The AI Daily Brief: Artificial Intelligence News and Analysis - The Significance of Ilya Sutskever Leaving OpenAI

Episode Date: May 18, 2024

OpenAI’s chief scientist and co-founder Ilya Sutskever has left the company, sparking widespread speculation about his departure. Was there a deeper concern that led to his exit, or is this a more routine departure than it seems? This episode of the AI Daily Brief delves into the timeline of events, the implications for OpenAI, and the broader AI community’s reaction. ** Check out the hit podcast from HBS Managing the Future of Work https://www.hbs.edu/managing-the-future-of-work/podcast/Pages/default.aspx Join Superintelligent at https://besuper.ai/ -- Practical, useful, hands on AI education through tutorials and step-by-step how-tos. Use code podcast for 50% off your first month! Check out https://useplumb.com/ to build complex AI pipelines simply. ** ABOUT THE AI BREAKDOWN The AI Breakdown helps you understand the most important news and discussions in AI.  Subscribe to The AI Breakdown newsletter: https://aidailybrief.beehiiv.com/ Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@AIDailyBrief Join the community: bit.ly/aibreakdown

Today on the AI Daily Brief, Ilya has officially left OpenAI. Before that on the headlines, speaking of OpenAI, they've cut a deal with Reddit and the market loves it. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. Check out the Discord link in the show notes to join the conversation. Welcome back to the AI Daily Brief headline edition, all the AI headlines you need in around five minutes. We kick off today with an announcement of a deal between OpenAI and Reddit. This is very similar to the deal that we heard about a few months ago between Reddit and Google, which was, of course, at the very center of Reddit's pre-IPO Roadshow.

Basically, Reddit was touting two investors that, in addition to the way they made money currently on advertising, in a world powered by AI, their data had value as well. Given that Reddit outperformed basically everyone's expectations on the first day of trading, it seemed like that message was resonant. Well, today, Reddit was up nearly 12% in pre-market trading after this OpenA licensing deal was announced. There are a few different parts of the deal. The thing that's perhaps most notable is that there was nothing mentioned about training data. Now, whether that means that OpenAI has already used all of Reddit's data in their training, or whether it was explicitly left off isn't clear.

But it's highly notable because that's different than the deal they cut with Google. So what is in this deal? Well, part of it is that OpenAI gets access to real-time content from Reddit's data API. This means that OpenAI can surface content from Reddit in Chatchip-T. If it's not clear to you by now, OpenAI is moving towards a model where they want to be able to reference the entire world of content explicitly in search results. In other words, it's very hard to see how ChatGPT just doesn't merge with a perplexity-style research use case in the future. Whether that gets instantiated as a totally new product or just an evolution of what ChatGPT is remains to be seen, but that's definitely the direction. Other parts of the deal including OpenAI

advertising on Reddit and Reddit bringing new AI powered features to Redditors and mods using OpenAI's models. Up next, we will have to see how Reditors take the deal. Moving on to our next headline earlier this week, the information reported that Stability AI looked to be forced into a sale. Well, now it appears that an investment consortium is coming together, which may have some new plans for the startup, and it seems that those plans may involve Hollywood and the entertainment industry. The information reports that Stability AI is in talks to raise money from a group including Sean Parker, who is of course famous for Napster and his early work with Facebook. Pram Akaraju is also involved who's the former CEO of Weta Digital and a previous co-founder with Parker of

Streaming Service, The Screening Room. The information writes, these media connections could help stability strike deals with film studios or other Hollywood companies looking to use its image and video generation software. However, as the information points out, that puts them directly in the competitive path with rivals like OpenAI and Google. This week, we got Google's answer to SORA, which was called Vio, and the way that it was announced was with Donald Glover, previewing the feature as a filmmaker himself. Still, I said before that I hoped that for the sake of the team at stability and the continued interesting competition in this space, that the situation all resolved well, and this is certainly more than an interesting lifeline.

Staying on the theme of the entertainment industry, Sony Music Group has sent letters warning more than 700 companies against using their content to train their AI models. Sony owns labels including Columbia, RCA, and Epic Records, and according to various outlets, they've started sending formal letters to more than 700 generative AI companies, prohibiting them from, quote, mining, text, or data, web scraping, or otherwise using any SMG content without explicit licensing agreements. That includes audio recordings, musical compositions, including lyrics, cover artwork, and metadata, and more. As complement, they also posted what they called a Declaration of AI Training opt-out. That letter reads, SMG is a passionate believer in the inherent

and paramount value of human artistry. Additionally, SMG has been embracing the potential for responsibly produced AI to be used as a creative tool, revolutionizing the way songwriters and recording artists create music. We support artists and songwriters taking the lead in embracing new technologies and support of their art. Evolutions and technology have frequently shifted the course of creative industries. AI will likely continue that longstanding trend. But the letter continues, don't use our crap to train your models. While there isn't any sort of direct legal threat here, it's clear that this is not just a statement of principle, but a legal argument being formed. And speaking of legal arguments, the latest AI lawsuit is from two voice actors who say that their voices were cloned without their permission by an

AI startup called Lovo. The New York Times writes, in their suit filed in federal court in Manhattan, on Thursday, the couple said anonymous Lovo employees had paid them for a few voice clips in 2019 in 2020 without disclosing how the clips would be used. They say Lovo, which was founded in 2019, is violating federal trademark law and several state privacy laws by promoting clones of their voice. These suits seek class action status, inviting other voice actors to join in. Now, this being the headline section of this, I'm not going to get into all the details, but suffice it to say, this is yet another example of murky legal territory in this new AI world that is going to have to be fought out and resolved in courts, and so it is worth watching

closely. For now, however, that is going to do it for the AI Daily Brief Headlines edition. Next up, the main episode. As a listener of this show, I have a strong feeling you like to stay up to date on all things artificial intelligence, including its impact on the workforce, which is why I highly recommend checking out managing the future of work, the chart-topping business podcast from Harvard Business School. HBS professors Bill Kerr and Joe Fuller talked to business leaders, technologists, and policymakers grappling with the forces like AI, globalization, and demographic shifts that are reshaping the nature of work. Recent guests include IBM's CHRO, Nicol Lamarro, on how Big Blue is adopting AI,

Morningstar CEO, Kunal Kapoor on how AI can raise the investment IQ, Microsoft Corporate Vice President Jared Spatero on how the tech giant is experimenting its way from AI assistance to autonomous agents, and many other prominent movers in business and the workforce ecosystem. So don't miss out. Follow managing the future of work on Apple Podcasts, Spotify, or wherever you're listening now. Hello, friends. Quick note before we get back to the show, I'm so excited to share that Super Intelligent is now live. Super Intelligent is a platform for fast, fun, and super practical,

useful AI learning. We have something like 300 video tutorials adding 30 to 50 each week, covering every topic in AI you can imagine from LLMs to image generators to case studies, use cases, basically everything that tells you how to use AI and what to use it on. In short, fast four to seven minute tutorial videos, which are paired with step-by-step instructions that help you actually use these tools as well. It's $20 a month for unlimited access, and I would love to see you there. Check it out at B-super.a.I. That's B-super.a-I. For months ever since the ouster and then return of Sam Altman last November, the refrain has been, what did Ilius see? Ilius Sutskiver had a very weird, it seems, part to play in the whole debacle. It was a

reported at various points that he was one of the main drivers of Sam initially being removed from OpenAI, but then at some point along the way, he retracted his position publicly and threw in his lot with the group that was trying to get Sam back. Perhaps because of that conversion, it wasn't clear if he was going to stick around, or if he was just living on inevitably borrowed time. What's more, the fact of his switch led to lots of confusion around what the actual cause for Sam's firing might have been. In other words, the natural assumption that many people jump to, was that there was some fundamental concern around a misalignment of priorities, open AI moving towards an AGI future that they couldn't control, starting to think about business

over the bigger philosophical mission. However, as many pointed out, if that was the case, it doesn't seem likely to be the thing that you would reverse position on. What's more, at various points, Ilya and others denied that that was what it was about. So all of this was really confused, and for months, people have basically just been waiting for the other shoe to drop. Well, Drop it finally did when on Tuesday, May 14th, I tweeted, After almost a decade, I have made the decision to leave OpenAI. The company's trajectory has been nothing short of miraculous, and I'm confident that OpenAI will build AGI that is both safe and beneficial

under the leadership of Sam, Greg, Mira, and now under the excellent research leadership of Jacob Pachaki. It was an honor and a privilege to have worked together and I will miss everyone dearly. So long and thanks for everything. I'm excited for what comes next, a project that is very personally meaningful to me, about which I will share details in due time. Two minutes later, Sam Altman tweeted, Ilya and OpenAI are going to part ways.

This is very sad to me. Ilya is easily one of the greatest minds of our generation, a guiding light of our field and a dear friend. His brilliance and vision are well known. His warmth and compassion are less well known, but no less important. Open AI would not be what it is without him. Although he has something personally meaningful he is going to go work on, I am forever grateful for what he did here,

and committed to finishing the mission we started together. I am happy that for so long I got to be close to such genuinely remarkable genius, and someone so focused on getting to the best future for he, humanity. Jacob is going to be our new chief scientist. Jacob is also easily one of the greatest minds of our generation. I am thrilled he is taking the baton here. He has run many of our most important projects, and I am very confident he will lead us to make rapid and safe progress towards our mission of ensuring that AGI benefits everyone. So clearly we have here the beginnings of a narrative. Both Sam and Ilya referenced as something personally meaningful that he's

going to go work on. On the one hand, it's a fairly clean, non-controversial way to end a relationship on a positive note, and so might be the type of thing that you would just say, even if it wasn't exactly true. On the flip side, it'll be pretty easy to tell whether it was true based on what Ilya does next. Greg Brockman also wrote effectively an emotional thank you letter to Iliah on Twitter as well. He wrote, when we started in late 2015, OpenAI was a nonprofit with a mission to make AGI go well, but without a credible plan of how to accomplish it. In the early days, I spent countless hours thinking hard about every aspect of culture, technical direction, and strategy. Together, we realized that we would need to raise a lot more capital than anyone had imagined

in order to build supercomputers of unprecedented size. We fundraised together in the nonprofit, raising more than others thought possible, but still far less than was needed. We proposed creating a for-profit structure and service of the mission, and once that had been created, we continued to align in shape what the company stood for. A lot of the comments are like this one from Vittorio that says, One last question remains, though. What did he see?

For a lot of folks, this was completely inevitable. The Verge, for example, writes Open AI chief scientist Ilius Sutskiver is officially leaving, who didn't see this coming after he tried to get Sam Altman fired. Along those same lines, superhuman founder Vivek Sodero wrote, It's been six months since the Sam, Ilya, and OpenAI board debacle. Would have sent shockwaves to the community if Ilya left when Sam rejoined. I'm willing to bet this was planned all along as Sam and the board convinced Ilya to stay for six months in order to manage optics. This did happen almost exactly six months from when that whole thing went down.

And I think Vivek is probably very close to right here. Now, maybe there was some question of whether Ilya would actually follow through with leaving. I wouldn't be surprised if, for example, they negotiated this single. month window, but were clear that as long as Ilya wanted to stay, he would have a place there. But it seems highly likely to me that Vivek's analysis of what was going on behind the scenes is at least in part correct. Now, what added some heightened drama to the whole thing, was when Jan Lakey, who co-leads the super alignment team at Open AI with Ilya, posted a few hours later, the very terse, no happy corporate boilerplate message, I resigned.

Adding even more intrigued to that was that Logan Kilpatrick, who also recently resigned from OpenAI, followed up on that tweet saying keep fighting the good fight, to which of course the natural question was, what is the good fight? For those of you who don't remember, the super alignment team was something that was announced in July of last year. Basically, it was an acknowledgement that we needed way more nonlinear progress to actually be able to steer AI systems that get smarter than humans. The super alignment announcement came with a time goal. They wanted to solve the problem within four years, as well as a dedication of compute effort, with Open AI committing to dedicate 20% of the compute that they had secured up to that date to that effort, and at the time, it seemed like

OpenAI ratcheting up their focus on this issue. For that reason, it's understandable why some folks, particularly on the AI safety side of things, would be concerned with this turn of events just 10 months later. Robert Masi H.co summed up many people's sentiments when he wrote, Five core safety experts at OpenAI, including Ilya, who is involved in firing Sam Altman, have left the company. Jan Lakey has since resigned, and that's just as concerning as Ilya leaving. Both are highly respected in AI, so losing them is a big deal. I believe the main reason for firing Sam was this group's fear that AI development was moving too fast without enough safety measures, even if the audit didn't support this. It's like a bank getting rid of its risk

management team and hoping loans don't default or a pharmaceutical company releasing new cancer drugs without proper testing. What's the Open AI board stance on this? Are they pushing full speed ahead? With no U.S. regulations and unlikely international ones were left hoping everything turns out okay. Like I said, Robert was far from alone with this sentiment. There have been a group of people who ever since the whole firing happened have been unwilling to accept any other explanation for what went down than these sort of safety-based concerns. There are some who think that if this is the case, they'd like to see more action. Jeffrey Miller, for example, writes, so Edward Snowden risk treason charges and life in jail to blow whistles on NSA mass surveillance

affecting 350 million Americans. I hope some of you guys quitting OpenAI will have the courage to blow the whistle on anything unsafe they're doing that could affect 8 billion people. However, there were many who also felt that the opposite might be true. Brian Atwood, for example, wrote, plot twist, the safety team left not because they saw something, but because they saw nothing. No real danger, only limitations, dead ends, and endless distractions with commercialization. No path to AGI with GBT, but the product is live and the bags must be pumped. AGI will be achieved, but not by Open AI. I think it's an interesting theory, although we don't have any particular evidence of this. And I thought, frankly, even more compelling an argument

from just a sheer rationality standpoint, came from Mason on Twitter who wrote, people are freaking out about the Ilya Jan resignations like the obvious thing anyone would do if their company was about to destroy humanity is resign and post-cryptic tweets about it. She continued, I'm not saying it's a nothing burger, but I am saying that public book cryptic resignations are obviously getting these very intelligent guys more bang for their buck than violating their NDAs. And I don't think that's compatible with the idea that we're all about to die. So, even as some questions get resolved, specifically the question of whether Ilya will be staying at OpenAI, the more significant questions absolutely remained. That, however, is going to do it for today's AI Daily Brief.

Until next time, peace.