its research was wrong. Quoting the verge, OpenAI has shared plenty of benchmark and test results for GPT4, as well as some intriguing demos,

but has offered essentially no information on the data used to train the system, its energy costs, or the specific hardware or methods used to create it. Many in the AI community have criticized this decision, noting that it undermines the company's founding ethos as a research organization and makes it harder for others to replicate its work. Perhaps more significantly, some say it also makes it difficult to develop safeguards against the sort of threats posed by AI systems like GPT4, with these complaints coming at a time of increasing tension and rapid progress in the AI world. I think we can call it shut on open AI. The 98-page paper introducing GPT4 proudly declares that they're disclosing nothing about the contents of their training set, tweeted Ben Schmidt, VP of Information Design at

Nomic AI, in a thread on the topic. Here, Schmidt is referring to a section in the GPT4 technical report that reads as follows. Given both the competitive landscape and the safety implications of large-scale models like GPT4, this report contains no further details about the architecture, including model size, hardware, training compute, data set construction, training method, or similar. Speaking to the verge in an interview, Ilya Sutskever, open AIs chief scientist and co-founder expanded on this point. Suskever said OpenAI's reasons for not sharing more information about GPT4, fear of competition and fears over safety were self-evident.

On the competitive landscape front, it's competitive out there, said Suskever. GPT4 is not easy to develop. It took pretty much all of open AI working together for a very long time to produce this thing. And there are many, many companies who want to do the same thing. So from a competitive side, you can see this is a maturation of the field. On the safety side, I would say that the safety side is not yet as salient a reason as the competitive side, but it's going to change. And it's basically as follows. These models are very potent,

and they're becoming more and more potent. At some point, it will be quite easy, if one wanted, to cause a great deal of harm with these models. And as the capabilities get higher, it makes sense that you don't want to disclose them, end quote. The closed approach is a marked change for OpenAI, which was founded in 2015 by a small group, including current CEO Sam Altman, Tesla CEO Elon Musk, who resigned from its board in 2018, and Sutskever. In an introductory blog post, Sutskever and others said the organization's aim was to, quote, build value for everyone rather than shareholders, and that it would, quote, freely collaborate with others in the field to do so. OpenAI was founded as a nonprofit, but later became a capped profit in order to secure billions in investment, primarily from Microsoft,

with whom it now has exclusive business licenses. When asked why OpenAI changed its approach to sharing its research, Sutskever replied simply, we were wrong. Flat out, we were wrong. If you believe as we do that at some point AI, AGI is going to be extremely unbelievably potent, then it just does not make sense to open source. It is a bad idea. I fully expect that in a few years it's going to be completely obvious to everyone that open sourcing AI is just not wise, end quote. Opinions in the AI community on this matter vary. Notably, the launch of GPT4 comes just weeks after another AI language model developed by Facebook owner Meta, named Lama, leaked online, triggering similar discussions about the threats and benefits of open source research.

Most initial reactions to GPT4's closed model, though, were negative. Speaking to the verge via DM, NAMIC AIs, Schmidt explained that not being able to see what data GPT4 was trained on made it hard to know where the system could be safely used and come up with fixes. For people to make informed decisions about where this model won't work, they need to have a better sense of what it does and what assumptions are baked in, said Schmidt. wouldn't trust a self-driving car trained without experience in snowy climates. It's likely there are some holes or other problems that may surface when this is used in real situations, end quote. William Falcon, CEO of Lightning AI and creator of the open source tool

pie torch lightning, told Ventrabeat that he understood the decision from a business perspective. You have every right to do that as a company, but he also said that the move set a bad precedent for the wider community and could have harmful effects, end quote. Lots of people are objecting to OpenAI saying, this stuff is so dangerous, we don't think we can trust it with just anybody. You should trust us, though, which may or may not be true. And certainly I can see the argument, as I've shared with you from recent pieces, that all these tools floating around out in the wild could lead to some wildly bad outcomes.

But their argument about the competitive landscape is also interesting and troubling because it essentially hands any competition out there, the nomics or anthropic or stability AI or whomever, a huge differentiating argument all of a sudden. It's the old argument, the one Apple used to make about IBM way back in the day and then about Microsoft more recently. The argument being, don't go with Big Brother, in the form of open AI, in this case, go with us because we're more open and more transparent and perhaps more friendly to your projects that you want to do on your own. Finally today, just a note that this finally happened. Stripe has raised more than $6.5 billion in a series I at a $50 billion valuation

to help its employees cover tax obligations for the expiration of their RSUs and to fund a new stock tender offer. Quoting Axios, this is much more than the company had expected to raise, albeit at a lower price. All proceeds will be used to help Stripe employees cover tax obligations related to the pending expiration of restricted stock units, plus to fund new stock tender offers for current and former employees. No proceeds will be used to fund Stripe's own tax obligations tied to the RSUs. The stock tender is voluntary with employees eligible to sell as few or as many vested shares as they want. Stripe had considered letting employees get liquidity via a public stock offering, but ultimately decided that the private transaction would be faster and give the company

more flexibility. The company which powers online payments for such companies as Amazon and Lyft was valued at around $95 billion by venture capitalists in March 2021, up from $35 billion just three months earlier. Early talk had been that the new round would be done at around a $60 billion valuation. New investors in the round included GIC, Goldman Sachs, Asset and Wealth Management, and Tamasek. Return backers included Andreessen Horowitz, Bally Gifford, Founders Fund, General Catalysts, MSD Partners, and Thrive Capital, end quote. Okay, because I am running late, I don't even have time to look up a song lyric for you, so talk to you tomorrow.