The AI Daily Brief: Artificial Intelligence News and Analysis - A New OpenAI Model Coming?

Episode Date: August 13, 2024

Rumors are swirling around OpenAI’s potential release of a new advanced reasoning model, codenamed “Strawberry” or “Q-Star.” Speculation is fueled by tweets from an anonymous account claiming insider knowledge, as well as cryptic hints from OpenAI’s own team. Plus an epic future of search team up between Perplexity and Polymarket.

Today on the AI Daily Brief, is OpenAI's more capable reasoning model coming tomorrow? Before then in the headlines, perplexity and polymarket come together for an epic future of search team-up. 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 edition, all the daily AI news you need in around five minutes. We kick off today around one of the themes that we've been exploring a lot lately, which is the future of search. It seems like there is a big bet that in a world newly enabled by generative AI, search is not just going to be a set of links like it is now with Google and like it has been for 20 years with

Google, but instead a combination of that and some amount of curatorial synthesis that gives you answers in complete written-up form. However, a new partnership from Perplexity and Polymarket shows that going forward search may also include other types of data in new ways. They tweeted earlier today, we're thrilled to announce our partnership with Polymarket. Now when you search for events on perplexity, you'll see news summaries paired with real-time probability predictions, such as election outcomes, market trends, and beyond. So Polymarket, for those of you who don't know, is a prediction marketplace that lets people bet on real-world events. It is the latest type of technology platform to draw the ire of Elizabeth Warren, and it is now going to be integrated into perplexity.

TechCrunch writes, when users click on an event on Polymarket, they will now see a summary of news related to the event based on search results from Perplexity. There's also a search box that you can use to ask more questions. Perplexity will also use some data from polymarket, such as election trends, to show visuals and answers. Now, interestingly, in this news story, there's also a little bit of insider baseball when it comes to Perplexity's business model. The chief business officer over at Perplexity told TechCrunch that, quote, the API business is not a priority for us as we are a consumer-focused company. But still, API usage is growing with more than 25,000 developers using our API. We have a unique offering that pulls in answers via different internet sources. The way we think about

API right now is that it's a means to grow the brand, but not the end. Could we see a perplexity that diversifies its consumer business model with a more B2B model through their API in the future? Certainly doesn't seem insane. Shane Copeland, the CEO of Polymarket, said, when I first built Polymarket, I would tell people I was building the search engine for the future. Now that's coming to life. Next up, another little team up to discuss. Invitya is collaborating with the state of California. to train California residents on AI. The partnership between the state and the company is aiming to help train 100,000 students,

college, faculty, developers, and data scientists to harness AI. Business Insider writes that as part of that training, educators in California can get a certification from the Nvidia Deep Learning Institute University Ambassador program. But of course, as much as this is about Nvidia making a bet on the state of California, it also comes at a time that California's relationship with AI is at an inflection point due to SB 1047.

If you want the background on what's going on that, go check out the episode that I dropped last week. I tried to provide a pretty broad overview without too much editorialization, although of course I couldn't avoid it entirely. But to those who are against it, they think that it could have major implications for whether businesses actually stick around to do business in California when it comes to AI. To more California companies in our next story, TechCrunch reports that Meta and Universal Music group have updated their licensing agreement and in it started to deal with AI music. TechCrunch again writes, Meta and UMG announced on Monday the expansion of their multi-year licensing agreement,

which enables users to share songs from UMG's music library across Meta's platforms without violating copyright. What's most notable about the new agreement is that it states that the two companies are addressing unauthorized AI-generated content. This refers to songs being scraped by AI systems, often without the consent of the original creators. Said Michael Nash, chief digital officer and EVP at UMG, we look forward to continuing to work together to address unauthorized AI-generated content that could affect artists and songwriters so that UMG can continue to protect their rights both now and in the future. Meta is treading carefully when it comes to AI-generated music, as TC writes, they're only releasing generative AI models, including audio craft music gen, and

Jasko that are trained with, quote, meta-owned and specifically licensed music. Now, I continue to think that markets are going to address this particular issue and that the record labels are going to do what they always do, which is swoop in to get a big cut of the new area, but exactly how that plays out we haven't yet seen. Lastly today, two updates from big tech companies in their AI strategy. analysts believe that Apple could charge up to $20 for some of its Apple intelligence features. Now, of course, Apple intelligence is being delayed slightly from when it was anticipated rolling out with iOS 18, but now analysts are speculating that these features are not going to be for free. Will that impact how much mainstreaming they actually do? Or is Apple such a

behemoth that even with a subscription model, they're still likely to be able to bring on a whole new raft of users? Finally, and yet another indication of how hard it is to be a startup creating AI features, Google Meet has announced a new Take Notes for Me feature, which seemingly imitates the functionality of a ton of startups out there who are AI note-taking apps for things like Google Meet and Zoom, but natively within the Google workspace. So is this a firefly slash otter killer? Does it just mean that all AI startups are destined to be replaced by functionality from the big tech companies? Only time will tell. That's going to do it for today's AI Daily Brief Headlines edition. Next up, the main episode. 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 company's 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. AIA Daily Brief listeners receive a 20% discount on Venice Pro.

Visit venus.a. slash NLW and enter the discount code, NLW Daily Brief. That's NLW Daily Brief. All one word. 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, to Fractional. The 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 superintelligent platform and we are working on something new as well. 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 Super Intelligent, the platform that helps teams maximize AI. Super is, of course, the platform that we've been building that pairs fun, fast, practical video tutorials with step-by-step instructions to get you actually using AI. And from there, unlocks information about hundreds of use cases that show you how people, are actually getting value out of AI right now. Now, we have just launched Super for Teams. This is a new add-on experience that allows teams to share more information about what AI

they're using, how it's working, and how to get more value out of it. Whether your company is 25 or 2,500, superintelligent is going to be the best platform for unlocking information about how to get the most out of AI right now. If you'd like to learn more about the superintelligent teams offering, go to BESuper.a.I. slash partner and send us a note so that our team can get right back to you. Once again, that's B-super.a.ai slash partner. Welcome back to the AI Daily Brief. We started talking about this last week, but boy, has the hype and rumor train around OpenAI's next model gotten nothing but louder since then. The main character in this particular drama is a Twitter account called at I Rule the World M.O.

This person or perhaps AI has been tweeting nonstop for a couple of weeks next to with increasingly specific predictions, interspersed with big, broad, grand vision tweets like this one. Imagine a world where every human has access to the power of a thousand minds. That's the promise of Project Strawberry, and it's closer than you think. So today we're checking in on the rumors, which of course could just be rumors and everything we know so far. So by way of background, Strawberry is apparently the latest name for what was originally called Q-Star. You might remember Q-Star from around the time that Sam Altman was fired from OpenAI, before being rehired, there were tons of speculation that this more advanced model had freaked some

people out internally, and they were concerned around how Sam Altman was likely to go try to commercialize it, and that was the motivation for his firing. Now, subsequently, no one has tried to make that claim, but try to tell that to the internet denizens. Then back in July, Reuters reported that the new name for Q Star was called Strawberry, and that yes, indeed, it was about a better reasoning model. All of this further heated up last week, when a new model showed up on the Limsys chatbot arena, which was called Anonymous Chatbot. Now, OpenAI has a history of releasing their models as this sort of anonymous chatbot

in advance of them being officially released until people were looking for evidence that this might be Q Star or now Strawberry. writes NDTV, Anonymous chatbot uses chain of thought reasoning, a stark step away from the current methods of reasoning like pattern recognition. If new chatbots are going to start showing up that use chain of thought and their reasoning capabilities

were likely to see another shift in the AI race. Then, of course, there's the fact that OpenAI's team seems to be part of the team. Sam Altman tweeted, I'd love summer in the garden with a picture of some strawberries. And when that same anonymous user I ruled the world wrote, welcome to level two. How do you feel? Referring to Level 2 on OpenAI's new chart of the path to AGI, which is reasoning level, Altman replied to that account, Amazing, TBH.

As Tom's guide points out, however, there's another reason to think that maybe OpenAI would do this this week. And that is that there is a Google event this week, made by Google, which happens tomorrow on Tuesday, August 13th. Tom's writes that the event has very few unknowns when it comes to what Google will likely announce, given that they've already shown off the Pixel 9 Pro and Pixel 9 Pro fold in videos, and so the speculation is whether the focus comes back to AI. As Tom's points out, Google has lost some of their steam when it comes to AI on the phone, and certainly as part of the larger generative AI conversation, and speculates around what some of the AI features that they might announce instead would be. So is Open AI just going to try to cut the media legs out from under this

Google event? Well, that same anonymous account, I Rule the World, has said that tomorrow at 10 a.m. Pacific Time is when some big announcement from OpenAI will be made. This morning, they wrote, attention isn't all you need. New architecture announcement, August 13th at 10 a.m. Pacific Time, the singularity begins. A couple days before that, that account had given more of what it argued was its insider information. On Project Strawberry slash Q Star, they said, this is what Ilya saw. It's what has broken math benchmarks. It's more akin to RLHF than throwing compute at the problem. Think of it as an LLM fine-tuned to reason like a human. This account also argues that GPT Next, which they say is internally called GPTX, but what we would call GPD5 is, quote, also ready to go.

Lots here relies on safety and what Google does next. It's difficult to say if competition will trump safety. Though red teaming is finished and post-training is done, this model is such an enormous leap in capabilities that's becoming impossible to make the model safe. If you have this particular model unlocked, you could easily disrupt the world on an unprecedented scale. When you mix in voice, SORA agents in the eye-watering capabilities, things heat up. They'll get the safety right and then they'll roll it out, I'm sure. They also say that in addition to it just being really expensive, those safety concerns are also delaying SORA. Ultimately, they say, we will get to a step change next week. It won't quite be GPT5. GPD 5. GpT.5 slash next slash X is more comparable to the jump

made from GPT1 to 4. This is why Sam feels great. Ilya was right. You can scale your way to a digital god with or without strawberries, but strawberries plus scale will cure world problems overnight. Now, there is more than a little skepticism out there. Google's Francois Chalet writes, there have been AGI achieved internally rumors spread by OpenAI every few weeks or months since late 2022, and you guys are still eating it up for the endth time. If you were actually close to AGII, you wouldn't spend your time shit posting on Twitter. Logan Kilpatrick, formerly of OpenAI now at Google, writes, anyone who thinks I rule the world MO is AGI or some advanced model has no idea how OpenAI operates as a company. News outlets are now trying to

dig into who this anonymous account is. James Poulos at Blaze writes, text most mysterious and controversial new account followed me on X. Is it the Antichrist? Is it a bot? Is it a person the internet can't say for sure? And techies are going nuts arguing over it all. Meet I rule the world M.O. An out of nowhere account that almost immediately drew replies from the likes of Sale Maltman, due to its seemingly acute insider knowledge of impending breakthroughs and so-called artificial general intelligence. The neuron pointed out that it was far from conclusive that the anonymous chatbot on Limsys chatbot arena was actually strawberry or even particularly good. As they write, reviews are mixed. Some see it as a

potential strawberry drop with improved logic. Others think it's a minor upgrade to chat GPT4O with little evidence of reasoning. Sometimes it performs better than other models. Sometimes testers prefer others. Some commenters even think chat GPT may have already received a secret upgrade. And indeed you see that if you float around on Twitter slash X as well. Hyperite CEO Matt Schumer writes, something might be going on with GPT40. For the first time, in a long time, it provided better vibes on an output than 3.5 Sonnet. Really surprised, we'll keep using it today to see if it continues. That's from earlier today Monday, August 12th. Maybe even more interesting were these tweets from Perplexity CEO Aravan Shrinivas that seemed to imply that they're running Strawberry slash Q Star. Again from Monday when

someone tweeted, Perplexity Pro is already running the strawberry. Arvon quote tweeted it with the eyes emoji and three strawberries. On August 11th, he posted a screenshot of a perplexity query around the question, how many R's in there are many strawberries in a sentence that's about strawberries? He added the caption, Guess What Model This Is? Again, Strawberry Emoji. He also had the classic psycho-entrepreneur phone battery life of 6%. As the AI safety memes account pointed out, Arvund is also tweeting out various examples purporting to show strawberry in action. One of them from a day ago, the berries are scheming, is the caption. Even for OpenAI, which does a lot of hinting and likes to release things a little bit in advance of saying that they're releasing them,

this is a lot of hinting. At this stage, while I don't have a ton of sense of who's behind this meme account and don't want to give it too much credence, it seems highly unlikely that we're not going to see something this week. At the same time, I also think it's pretty unlikely that most people won't be disappointed based on just how hyped it has gotten. However, as soon as it comes out, whatever it is, you will of course hear about it first. Until then, appreciate you listening or watching as always. And until next time, peace.