Apple is Getting an OpenAI Board Observer Seat

Today on the AI Daily Brief, Figma is pulling its AI features and Apple is getting an OpenAI board observer seat. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show. Notes. Hello AI Friends, A quick note before we dive into today's episode. Before because of the holiday week, I'm doing this episode a little bit differently. It is an Extended Headlines edition. We have four or five different stories that we cover in a little bit more depth but not necessarily a full feature length depth. Anyway, some really interesting stuff in here, but just wanted to give you that heads up. Usually we take holidays off here, but I will have a special bonus interview episode coming out tomorrow and then we will do a long read over the weekend as well. For now though, let's dive into today's episode. Welcome back to the AI Daily Brief. This episode is an Extended Headlines edition. Well, one of the big stories from last week was that Figma had launched a slew of new AI features as part of its big update at its annual config conference. This week though, things have taken a rough turn. So what happened is Andy Allen from Not Boring Software shared a bunch of videos of using Figma's new Make Something AI, which is basically its answer to the blank page problem. When asked to create a weather app, it created something that looked a lot like Apple's weather app. Andy says Figma AI looks rather heavily trained on existing apps. This is a weather app using the new Make Designs feature and the results are basically Apple's weather app tried three times, same results. So for those of you who are looking at the screen right now, over here on the left we have Apple's existing weather app and these three from Figma, which are all generations with the AI tools. This was a big enough deal that Dylan Field from Figma took to Twitter to defend the company. He wrote, as we shared at Config last week, as well as on our blog, our website and many other touch points, the Make Design feature is not trained on Figma content, community files or app designs. In other words, the accusations around data training in this tweet are false. How does make design work? As we have explained publicly, the feature uses off the shelf LLMs combined with design systems we commissioned to be used by these models. The problem with this approach which I outlined in my keynote last week is that variability is too low. Within hours of seeing this tweet we identified the issue which was related to the underlying design system that were created. Ultimately, it is my fault for not insisting on a better QA process for this work and pushing our team hard to hit a deadline for config. I've asked our team to temporarily disable the Make Design feature until we are confident we can stand behind its output. This feature will be disabled when our US team wakes up in a few hours, and we will re enable it when we have completed a full QA pass on the underlying design system. As I discussed at config, we believe that craft is your competitive advantage and it's more important than ever to create unique designs. We live in a world where any off the shelf LLM can help you go from prompt to software in one shot. Designers will need new tools to explore the option space of possibilities and we intend to build these tools. My belief is that if we build these tools the right way, then designers will be able to have more creative expression, not less. I hate missing the mark, especially on something that I believe is so fundamentally important to the future of design more soon. So the thing that lots of people have noted here is that there is a new sort of deniability when it comes to AI training. The Verge, for example, wrote in an interview Tuesday with Figma CTO Chris Rasmussen. I asked him point blank if Make Designs was trained on Apple's app designs. His his response? He couldn't say for sure. Figma was not responsible for training the AI models it used at all. Basically, the blame is being passed to the OpenAI's of the world. According to Rasmussen, for example, the key AI models that power make designs are OpenAI's GPT4O and Amazon's Titan Image Generator G1. OpenAI and Amazon have not responded to a request for comment. Given how big figma is in the design world, this could have a pretty significant impact on how designers think about putting into practice these AI tools. We'll see how it evolves from here, but for now, not great. Staying on Apple for just a minute A big scoop from Bloomberg as part of their deal with OpenAI to put ChatGPT on the upcoming iOS 18, Apple is going to get an OpenAI board observer role, according to Bloomberg's Mark Gurman sources. Phil Schiller, who's the head of Apple's App Store and its former CMO, is the person who will be joining OpenAI in that board observer role. Of course, the incredibly notable thing here is that Microsoft, who has invested $13 billion into OpenAI, also has a board observer role. And so, as Bloomberg puts it, the board observer role will put Apple on par with Microsoft, Gurman goes on. Having Microsoft and Apple sit in on board meetings could create complications for the tech giants, which have been rivals and partners over the decades. Some OpenAI board meetings will likely discuss future AI initiatives between OpenAI and Microsoft, deliberations that the latter company may want Shiller excluded from. However, as Gurman points out, board observers do often oblige and exit meetings during discussions that are seen as sensitive. A lot of the commentary is represented in this tweet from MG Siegler. Microsoft cost for an OpenAI board observer role around 15 billion Apple costs for an OpenAI board observer role $0 jokes aside, this is pretty wild, paul Thurot writes. Phil and the Microsoft observer will employ rock, paper, scissors when they disagree on strategy. Now obviously there are ways to solve the conflicts here, and Microsoft's deal is much bigger than their relationship with OpenAI's board. Remember, they didn't have a role on the board at all until the whole thing with Sam Altman happened last November. But still, it does show just how complicated this relationship that runs through OpenAI could be. And it's likely that it could be a little bumpy as things get going. Speaking of Microsoft, you might remember that they signed a deal to make a $1.5 billion minority investment into the UAE's G42. At the time, it was very clear that the Department of Commerce was involved with the deal. In fact, Commerce Secretary Gina Raimondo was one of the people who announced it. Speculation has largely been that the deal was in fact in large part facilitated by the US government as a way to have greater visibility into G42's books. G42 has been at the center of the conversation in the Middle east about whether companies in that region can deal simultaneously with the US and China. In short, the US's answer is that no, they cannot, and G42 has been increasingly focusing their attention on the US market. It had seemed that the Microsoft deal was the culmination of that, but now Bloomberg is reporting that there are still some in Washington who think the deal shouldn't go through because of their national security concerns, writes Bloomberg. The linchpin of the two companies relationship is an agreement blessed earlier this year by the Biden administration for Microsoft to invest 1.5 billion in the UAE firm in exchange for G42's agreement to cut ties with Chinese companies. Pentagon Officials are skeptical G42 will entirely uncouple from China, according to people familiar with the situation, Microsoft's chronic cybersecurity failures enumerated recently in a scathing government report, have also stoked fears that foreign powers could access sensitive U.S. networks. Other officials wonder if it's a good idea shipping advanced AI to Gulf states with questionable human rights records, Bloomberg continues. The resistance to the G42 Microsoft deal could hinder the administration's efforts to help US companies grab a global lead in the development of AI, which is already emerging as the next front in Washington's cold war with Beijing. US Officials want to make sure that American firms, not their Chinese rivals, benefit from Gulf states ambitions to fund and build big AI projects. By dangling the promise of access to American technology, they hope to persuade Middle Eastern states to sever ties to Beijing. It's hard for me, from where I'm sitting, to really wrap my head around this. It seems likely to me that this is not necessarily the whole government changing its opinion of this deal, but instead reporting that reflects what a lack of consensus there is around the right way to handle these issues in general. In other words, it could be simultaneously true that the Commerce Department was trying to use this deal as a way to bring G42 closer into the fold, but at the same time that other officials think that that's a crazy and bad idea, and now the reporting that we're getting is just showing all sides of this particular issue. Whatever the case, it is a fascinating wrinkle in what remains a key geopolitical story for AI, and one that I will continue to watch closely. Today's episode is brought to you by Super Intelligent, the platform for fun fast AI learning. Super has a ton of new things going on. We recently announced our partnership with Spotify, through which users of that app can now access super intelligent content directly from their mobile apps. We've also just launched the AI Learning feed. In addition to seeing the tutorials that we're dropping, there are polls, news items with related lessons, and a chance for people to show off the projects and use cases that are making AI come alive for them. We've also just kicked off the Super Summer Challenge, where each week we'll share a new challenge that you can use to discover new AI tools and use cases. Go to BeSuper AI and use code SUPERFUND for 50% off your first two months. That's BeSuper AI. Today's episode is brought to you by Venice. Venice is a private, uncensored, generative AI app. It accesses open source models to enable text, image and code generation without the fear of being spied on or having your data exploited. Discuss anything with Venice without concerns about it being monitored, sold, or given to advertisers and governments. Venice is different because your conversations and creations are kept securely within your own browser, never stored or accessible by Venice. Unlike other AI apps, Venice won't tell you what's okay to say or not. Venice won't patronize you. It simply provides direct access to machine intelligence. No topics are off limits, no ideas are taboo. With Venice, you're in control of the AI, as you should be. Pro subscriptions are available for $49 a year or $8 per month. Try it for free without an account at Venice AI. Next up, an interesting story from the world of finance. Bridgewater is launching a new $2 billion hedge fund that will be run entirely by machine learning. So basically, Bridgewater, which is of course a huge hedge fund, has been working on proprietary models for investing AI for some time. However, they've recently broadened that effort to include models from OpenAI, Anthropic Perplexity and others, and they finally started to use external capital after testing it with a $100 million fund of their own money over the past year, said CEO Nir Bardia. The push into machine learning is a good manifestation of us taking the flag and putting it at the top of the mountain. This is maybe the most significant and pure manifestation of the moment we're in, said Co Chief Investment Officer Greg Jensen. The big jump here is using machine intelligence to generate the alpha. That is a leap. So will humans be involved at all? According to Yahoo Finance, Bridgewater's humans will still be overseeing a number of functions, including risk management, data acquisition and trade execution. But this is definitely leaning on machines in a huge way. So is this the future of investing? We will have to wait and see. Moving the investing conversation back to startups A couple interesting stories around AI startup fundraising. Runway, the video generation startup which just released their Gen 3 model, is, according to the information in talks to raise at a $4 billion valuation. According to the information sources, Runway is looking to raise around $450 million at that valuation, and General Atlantic, a New York PE firm, is in talks to lead the round. Up to this point, Runway has raised around $230 million and was most recently valued at 1.5 billion in June of last year. The information points out that if Runway gets this deal, it would make the deal relatively expensive, they write. The $4 billion valuation implies that investors value the startup at about 160 times its end of 2023 arrows, which is significantly higher than the multiple of other in demand AI startups, which have been raising it 50 to 100 times their forward revenue. From a completely biased perspective. I love Runway and what they're building, so I hope they get all the money that they need. Now another deal which is perhaps even more interesting in terms of what it says about the state of the industry is a new round for Harvey to get a sense of the tone of this reporting, the Information writes. Harvey nears new smaller deal with GV after its acquisition plans collapse Basically, Harvey is an AI legal company that had been looking to raise $600 million at a valuation of at least $2 billion. As part of that deal, they were hoping to acquire Velix, which is a 25 year old legal research firm. But, the Information writes, the acquisition fell apart and the funding round shrank. So now it appears that Harvey is going to raise around $100 million, obviously a far cry from that 600 million at a valuation of around one and a half billion, the Information writes. One investor who spoke with Harvey executives was concerned that the acquisition of Velex wouldn't ensure Harvey would catch up to one of its main competitors and and would have spent a lot of money to boot. That rival, Casetext, now has easy access to enormous amounts of legal data, thanks to its acquisition last June by Thomson Reuters. The Information also reports that investors in this round were able to win some special protections that haven't been as common recently in AI fundraising. Whenever we get these types of stories, the question is, is there something going on in terms of trends in AI fundraising, or is this all specific to the particulars in this case the company, Harvey and the legal AI space more generally? I don't know the answer to that, but I do know that the Information is also reporting that Character AI, as popular as it is, is thinking about doing deals with its arch competitors in Google and Meta. Once again, the Information writes. A year ago, Character AI had the wind at its back. Founded by two AI pioneers from Google, the startup had launched AI chatbots that offered what its far larger rivals hadn't yet provided something fun. Millions of people had signed up for its website and apps, intrigued by the chance to text with facsimiles of anime characters, TV personalities and historical figures in hyperrealistic roleplay. Executives discussed raising additional funding four months after Andreessen Horowitz and others had sunk $150 million into the startup at a $1 billion valuation. That head start has since receded. In its place are signs Character, founded in November 2021, could follow other AI startups into the arms of a larger tech company so the point of this article is one that competition has increased significantly on this show we've talked about before. The fact that Meta has recently started testing chatbots made by creators, and that Google has also been recently developing a Character AI style product. Because of this, reports are that Character has held exploratory talks over, quote, possible research partnerships. These include potential collaborations with Google and Elon Musk's xai. Of course, what these quote unquote research projects actually mean is a little more questionable in a world where the deals that we've been seeing recently are things like Microsoft's non acquisition acquisition of inflection and Amazon's more recent non acquisition acquisition of Adept. If Character does end up having to go to either Meta or Google, I do think it would be a strong indicator of just how hard competition is in this AI space and the level of resources you need to really compete. For now, though, this is all just reporting and so we'll have to see how it all shakes out. That, however, is going to do it. For today's Extended Headlines edition of the AI Daily Brief, until next time, Peace, Sam.