The AI Daily Brief: Artificial Intelligence News and Analysis - Why OpenAI's $2000/Month Model Isn't Crazy

Starting point is 00:00:00 Today on the AI Daily Brief, why OpenAI's plan to potentially charge up to $2,000 for its advanced models might not be as crazy as it sounds. Before that in the headlines, are we on the verge of seeing an XAI Tesla 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. When people have explored why Elon Musk's XAI might be an interesting competitor in the LLM space, there's two things you tend to hear. The first is access to the unique data and distribution channel of Twitter slash X,

Starting point is 00:00:43 in other words that GROC gets a source of information that no one else has, or at least no one else has easily, as well as a place to put it, and the other is the potential integrations with the other family of Elon projects. Chief among those has to, of course, be Tesla. A Tesla X-AI tie-up seems like it's more than potential. The Wall Street Journal writes, Musk's X-AI has discussed a deal for share of future Tesla revenue. Under a proposed arrangement, the startup would give Tesla access to its AI technology in exchange for a slice of the carmaker's software revenue. The deal is not particularly complicated, or at least there's not that much

Starting point is 00:01:16 information yet. The information sums it up. The two companies have discussed Tesla licensing the startup's AI models to help the carmaker develop its full self-driving software, and in return, Tesla would share some revenue with XAI. The driver assistance software aims to move Tesla's autonomous driving technology beyond highways onto city streets. Now, Musk for his part, has of course responded on X, saying, haven't read the article, but the above is not accurate, referring to an ex-user summary of the Wall Street Journal piece. He added, there's no need to license anything from XAI, and shortly after he posted, WSJ is talking nonsense. Regardless of the truth of this particular rumor, investors are for sure

Starting point is 00:01:51 watching the relationship between these two companies closely. As the journal puts it, Musk is already shifting talent and hardware between XAI and Tesla, raising concerns about potential conflicts of interest. There is particular scrutiny around how Musk shares resources of Tesla, which is publicly traded. Several Tesla shareholders have filed suits claiming that the shift in resources to XAI has hurt the carmaker's investors. So whether this is a true report or not, it seems like there will be more of this in the future. Speaking of X and Grock, when it comes to that advantage of being able to use tweets as training data, Elon appears to have hit a snag in Europe. Ireland's Data Protection Commissioner or DPC took Twitter slash X to court in Ireland,

Starting point is 00:02:28 the result of which is that X has agreed to stop processing EU users' data for GROC AI, said the DPC. The DPC had significant concerns that the processing of personal data contained in the public posts of XEU and EEA users for the purpose of training its AI GROC gave rise to a risk of the fundamental rights and freedoms of individuals. And these types of rules may be why we're also getting reports that AI lobbying in the U.S. is way up this year. A couple of key stats from Open Secrets, in 2023 organizations spent money on federal lobbying around AI. That's 302 more than 2022 and a 7,567% increase back from 2016, when only six organizations were lobbying the government on AI-related issues. In the first quarter of 2024, this year, Open Secrets identified more

Starting point is 00:03:13 than 90 organizations that reported lobbying on AI issues for the first time. The total so far this year appears to be 462, just a little above last year. Of the organizations that, the organizations that lobbied on AI issues last year. 85 were from the electronics manufacturing and equipment industry, 46 were educational organizations, 31 were from the internet industry, and 22 came from pharmaceuticals and health care. In terms of some specifics, Open Secrets noted that OpenAI spent $340,000 on lobbying fees in the first quarter of this year, with scale spending 180K in that period, and Nvidia spending 160,000 during that time. My general feeling about lobbying tends to be don't hate the player, hate the game, and that it's no surprise to see an increase in this incredibly important area,

Starting point is 00:03:54 especially now as rules remain ill-defined. Now, one thing that we are looking forward to later today, Apple is getting ready to roll out its new iPhones with the promise that AI is right at the center of them. I'm sure that tomorrow we will have a chance to get a better sense of exactly how deeply integrated AI is and when those products are actually coming to market. And speaking of products coming to market and announcements from big tech, on September 16th, Microsoft has scheduled an announcement called Microsoft 365 copilot Wave 2. No more information really than that, but it should be interesting to see how Microsoft positions co-pilot now, as opposed to when it was announced last year. For now, though, that is going to do it for today's AI Daily Brief Headlines

Starting point is 00:04:32 Edition. Next up, the main episode. 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. Discussed anything with Venice without concern about it being monitored, sold, or given to advertisers and governments. Venice is different because your conversations and creations are kept securely within the 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 or taboo. With Venice,

Starting point is 00:05:09 you're in control of the AI as you should be. Pro subscriptions are available for $49 a year or $8 dollars per month. AI Daily Brief listeners receive a 20% discount on Venice Pro. Visit venice.a.I 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 Super Intelligent, 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 slash back to work, we are running our best promotion ever when you sign up for Super Intelligent between now and the

Starting point is 00:05:53 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 Superintelligent between now and the end of the month using code so back, you will get your first month 100% free. Go to B-Super.a.I. And check it out today. Rumors are swirling that OpenAI is considering an extremely high-price plan for its future

Starting point is 00:06:37 more advanced models. And by high-priced, I'm talking $2,000 a month. On today's episode, we are going to talk about why that is somewhat less crazy than it seems. At least I think there are ways to look at it as such. To kick off, however, we're also going to do a little bit of catch-up on previous OpenAI stories that we covered over the last week or so. One of the big stories last week was that presentation from OpenAI Japan that listed GPT Next as 100 times more powerful than GPT4. If you remember in that episode, one of the things that I was wondering is whether

Starting point is 00:07:07 GPT Next was referring to an actual specific model or just a category of GPD Next. general things in the future. Now, I noted that even if it was that, which seemed most likely to me, it probably wasn't a great idea to put that image on a slide. And indeed, it does appear that GPT Next is not a new model, but just a general future state. From Mashable, an open AI spokesperson has confirmed to Mashable that the term GPT Next, written in quotations on the slide, was simply a figurative placeholder to indicate how OpenAI's models could evolve exponentially over time. The spokesperson also clarified that the line graph in the slide was illustrative, not an actual timeline of OpenAI's plans. So, so much for getting a 100x more powerful model this year.

Starting point is 00:07:46 Another interesting story that's been swirling around is OpenAI's latest raise at something like a $100 billion plus valuation. Thrive Capital has been widely reported to be the lead of that round, leading to a recent piece in the information called why Josh Kushner became OpenAI's fundraising Sherpa. The piece begins, in mid-July, CEO Sam Altman, flew into Idaho to mingle with entertainment and tech elites at the annual Sun Valley retreat. Among the those by Altman's side was Josh Kushner, founder of Thrive Capital, the New York VC firm that had led a recent share sale valuing the chat GPT maker at $86 billion. Less than a month later, Altman gave Kushner a pat on the back, replying to news that Thrive had just raised $5 billion in new

Starting point is 00:08:24 funds, Altman posted on X that among the, quote, many great investors he's worked with, there's no one I'd recommend more highly than Josh. So what else do we learn from this round? Well, a couple of things. First, it sounds like the valuation might be up over $120 billion, not just $100 billion. Second, it sounds like part of the role that Kushner is playing is not just investing capital, but also organizing more capital. Again, from that same piece, OpenAI requires so much capital that Kushner could play a key role in arranging funding as well. Others also speculate that it's at least a little bit about raising the profile of Josh Kushner

Starting point is 00:08:54 and Thrive in general. The founder of another large VC firm said that Thrive's massive investment is, quote, Josh's effort to be the guy you call an AI. It makes him very visible and gives Thrive the appearance of having, quote, unlimited check-writing power. Interesting stuff if a little bit insider baseball, but not the thing that we want to focus on today. What we want to focus on is shifts in how OpenAI is thinking about its business model, and one more story before we get to this theoretical $2,000 price point.

Starting point is 00:09:19 At the end of last week, OpenAI announced that it had hit 1 million paid users for its enterprise offerings. That's up from 600,000 corporate chat GPT accounts back in April. In terms of additional real information, there's not that much. For example, one of the things that we don't know is whether the increase came from new clients or expansion within organizations. We did find out that just under half of the corporate users are based in the United States, with Germany, Japan, and the United Kingdom being the other popular markets. All right, and so this gets us to this $2,000 a month report. This comes once again from the information, where they wrote last week, how much would you be willing to pay for

Starting point is 00:09:56 chat GPT every month? 50, $75? How about $200 or $2,000? That's the question facing Open AI, whose executives we here have discussed high-price subscriptions for upcoming large language models, such as OpenAI's reasoning-focused strawberry and a new flagship LLM dubbed O'Ryan. The piece continues, in early internal discussions at OpenAI, subscription prices ranging up to $2,000 per month were on the table, said one person with direct knowledge of the numbers, though nothing is final. Now, the information seized on the idea that this represented,

Starting point is 00:10:23 an indication that revenue wasn't keeping up with costs. They write, it suggests that the paid version of ChatGBT, which was recently on pace to generate $2 billion of revenue annually, largely from $20 per month subscriptions, may not be growing fast enough to cover the outsized costs of running the service. They continue, and more advanced models such as Strawberry and Orion may be more expensive to train and run than prior models. The flip side, they say, is that a high price point would also mean OpenAI believes its existing white-collar customers of ChatGBT,

Starting point is 00:10:51 will find these upcoming models a lot more valuable to their coding analytics or engineering work. And this is largely the two sides of the conversation that you saw on Twitter slash X. X user Bob E. Hayes writes, Tell me things are bad without telling me things are bad. Report, OpenAI considers $2,000 monthly LLM subscriptions. OKE egg writes, OpenAI's $2,000 a month feels like a clumsy way to address the dilemma. One, achieving AGI requires monstrous investments. Two, need a mass consumer product to recoup.

Starting point is 00:11:19 Three, must limit genuine AGI to a select elite. AI entrepreneur Bendi Ready writes, OpenAI is considering 2K a month to access their top models. All kidding aside, this will be a Vision Pro-level disaster. I hope it's a joke. Digital spaceport writes, a $2,000 a month subscription is going going to appear to those on lower tiers very negatively. Rich-get-richer stratification, maximalism is how it will be labeled. So in these few tweets, you get a pretty good breadth of the conversation on the negative side,

Starting point is 00:11:45 that on the one hand, maybe this suggests that there's trouble with the Open AI business model, or even if there's not, that it would somehow represent a failure of the market to only make the most advanced models available to those who already have every advantage, at least financially speaking. That wasn't the only take, though. Professor Ethan Malik writes, I have no idea if the 2K a month price rumors about advanced open AI models are true, but it would represent a move from thinking about AI as a co-intelligence for each person to

Starting point is 00:12:10 AI as independent worker. And I think this is the nut of it. Swicks, for example, writes, I think it is a very bullish sign that OpenAI is considering $2,000 a month subscriptions. This is what taking AI employees seriously looks like, a level 3 AI product. Now, what he's referring to is OpenAI recently announced, or better put recently leaked, levels or stages of artificial intelligence. Level 1 is chatbots, which are AI with conversational language. Level 2 is reasoners with human-level problem-solving. Level 3 is agents or systems that can take actions.

Starting point is 00:12:41 Level 4 are innovators AI that can aid an invention. And level 5 are organizations or AI that can do the work of an organization. And basically, Sean's argument here is that if an AI is powerful enough to do the work of a human, $2,000 a month starts to look a lot more reasonable. You see that conversation play out everywhere. On that same Bindu Ready post that I mentioned before, someone responded if it can replace an entry-level software engineer that's still a four-to-six-six cost reduction. To which bin due responded, it can't. And so indeed we have some skepticism of actual

Starting point is 00:13:10 capabilities here as well. John Gilhuli writes, what is the next limit of human productivity? OpenAI's rumored $2,000 a month price tag moves it from consumer tool to employee replacement. This got me thinking. If the new reasoning models could work perfectly, could they actually replace an employee? They'll still need someone giving them direction, so the benefit has to be either in the quality of output or the speed, likely both. Imagine having a magic button that completes most base work for you in a few seconds. You'd still need to give clear instructions and review outputs, as you'd have with most entry employees, but now your cycles are much, much faster. Suddenly, you're going from making a few decisions a day, to dozens or hundreds. Your quote-unquote

Starting point is 00:13:44 workers no longer need time to go off and complete tasks. They just immediately deliver results and can move on to the next decision. Decision fatigue is going to become the new productivity limiting factor. The human brain can only handle so many decisions before it starts to falter. Sure, AI will continue to improve in its ability to handle more high-level instructions like Find Me More Users, but it will still require direction. It will still be a while before we reach the its AI turtles all the way downstage of the world. And between here and there, we'll have a world where those who can make hundreds of well-reasoned decisions per day are going to outperform those who simply produce high-quality granular work. Mastering the art of rapid, continual reasoned

Starting point is 00:14:17 decision-making could become the next career challenge for many. Now, this moves us into a conversation around what the destiny of today's workers is in a world where we really have AIs that can do independent work. Later this week or this weekend, we'll do a long-read style episode from a piece from March from Noah Smith called plentiful high-paying jobs in the age of AI. However, at this point, you may be also asking, what are the chances this is just a dumb report, and people are speculating about nothing? It am a cloud jokingly summarized it this way. He said, the information says, the absolute highest and most absurd price floated for chat GPT was 2000 a month, but the idiot open AI employee who suggested that has now been fed to the basement Shogoth and will never speak again.

Starting point is 00:14:55 Seven OpenAI spokespeople have reached out and confirmed they will not price it this high. Twitter, holy crap, OpenAI is crazy. Did you hear they're going to ship strawberry tomorrow for 2K a month? Morons, nobody will pay that. To which Sean Ralston, who I believe used to work at OpenAI, although I don't think does anymore, responded was an internal convoy about enterprise accounts with hundreds of licenses, not individual users. And whatever the right of this Sean has or not, it certainly seems possible that this is a big report about a very exploratory meeting where someone decided to humor someone else and write $2,000 a month down on a whiteboard, and that became a whole story that we're discussing. However, I think why it's resonant and worth the discussion is that it

Starting point is 00:15:30 really does get us thinking, again, as Ethan Mollock put it, about the move from AI as co-intelligence for each person to AI's independent worker. There are some who believe that will never happen. There are many who believe it will happen in a fitful, non-linear way, but it's certainly a conversation that's going to come up a lot more as capacities increase. For now, I think it's fairly safe to say that at these levels, I don't think we're going to be seeing a $2,000 a month chat Shp.T. Anytime soon, but you never know. Stranger Things have happened. That is going to do it for today's AI Daily Brief.

Starting point is 00:16:00 Until next time, peace.

The AI Daily Brief: Artificial Intelligence News and Analysis - Why OpenAI's $2000/Month Model Isn't Crazy

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.