Monologue: OpenAI Is Getting Desperate - Better Offline

Summary

Better Offline: Monologue - "OpenAI Is Getting Desperate"

Episode Overview: In the episode titled "Monologue: OpenAI Is Getting Desperate," host Ed Zitron delves deep into the current state of OpenAI, scrutinizing its latest developments, financial health, and strategic decisions. Released on March 7, 2025, this monologue offers a critical examination of OpenAI's trajectory amidst a rapidly evolving tech landscape.

1. Introduction and Upcoming Topics

Ed Zitron opens the episode by setting the stage for future discussions while briefly introducing the focus of the current monologue.

Upcoming Two-Part Series: Ed mentions an upcoming two-part series that will explore Microsoft's data center reductions and the precarious funding situation facing OpenAI.

2. OpenAI's GPT-4.5 Launch and Features

The core of the monologue revolves around OpenAI's latest model, GPT-4.5, and its implications.

Comparison with Previous Models: Ed contrasts GPT-4.5 with its predecessor, GPT-4.0, highlighting the ambiguity surrounding its distinct features.

"GPT 4.5 was the first model that feels like talking to a thoughtful person to him, which makes me wonder what the other models have been like." [00:53]
Sam Altman's Statements: CEO Sam Altman touts GPT-4.5 as a "giant, expensive model" requiring substantial GPU resources, signaling a shift from previous accessibility.

"GPT 4.5 isn't a reasoning model and won't crush benchmarks on account of it being a different kind of intelligence." [05:15]
Technical Ambiguity: Ed expresses confusion over the actual enhancements of GPT-4.5, noting a lack of clear differentiation from earlier models.

3. Cost Implications and Pricing Strategies

A significant portion of the discussion centers on the escalating costs associated with GPT-4.5 and OpenAI's pricing strategies.

Cost Analysis: Ed cites Benj Edwards from Ars Technica, revealing the exorbitant costs of GPT-4.5—$75 per million input tokens and $150 per million output tokens. This marks a 3000% increase for input tokens compared to GPT-4.0.

"GPT 4.5 costs an incredible $75 per million input tokens... that's roughly 3000% more expensive for input tokens than GPT4O." [03:30]
Subscription Models: OpenAI introduces a tiered subscription model where the $200/month ChatGPT Pro and the upcoming $20/month Plus plan will have varying access to GPT-4.5, with potential future costs for high-tier subscriptions.

"Sam Altman posted an idea for paid plans where your $20 plus subscription converts the credits you can use across features like Deep Research, Zero1, GPT 4.5, SORA..." [07:45]
Rent-Seeking Concerns: Ed criticizes OpenAI's approach as a rent-seeking model, aiming to maximize revenue from power users through additional credit purchases.

4. OpenAI's Financial Health and Strategic Desperation

Ed Zhengiton paints a bleak picture of OpenAI's financial stability and strategic direction.

Dependence on SoftBank: OpenAI's main backer, SoftBank, is reported to be borrowing billions to support OpenAI, indicating financial strain.

"The larger picture for OpenAI right now is pretty fucking dire considering their main backer, Softbank, has to borrow billions of dollars to fund them." [04:55]
Desperate Measures: Ed accuses OpenAI of resorting to opaque and convoluted pricing structures to sustain operations, labeling the company's actions as signs of desperation.

"OpenAI is unprofitable, unsustainable and deeply, deeply lost. These are the actions of a desperate company run by a desperate man." [08:55]
Questioning Product Value: The introduction of high-priced agents, such as $2,000/month for low-end agents and up to $20,000/month for PhD-level research agents, is met with skepticism regarding their actual utility.

"What the fuck does that mean? What would it do? Why do these companies. I get emails every week having to justify my fucking cynicism..." [07:30]

5. Public Relations and Communication Issues

Ed highlights OpenAI's questionable communication strategies and public relations missteps.

Inconsistent Messaging: Sam Altman's remarks about GPT-4.5 being "magic" without clear explanations undermine confidence in the model's capabilities.

"It's pretty difficult to find exactly what it is that GPT 4.5 does differently... what it's good at, or indeed really anything about it." [03:15]
Internal Contradictions: Ed points out the disconnect between OpenAI's high expenditures and the transparent communication expected from a leading tech company.

"OpenAI is getting desperate. And now the information is reporting that OpenAI executives have told some investors they will be charging $2,000 per month for their low end agent product." [07:10]

6. Conclusion: A Dire Outlook for OpenAI

Ed Zitron concludes with a stark assessment of OpenAI's future, emphasizing the company's unsustainable practices and leadership challenges.

Unsustainable Business Model: OpenAI's reliance on exorbitant pricing and unclear product value propositions are unsustainable in the long term.

"This is not a company that's cooking. This is not a company that's worked out anything. OpenAI is unprofitable, unsustainable and deeply, deeply lost." [09:00]
Leadership Critique: Ed suggests that Sam Altman's leadership is faltering, lacking the necessary insight and strategic vision to steer OpenAI towards stability.

"These are the actions of a desperate company run by a desperate man." [09:10]
Final Remarks: Despite facing backlash and skepticism, Ed remains committed to analyzing and criticizing OpenAI's trajectory, underscoring the importance of accountability in the tech industry.

Notable Quotes with Timestamps:

Ed Zitron on GPT-4.5's Ambiguity:

"It's pretty difficult to find exactly what it is that GPT 4.5 does differently, or what it's good at, or indeed really anything about it." [03:15]
Sam Altman's "Magic" Statement:

"It’s magic. It’s. It's literally magic. I made magic. Now what does the magic do?" [05:20]
Ed on OpenAI's Desperation:

"OpenAI is unprofitable, unsustainable and deeply, deeply lost." [09:10]
Critique of Expensive Agents:

"$20,000 for a PhD level agent? What the fuck does that mean? What would it do?" [07:45]

Final Thoughts: Ed Zitron's monologue offers a critical lens on OpenAI's recent developments, questioning the sustainability of its business model, the clarity of its product offerings, and the efficacy of its leadership. By dissecting the launch of GPT-4.5 and the accompanying financial maneuvers, Ed underscores the potential pitfalls facing one of the tech industry's most influential players.

Transcript

Eric Andre (0:00)

Hey, what's up y'all? This is Eric Andre. Well, I made a podcast called Bombing about absolutely tanking on stage. I tell gnarly stories and I talk to friends about their worst moments of bombing in all sorts of ways. Bombing on stage, bombing in public, bombing in life. I want to know what's the worst way they've ever bombed? Or have they ever performed way too drunk or high? Or was there ever a time where they thought they were going to crush and they stunk it up? Listen to Bombing with Eric Andre on Will Ferrell's Big Money Players Network on the iHeartRadio app, Apple Podcasts or wherever you get your podcasts.

Ed Zitron (0:41)

Hello you. It's your better Offline monologue and I'm your host, Ed Zitron.

Jay Shetty (0:50)

Better Offline.

Ed Zitron (0:53)

Now. Next week I'm gonna have a two parter that digs into Microsoft's data center pullback and OpenAI's shaky new funding situation. But this week's monologue focuses on OpenAI's new model, GPT 4.5. You may be wondering what it does differently to GPT 4.0 or Claude Sonnet 3.7 or any number of other large language models. And if I'm honest, I have absolutely no idea. Thankfully, neither does Sam Altman, CEO of OpenAI, who said, and I quote, that GPT 4.5 was the first model that feels like talking to a thoughtful person to him, which makes me wonder what the other models have been like. So I went back and look so compare this to the launch of GPT4.O, which Altman called OpenAI's best model ever, saying that it was fast, smart, natively multimodal, referring to the ability to accept text as well as audio and video and photos as well, and available to all ChatGPT users, including on their free plan, adding that it was a very good model, especially at coding. By contrast, Altman summarized GPT 4.5 as a giant, expensive model, one that required hundreds of thousands of GPUs to launch beyond ChatGPT Pro. It started there. It's still not as I record this out to plus users or free users now. GPT Pro, of course, is OpenAI's $200 a month subscription, and it's unclear when plus which is the 20 buck a month one will get it, but apparently it's in the next few days. Altman also added that GPT 4.5 isn't a reasoning model and won't crush benchmarks on account of it being a different kind of intelligence that has and all of these are quot magic to it that Sam Altman had not felt before. Yeah, just, you know, shit's not doing well when you have to just be like, it's magic. It's. It's literally magic. I made magic. Now what does the magic do? I'm really not sure. In fact, it's pretty difficult to find exactly what it is that GPT 4.5 does differently, or what it's good at, or indeed really anything about it. Benj Edwards over at ars Technica had one developer call it a lemonade. GPT 4.5 costs an incredible $75 per million input tokens, prompts and data pushed into a model and $150 per million output tokens, as in the thing it creates. A token is like 0.7. I think one token is maybe three words. Someone will get up my ass for this. Nevertheless, this seems like a lot. It isn't when you're running a company. And by the way, this is roughly 3000% more expensive for input tokens and 1500% more expensive for output tokens than GPT4O for results that OpenAI co founder and Andrej Karpathy described as a little bit better and awesome, but also not exactly in ways that are trivial to point to. That translates to it's a little bit better, but I can't really tell you why. And yes, you're gonna hear me say something similar in next week's episode because the larger picture for OpenAI right now is pretty fucking dire considering their main backer, Softbank, has to borrow billions of dollars to fund them. Nevertheless, back to 4.5 since launch, which was for some reason on the day that Sam Altman's child was being born in. He's been posting some really weird shit since, though. A few days after launch, Altman claimed that GPT 4.5 was the first time people had been emailing with such passion asking that OpenAI promised never to stop offering a specific model or even replace it with an update. At which point I assume everybody in the room started clapping and they saluted Sam Altman and said thank you sir, for making this happen. And by the way, what I'm suggesting is that no one's ever done this, or like one freak did, or maybe Altman emailed it to himself. Just shut up. Just. Your company burns $5 billion a year and the best you've got is this warmed up dog shit about people marine todding you over your model and never taking it away. Has OpenAI ever even taken away a model Jesus fucking CR, these companies. Anyway, a few days later Altman posted a conversation where he asked GPT 4.5 if it believed it was real, leading to a series of bullet points with things like what do we mean by real? Only for GPT 4.5, saying that it believed that not an independent consciousness, but rather a structured experience happening within your consciousness, referring to Sam Altman, which is the kind of shit that's only impressive if you're an imbecile or so stoned you've texted eight of your friends the question what if the Joker was Batman? And by the way, the answer to that is called the Batman who Laughs and it's one of the worst comics ever written. If you want to talk to me about DC Metal, please email me at ez, that's E Z. Or Z if you're Canadian or british@betteroffline.com I really if you are working for DC Comics right now and you had anything to do with Death Metal or the Batman who Laughs, you and I have a grievance. You and I need to talk. Sorry, what? This is a tech pop ups room right back to OpenAI. More worryingly, Sam Altman posted an idea for paid plans where your $20 plus subscription converts the credits you can use across features like Deep Research, Zero1, GPT 4.5, SORA and so on with no fixed limits per feature and you choose what you want. If you run out of credits you can buy more. This, to be clear, is an attempt to raise prices without actually raising them. By attempting to limit usage of OpenAI's more expensive models, ChatGPT plus and other subscriptions give you a limit, for example a limit of 80 messages every three hours on GPT4O. But using one doesn't limit your use of other products here. OpenAI is trying to create a rent seeking model where power users have to pay for more credits if they want to use say OpenAI's more expensive models like Sora and Zero1. And I imagine any situation like this will be one where they hope that people simply won't use their credits or overuse them and have to pay for top ups. This is of course all theoretical, but it heavily suggests that OpenAI is getting desperate. And now the information is reporting that OpenAI executives have told some investors they will be charging $2,000 per month for their low end agent product. And yes, that's a quote sold to. And again I quote high income knowledge workers with supposed mid tier agents for software development costing possibly $10,000 a month with supposed PhD level research agents costing 20,000 dol thousand dollars a month. And I will tell you, the PhDs I know would probably do it for half and they'd even work for an annoying asshole like Sam Altman. Now you may wonder what any of these things do and the answer is that neither I nor the information know. As of right now, the only operational agent OpenAI has is operator OpenAI's agent that sometimes successfully uses a web browser to search for something in minutes, which would usually take you seconds. The information attempted to suggest that the 2,000amonth agent would be some sort of thing that could sort through and rank Sal, but I'm sorry, do I really have to read this shit with a straight face? $20,000 for a PhD level agent? What the fuck does that mean? What would it do? Why do these companies. I get emails every week having to justify my fucking cynicism, but these shitheads, they're allowed to just make up stuff and leak it to the information the information publishes there. We're all meant to be impressed. What the fucking. What the fuck? I'm allowed to rant on these. They're allowing me to rant on these. It's just, it sickens me. I have had this week at least five people email me and be like, well Ed, what would it take to change your mind about this stuff? Why do I have to fucking do it? Why do I? The multi billion dollar companies do a dog shit job of actually explaining this stuff or selling it. They lose billions of dollars, but I'm the guy who has to justify myself. Oh well, I'll keep doing it. Nevertheless, nestled at the bottom of this article is a far more obvious pale horse. OpenAI is planning to charge 20% to 30% of pro customers the $200 a month subscription that loses them money every time a higher price research queries they're doing with Altman according to the information, suggesting some sort of hey, guess what? A la carte or pay as you go approach. I want to be clear about something. This is not a company that's cooking. This is not a company that's worked out anything. OpenAI is unprofitable, unsustainable and deeply, deeply lost. These are the actions of a desperate company run by a desperate man. If only Sam Altman had a thoughtful friend to talk about all these problems to.