OpenAI's Next Big Model is Called Orion

Today on the AI Daily Brief, new reports about OpenAI Strawberry and what it means for a new model that they're codenaming Orion. Before that in the headlines, Elon Musk has come out in favor of SB 1047. 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 welcome back to the AI Daily Brief Headlines all the daily AI headlines you need in around five minutes. We kick off today with the latest development in the conversation around California AI safety legislation SB 1047, where Elon Musk, the CEO of Twitter and XAI, has come out in favor of the bill. Last night he tweeted, this is a tough call and will make some people upset. But all things considered, I think California should Probably pass the SB 1047 AI safety bill. For over 20 years I have been an advocate for AI regulation, just as we regulate any product or technology that is a potential risk to the public. So this sent shockwaves through the community, both good and bad. Safety advocate Eliezer Yudkowski writes, honestly, I haven't studied SP 1047 hard enough. I don't think it saves the world by itself. So the main impact is political consequences on which I'm not expert. But given how it selectively hits big AI companies like yours, this looks to me like an act of integrity by you. Dan Hendricks, the director of the center for AI Safety, who was one of the voices most influential in drafting the legislation, wrote, you're the best, Elon, and used it as a chance to provide a TLDR of the bill. The much more common response, though, was one of incredulity and accusations of hypocrisy. BOK CEO Aaron Levy writes, the unintended consequences of this bill in many years from now would likely be quite bad. The amount of risk that model providers take on would almost necessarily slow down the rate of AI progress and reduce competition overall. And as we know with regulation, it's often a one way door. If this bill had been written two years ago with the same relative numbers, we likely would have stalled the major breakthroughs we've seen since. Bindure writes, it's premature to pass a bill when we don't understand exactly how AI can be dangerous. We don't have any proof that it has been dangerous so far. Ari Paul writes, yeah, we get it. You finished the data center and now want a regulatory moat so you can extract rents without facing competition. As always, the principles go out the window as soon as the grift is ready, developer Nick Dobo said something similar. Someone finished building a data center and now wants regulatory capture. The other common theme was Elon setting fire to the door on his way out of California. Martin Casado from A16Z writes, Step one move out of California. Step two support legislation that'll hurt California. Well played Mr. Musk. Well played. 1. Move X and Xai out of California 2. Build AI superclusters in Texas and Tennessee 3. Support California build at Nukes Competition 4. Profit Clem DeLang from Hugging Face writes, is Elon Musk Replicating the OpenAI playbook from open source Grok 1 to closed source Grok 2? From pro startups to regulatory capture? If you can't beat them, join them I guess. Ha ha. Anyways, for now it is just one more data point and we will see how it all shakes out when the vote finally happens. Now, it's not the only legislation that's being considered right now. In fact, a group of companies, including some that have come out against SB 1047, have come out in favor of something else. OpenAI, Adobe and Microsoft are among the companies supporting AB3211, which requires watermarks in the metadata of AI generated photos, videos and audio. Interestingly, back in April, a trade group that represented Adobe and Microsoft, among others, had opposed AB3211, calling it unworkable and overly burdensome. But amendments to the bill seem to have changed the situation as they are now in support. This certainly feels, at least in part, a way for these companies to show that they're not against any regulation. They just don't want what SB 1047 has to offer. Staying on the big company theme for a minute, Amazon is finally getting set to release an update to a new AI version of Alexa. According to internal documents obtained by the Washington Post, the company is preparing to release AI Alexa in October, but it will require a paid subscription to have access to it, according to the documents, Classic Alexa the non AI version will remain free to use, but the new assistant could cost as much as $10 a month for things like AI features that help customers curate, summarize and explore current events, which they say was rated as one of the top customer requests. I could be completely wrong here, but I am extraordinarily skeptical that Amazon is going to be able to get people to pay $10 a month if the primary value proposition being advertised is news summaries, it's hard not to feel that Amazon continues to be a little bit behind the eight ball when it comes to their AI strategy. And frankly, if the big launch of Alexa in October, more than a year after they announced these updates, is a news summarizer, that is going to be painful. Moving on, you might remember that most of the Inflection team went over to Microsoft with Inflection CEO Mustafa Suleyman now running AI internal to Microsoft. But the remainder of Inflection is still cooking and had a bit of news today, the most interesting piece of which is an attempt to develop a new mechanism for exporting data from chatbots that could be standardized, that could lead the way to an eventual industry standard. Starting from right away, users of the Chatbot PI will be able to export their previous conversation history and they're hoping that this becomes more normalized in the future. Finally, it appears that we have a date for the next big iPhone launch, which of course for those of you listening to this podcast, is much more notable as a platform for Apple's new Apple Intelligence AI rather than just as a new piece of hardware. September 9th at 10am Pacific time you can tune in at apple.com, although rumor is that the Apple Intelligence features won't be shipping right away and so we may still have a little while to wait after that. For now though, that is going to do it for today's AI Daily Brief headlines. Next up, the main episode. Today's episode is brought to you by Fractional. When we wanted to build an AI powered feature of Superintelligent, our AI tool Finder, I went straight to Fractional. Fractional Team is a group of senior engineers in San Francisco working on some of the most exciting projects in applied AI. Working with them is basically like hiring an absolute top flight AI engineering team, but in a way that you can customize exactly for your particular needs. Like I said, that AI Tool Finder feature that we built with them is already a key part of the super intelligent platform and we are working on something new as well. As Fractional works with everyone from startups to the Fortune 500 and to request a free consultation, you can go to Fractional AI. If you want help identifying and building AI projects for your business, then I highly recommend you hit pause, open a web browser and go to Fractional AI to request a free consultation. Today's episode is brought to you by Venice. The leading AI companies store your entire conversation history and attach it to your identity forever. That's every question you ask, every answer you receive, every image you generate, every thought you share with the machine. It's all being spied on. If you trust all the companies, hackers and NSA board members that will ever have access to your AI conversations then rejoice for you are well served. For the rest of us, Venice is an alternative. Venice is a powerful AI app for text, image and code generation that respects you as a sovereign individual and believes privacy and free speech are not only human rights but necessary for civilizational advancement. Private, permissionless and uncensored, you can try it for free without an account. A Daily Brief listeners receive a 20% discount on Venice Pro. Visit Venice AI NLW and enter the discount code nlwdailybrief. That's NLW Daily Brief all one word today's episode is brought to you by Superintelligent, which is of course our platform that helps you learn how to use AI tools and perhaps even more importantly, gives you ideas on the best use cases that are actually going to help you achieve whatever it is you want to achieve. To recognize the end of Summer and Back to school. Back to work. We are running our best promotion ever. When you sign up for Super Intelligent between now and the end of August using code so back, your first month will be 100% free. The platform features over 600 fun, highly practical AI tutorials that get you using AI fast and with an eye to actually transforming how you get things done. We've just launched super for Teams, so if you have a group of people at your company that want to figure out how to use AI together, I highly suggest you check it out. But for those of you who are using superintelligent as an individual, once again, if you sign up for Super Intelligent between now and the end of the month using code so back, you will get your first month 100% free. Go to be super AI and check it out today. Welcome back to the AI Daily Brief. A few weeks ago there were serious rumors flying around about OpenAI and their strawberry reasoning model. The quick TLDR on all of this goes all the way back to around Sam Altman's firing, when there was speculation that part of the reason that people had gotten spooked internally was a new technical breakthrough around a reasoning AI that they were calling Q. A few months later, Q was reported to be renamed Strawberry, and at some point over the last month and a half or so, Sam Altman and others from OpenAI started posting pictures of strawberries on Twitter. There was also a new theoretical leaker who emerged, making all sorts of prognostications around forthcoming information, which by the way, never materialized. Ultimately all of that dissipated, but now we have some actual sourced information in an Exclusive report from the Information the piece reads, OpenAI races to launch Strawberry Reasoning AI to boost chatbot business so what is going on here? First of all, let's talk about the new information we have about Strawberry. The examples that the information gives, which are sourced from two people who have been involved in the effort with Strawberry, are that it can do things like solving math problems that it hasn't seen before, solving complex problems involving programming, solve the New York Times connections, a complex word puzzle, or reason to create more comprehensive answers around subjective topics such as product marketing strategy. The next bit of information from this report is that Strawberry could be coming as soon as this fall, that it appears that the race is on to release this, possibly within ChatGPT. A third interesting piece of information is that it appears that part of the reason that they're so eager to get Strawberry online is in order to improve upon GPT4. We have a new codename for a new OpenAI flagship LLM, and that codename is Orion. One of the people who talked with the information said that OpenAI is using a big version of Strawberry to generate data for training Orion. In other words, if OpenAI is facing limitations in obtaining enough high quality data to train Orion, they're looking to synthetically created data from Strawberry to fill in that gap. Strawberry is also seen as a potential aid for upcoming OpenAI agents, according to these sources. Of course, this is another frontier that is opening up in this battle, with agents widely expected to at some point represent another phase change when it comes to what generative AI can do now. Another interesting part of the framing of this piece is that they set up a certain urgency around how this could help with a new fundraising process. Despite OpenAI being the most monetarily successful generative AI product in ChatGPT, it's still widely suspected to be operating at a loss. The tone throughout this piece is very different than something we've seen around OpenAI before, which is a recognition that expectations are exceedingly high on whatever they call their post GPT4 model, and that the company's destiny could actually be tied up in how successful it is or not. For example, OpenAI's prospects rest in part on the eventual launch of a new flagship LLM. That model seeks to improve upon its existing flagship LLM GPT4. By now, other rivals have launched LLMs to perform roughly as well as GPT4. Later in the piece, they write, expectations are running high for Orion as OpenAI looks to stay ahead of its rivals and continue its remarkable revenue growth. Earlier this month, for instance, Google beat OpenAI to launch an AI powered voice assistant flexible enough to handle interruptions and sudden topic changes from users. Despite OpenAI first announcing its version in May. Now nothing in this piece says explicitly that OpenAI is actually out fundraising, and it's entirely possible that the tone of urgency comes just from the reporters, not from their sources. However, it is also possible that the sources themselves expressed a sort of nervousness that might be more emblematic of what's going on behind the scenes at OpenAI. Now, one more piece of the story that was interesting. In a separate post on the information, the publication reported that the company had shown Strawberry to the federal government. From the post this summer, Sam Altman's team demonstrated Strawberry to American national security officials, said a person with direct knowledge of those meetings. As the Information notes quote, by demonstrating an unreleased technology to government officials, OpenAI could be setting a new standard for AI developers, especially as advanced AI increasingly becomes a national security concern. The demonstration could be part of OpenAI's push to be more transparent with policymakers, who could cause the company problems if they feel threatened by its technology. However, they also point out the demonstrations also likely have another purpose, starting a conversation with officials about how the company can secure the technology so that foreign adversaries like China can't steal it, and maybe also take a shot at meta platforms for releasing open weight AI that China and everyone else can access. This is obviously extra interesting in the context of the larger debate around SB 1047 and what the appropriate relationship is between regulators and models. Is it also possible that part of OpenAI's logic is to show that if the big Frontier labs are engaged in this way with the government, particularly around concerns around national security, does it buy the whole industry time for everyone to learn a little bit more about AI before we go off and regulate it? Not sure, but it's an interesting point of speculation. Last note for those of you who are keeping track of who the real leakers are out there, AI for Success on Twitter reposted a tweet from leaker Jimmy Apples from back in November with a picture of the Orion constellation and the caption let's conquer the Cosmos. Ashutosh writes, it's wild that Jimmy gave us a hint about OpenAI Orion last November. Jimmy is probably the only person I trust with OpenAI leaks. Anyways, friends, it seems like as fall is right around the corner we may finally get more real information about Strawberry and even potentially the chance to get our hands on it. But for now, that is the story from here. Appreciate you listening as always. And until next time. Peace, Sam.