Scam Fraudulentman is on the way to become the first trillion dollar scammer in recorded human history & very close to that title. Clearly worse than the theranos scammer lady.
There is another signal of how risky this is: Apple has decided to stay out of it. Bloomberg had an analysis of this earlier this week showing that Apple's stock was moving up while others in the AI industry space stocks were falling. Here is how I see it: Apple also sees the lack of a moat, also known as sustainable competitive advantage, for OpenAI. Apple also understand the role of China: both in terms of building data center capacity and evolution of AI system innovation. In the longer term (possibly mid-term) those advantages that China has will have to play out. It will lead to price reductions in the AI industry making profit projections for OpenAI and other really hard to meet (no U.S. sanction or trade barriers will save them). Apple also understands that owning the final customer through control of the delivery device (and the innovation to come there at AI-infused products evolve) is the best place for Apple to play since it has a history of creating breakthrough high-end premium products in that space. Finally, when the next deep system software innovation arrives in AI (see Yann LeCun here), this whole thing will blow up into a massive market implosion. The next level of innovation will likely be cheaper, faster and better...it always is.
You're not wrong, but I'd like to point out that there is no mythic monolithic "China" to be worried of; with the recent copyright (or rather: copyleft) rulings, distillation of frontier models is fine for everybody. Together with the fact that reasoning models - to the extent that it works at all - can be distilled down into quantized single-mode models, we'll see more and more open weight models at the same "quality" level, but way cheaper and even in edge computing. Point is: You don't have to invest 110 billion, when you can just run some prompts on free OpenAI accounts*.
* They cannot and will not remove a free option. But that's not even the problem. A fresh subscription with a fresh or fake account in any way gives you enough tokens before banning that you can extract a substantial part of whatever "edge" they may have. And since OpenAI has to retrain their stuff all the time just to stay clear of the world moving about, their costs will not go down.
Great stuff!!! Thank you!!! I am not an expert, but I want to keep an eye on so-called small language models that can run on a distributed basis (PC I am thinking) and deliver value in AI-infused consumer devices. My guess is Apple is studying this, maybe connected to the Apple Watch.
I can see no reason why it will not become a commodity; just how will it be priced and paid for over the developing years. There are no real barriers to entry other than brainpower. Information technology will continue to become cheaper and more powerful for a while. I just wonder what products (wearables?) Apple will infuse AI with and take advantage of what will likely be declining system access costs. I have piece in my Substack on the longer term if you are interested. Gerald968 on Substack. Thank you!
It's ideological at this point, those people are voluntarily committing themselves into a death spiral just in hopes they see something out of it that proves their world view. If you listen to the way Sam compared humans to AI in India concerned you will clearly see it.
OpenAI's prospects of achieving AGI in the next year are nil. The fundamental approach being pursued by OpenAI has no chance at all of achieving it EVER! It would take some brand new architecture that has yet to be described, determined, or tested. Scaling will not do it. They are building language models, not intelligence models and there is an insurmountable chasm between the two.
They do not have a way even to judge whether they have achieved AGI. The best method they have is to claim that the model passes benchmark tests. But that method is a fallacy. There are many ways that one can pass a test (including memorizing answers) and one cannot infer the means from the observation that test was passed. Reciting words consistent with passing a test does not imply that the model has the intelligence they wish to infer.
Every benchmark includes tasks that are easy to score, but intelligence implies discovery of new insights, new solutions to unstructured problems, new open-ended questions. These cannot be evaluated using simple, easy to score, tests.
They may claim to have solved AGI, but so has Anthropic until they were confronted with owning up to it at the Pentagon. Amodei had to admit that his model was not up to human-level intelligence. He had to admit that language models were not up to the task of autonomous target selection. Would you be more willing to entrust ChatGPT with a weapon? I would not be willing to trust it even with a credit card.
One more data point. According to Statista, the entire market for AI is projected to be $347 billion by OpenAI is valued at more than twice the value of the entire market ($840 billion). How does that work? The mind boggles.
"Too big to rescue" maybe. These firms aren't key to the economic system as the banks and insurers were in 2008. Dotcom failures weren't;bailed out either.
First: "Stock prices follow earnings. Always have. Always will" is a statement that misses the vast majority of history in the stock market (possibly all of it). The stock market is speculative. Always has been. Always will be.
Second: It's a bet for global domination.
Money doesn't matter when you control the keys to a tool that matches the description of a James Bond or Marvel villain's ultimate weapon. You can hold the infrastructure of every government, every business, every person hostage. You can target specific individuals in every direct and insidious way to ensure their lives fall apart - based on a behavioral profile that doesn't even require them standing up to you.
That's what they're betting on. They want the keys to AGI. They want the ability to create all marketing material without the input of artists. They want the ability to squeeze every spare second and dollar from every person's lives who could fuel their power machine.
Now, there are problems with that "ultimate weapon" that would keep it from being as effective as they are betting it'll be. But they're big picture people - they're not worried about the nuances of reality by and large.
But it's now just like every other insane investment scheme in history listed above - just so much worse. It's "too big to fail" because so many companies have invested into it. But it's also not too big to fail, because fail it will - just like every other insane investment scheme in history. The question is, what do we do when the bubble generating 1/3 of the entire US's economy turns out to be a circle jerk of failed companies?
My bet is on the bipartisan action that's happened in the past few decades...
- the dotcom burst resulted in major bailouts for corporations
- the 2008 housing bubble burst resulted in major bailouts for corporations
- the covid pandemic resulted in major bailouts for corporations
No new regulations. In fact, there were new protections for the corporations who took advantage of American citizens. And insanely large bailouts on unrealized gains.
We are paying significantly more for computer parts so that computer companies can develop software that will force us out of jobs so they can control us more while screwing with our economy so that our government decides to use our money to bail them out when it inevitably fails.
Not according to every person who is professionally building it. "Easy button" AI, which is crucial for downsizing beyond what they've already done, is not a thing, nor will it ever be without AGI.
The one that will be standing strongest at the end of the GenAI hype, seems to be Google, who can fund this from their ad monopoly position, and weather the upcoming shakeout. And they seem to have been more clever technologically, with their TPU hardware (and I noticed during a test they did caching as part of their RAG, also a lot cheaper). Google has been focused a lot on cost of inference.
OpenAI probably survives 2026 on investments/subsidies. The moment of truth may then come in 2027.
What to think of the following outcome: Google survives as the top player in GenAI, a market that ends up to be a lot smaller than thought with no AGI. Claude and OpenAI fail financially, and scraps of each get gobbled up by one of Microsoft, Amazon, maybe Meta, maybe Apple. Coreweave and others go belly up and the scraps get taken over by established players. Nvidia survives bit its valuation comes back to earth. Yes, there will be use for their chips, but they will be far less profitable for a long period. Oracle takes a huge hit, but the Ellisons are already fleeing into media (destroying the plurality of information) and their database business keeps making money.
When the bubble bursts, a lot of investments, a lot of it circular, will turn out to be smoke. After the smoke has cleared, some will remain and grow.
At this point, from a financial perspective, I would be surprised if OpenAI independently survives beyond 2027 and the same is probably true for Anthropic. But who knows, maybe the train will keep running a bit longer burning all that cash. And who knows, maybe lots of companies will get into coding, expecting to replace most of their IT engineers with a $200/month subscription and it will take a while before that turns out to be a mistake.
I still (since 2023) expect the bubble to burst in 2027 plus or minus a year. I wonder if OpenAI and Anthropic aren't too late with their IPO to profit from the craze. And they can't go early because their numbers aren't good.
Funding any of these businesses doesn't make sense and neither does their over the top announcements. OpenAI has and will always have competition from Anthropic, Google, xAI, and others besides the ever improving Chinese models. Competition constrains their ability to raise prices. If their jobs armageddon is true (I have serious doubts) they'll be looking at massive deflationary pressures making things even worse. Ads will erode credibility. They're burning money like an old fashioned coal-driven engine burns coal and there is no clear path to profitability - for any of them, especially the b2c model OpenAI has focused on.
The answer is yes, it makes total sense, if the company is a scam.
The end-state here is an IPO. The current "invenstors" get to cash out and your 401K ends up holding the bag. A very, very, big bag. Sam and Dario should be in jail.
I don't think we'll see an OpenAI IPO. While you're right that some fonds may have to buy into it for structural and ideological reasons, I seriously doubt that they could raise a TRILLION for... nothing.
These people (Musk, Amodeo, Altman, Andreeson, Kurzweil and others) are basically insane religious fanatics.
Their religion is the "singularity", the idea that we will develop a sentient AI that will take over the universe. Not the world, the universe. They claim it is coming very soon and we must prepare now.
Climate change doesn't matter, rampant disease doesn't matter, inequality and social unrest don't matter. Nothing matters but making sure that the singularity happens because that will solve all other problems and make us all into demi-gods (or dirt, depending).
Now, maybe they don't really believe it. On the other hand they have spent lots of time and money on books, white papers, symposia, think-tanks, and related propaganda discussing it. Dario quit OpenAI supposedly because they weren't worried enough about "AI Alignment", which is making sure God, when we invent it, doesn't kill us all right away.
So I'm not holding out a lot of hope that they will behave rationally.
Time will tell but I suspect the fantastical narrative that AGI will be realized (based on LLMs) will prove to be the most costly financial mania mass delusion in US financial history, perhaps in all financial history. When this massive credit bubble blows, as they all do, so will associated nonsense narratives.
Worth noting, even if AGI were to be realized this financial mania will collapse because it is based on decades of monetary and fiscal distortions, meaning choices that will prove unsound, which must be wrung from the system at some point. All such cycles are fundamentally the same.
Sounds like a ponzi scheme, looks like a ponzi scheme, so it must be a ponzi scheme.
Scam Fraudulentman is on the way to become the first trillion dollar scammer in recorded human history & very close to that title. Clearly worse than the theranos scammer lady.
So that tested the blood, this one sucks it.
“Achieving AGI” … only the lawyers will make money on that.
There is another signal of how risky this is: Apple has decided to stay out of it. Bloomberg had an analysis of this earlier this week showing that Apple's stock was moving up while others in the AI industry space stocks were falling. Here is how I see it: Apple also sees the lack of a moat, also known as sustainable competitive advantage, for OpenAI. Apple also understand the role of China: both in terms of building data center capacity and evolution of AI system innovation. In the longer term (possibly mid-term) those advantages that China has will have to play out. It will lead to price reductions in the AI industry making profit projections for OpenAI and other really hard to meet (no U.S. sanction or trade barriers will save them). Apple also understands that owning the final customer through control of the delivery device (and the innovation to come there at AI-infused products evolve) is the best place for Apple to play since it has a history of creating breakthrough high-end premium products in that space. Finally, when the next deep system software innovation arrives in AI (see Yann LeCun here), this whole thing will blow up into a massive market implosion. The next level of innovation will likely be cheaper, faster and better...it always is.
"those advantages that China has"
You're not wrong, but I'd like to point out that there is no mythic monolithic "China" to be worried of; with the recent copyright (or rather: copyleft) rulings, distillation of frontier models is fine for everybody. Together with the fact that reasoning models - to the extent that it works at all - can be distilled down into quantized single-mode models, we'll see more and more open weight models at the same "quality" level, but way cheaper and even in edge computing. Point is: You don't have to invest 110 billion, when you can just run some prompts on free OpenAI accounts*.
* They cannot and will not remove a free option. But that's not even the problem. A fresh subscription with a fresh or fake account in any way gives you enough tokens before banning that you can extract a substantial part of whatever "edge" they may have. And since OpenAI has to retrain their stuff all the time just to stay clear of the world moving about, their costs will not go down.
Great stuff!!! Thank you!!! I am not an expert, but I want to keep an eye on so-called small language models that can run on a distributed basis (PC I am thinking) and deliver value in AI-infused consumer devices. My guess is Apple is studying this, maybe connected to the Apple Watch.
Do you think Apple is waiting for ai to become a commodity?
I can see no reason why it will not become a commodity; just how will it be priced and paid for over the developing years. There are no real barriers to entry other than brainpower. Information technology will continue to become cheaper and more powerful for a while. I just wonder what products (wearables?) Apple will infuse AI with and take advantage of what will likely be declining system access costs. I have piece in my Substack on the longer term if you are interested. Gerald968 on Substack. Thank you!
It's ideological at this point, those people are voluntarily committing themselves into a death spiral just in hopes they see something out of it that proves their world view. If you listen to the way Sam compared humans to AI in India concerned you will clearly see it.
Exactly!
OpenAI's prospects of achieving AGI in the next year are nil. The fundamental approach being pursued by OpenAI has no chance at all of achieving it EVER! It would take some brand new architecture that has yet to be described, determined, or tested. Scaling will not do it. They are building language models, not intelligence models and there is an insurmountable chasm between the two.
They do not have a way even to judge whether they have achieved AGI. The best method they have is to claim that the model passes benchmark tests. But that method is a fallacy. There are many ways that one can pass a test (including memorizing answers) and one cannot infer the means from the observation that test was passed. Reciting words consistent with passing a test does not imply that the model has the intelligence they wish to infer.
Every benchmark includes tasks that are easy to score, but intelligence implies discovery of new insights, new solutions to unstructured problems, new open-ended questions. These cannot be evaluated using simple, easy to score, tests.
They may claim to have solved AGI, but so has Anthropic until they were confronted with owning up to it at the Pentagon. Amodei had to admit that his model was not up to human-level intelligence. He had to admit that language models were not up to the task of autonomous target selection. Would you be more willing to entrust ChatGPT with a weapon? I would not be willing to trust it even with a credit card.
One more data point. According to Statista, the entire market for AI is projected to be $347 billion by OpenAI is valued at more than twice the value of the entire market ($840 billion). How does that work? The mind boggles.
We’ve officially entered the “too big to fail” territory. Which was their strategy all along.
"Too big to rescue" maybe. These firms aren't key to the economic system as the banks and insurers were in 2008. Dotcom failures weren't;bailed out either.
First: "Stock prices follow earnings. Always have. Always will" is a statement that misses the vast majority of history in the stock market (possibly all of it). The stock market is speculative. Always has been. Always will be.
Second: It's a bet for global domination.
Money doesn't matter when you control the keys to a tool that matches the description of a James Bond or Marvel villain's ultimate weapon. You can hold the infrastructure of every government, every business, every person hostage. You can target specific individuals in every direct and insidious way to ensure their lives fall apart - based on a behavioral profile that doesn't even require them standing up to you.
That's what they're betting on. They want the keys to AGI. They want the ability to create all marketing material without the input of artists. They want the ability to squeeze every spare second and dollar from every person's lives who could fuel their power machine.
Now, there are problems with that "ultimate weapon" that would keep it from being as effective as they are betting it'll be. But they're big picture people - they're not worried about the nuances of reality by and large.
But it's now just like every other insane investment scheme in history listed above - just so much worse. It's "too big to fail" because so many companies have invested into it. But it's also not too big to fail, because fail it will - just like every other insane investment scheme in history. The question is, what do we do when the bubble generating 1/3 of the entire US's economy turns out to be a circle jerk of failed companies?
My bet is on the bipartisan action that's happened in the past few decades...
- the dotcom burst resulted in major bailouts for corporations
- the 2008 housing bubble burst resulted in major bailouts for corporations
- the covid pandemic resulted in major bailouts for corporations
No new regulations. In fact, there were new protections for the corporations who took advantage of American citizens. And insanely large bailouts on unrealized gains.
We are paying significantly more for computer parts so that computer companies can develop software that will force us out of jobs so they can control us more while screwing with our economy so that our government decides to use our money to bail them out when it inevitably fails.
This assumes AI is as good as they claim
it is.
Not according to every person who is professionally building it. "Easy button" AI, which is crucial for downsizing beyond what they've already done, is not a thing, nor will it ever be without AGI.
The one that will be standing strongest at the end of the GenAI hype, seems to be Google, who can fund this from their ad monopoly position, and weather the upcoming shakeout. And they seem to have been more clever technologically, with their TPU hardware (and I noticed during a test they did caching as part of their RAG, also a lot cheaper). Google has been focused a lot on cost of inference.
OpenAI probably survives 2026 on investments/subsidies. The moment of truth may then come in 2027.
What to think of the following outcome: Google survives as the top player in GenAI, a market that ends up to be a lot smaller than thought with no AGI. Claude and OpenAI fail financially, and scraps of each get gobbled up by one of Microsoft, Amazon, maybe Meta, maybe Apple. Coreweave and others go belly up and the scraps get taken over by established players. Nvidia survives bit its valuation comes back to earth. Yes, there will be use for their chips, but they will be far less profitable for a long period. Oracle takes a huge hit, but the Ellisons are already fleeing into media (destroying the plurality of information) and their database business keeps making money.
When the bubble bursts, a lot of investments, a lot of it circular, will turn out to be smoke. After the smoke has cleared, some will remain and grow.
At this point, from a financial perspective, I would be surprised if OpenAI independently survives beyond 2027 and the same is probably true for Anthropic. But who knows, maybe the train will keep running a bit longer burning all that cash. And who knows, maybe lots of companies will get into coding, expecting to replace most of their IT engineers with a $200/month subscription and it will take a while before that turns out to be a mistake.
I still (since 2023) expect the bubble to burst in 2027 plus or minus a year. I wonder if OpenAI and Anthropic aren't too late with their IPO to profit from the craze. And they can't go early because their numbers aren't good.
Maybe SA is telling investors he’ll do what Anthropic won’t and that he’s got a commitment from Whiskey Pete.
Or maybe just signaling frantically in that direction. Warfighters-R-us, and such.
Funding any of these businesses doesn't make sense and neither does their over the top announcements. OpenAI has and will always have competition from Anthropic, Google, xAI, and others besides the ever improving Chinese models. Competition constrains their ability to raise prices. If their jobs armageddon is true (I have serious doubts) they'll be looking at massive deflationary pressures making things even worse. Ads will erode credibility. They're burning money like an old fashioned coal-driven engine burns coal and there is no clear path to profitability - for any of them, especially the b2c model OpenAI has focused on.
You don’t need to lie about your product if it’s good. Altman was cooked by the time he tweeted about strawberries.
The answer is yes, it makes total sense, if the company is a scam.
The end-state here is an IPO. The current "invenstors" get to cash out and your 401K ends up holding the bag. A very, very, big bag. Sam and Dario should be in jail.
I don't think we'll see an OpenAI IPO. While you're right that some fonds may have to buy into it for structural and ideological reasons, I seriously doubt that they could raise a TRILLION for... nothing.
That doesn't mean they won't try.
These people (Musk, Amodeo, Altman, Andreeson, Kurzweil and others) are basically insane religious fanatics.
Their religion is the "singularity", the idea that we will develop a sentient AI that will take over the universe. Not the world, the universe. They claim it is coming very soon and we must prepare now.
Climate change doesn't matter, rampant disease doesn't matter, inequality and social unrest don't matter. Nothing matters but making sure that the singularity happens because that will solve all other problems and make us all into demi-gods (or dirt, depending).
Now, maybe they don't really believe it. On the other hand they have spent lots of time and money on books, white papers, symposia, think-tanks, and related propaganda discussing it. Dario quit OpenAI supposedly because they weren't worried enough about "AI Alignment", which is making sure God, when we invent it, doesn't kill us all right away.
So I'm not holding out a lot of hope that they will behave rationally.
It’s almost as if the game is rigged…
Time will tell but I suspect the fantastical narrative that AGI will be realized (based on LLMs) will prove to be the most costly financial mania mass delusion in US financial history, perhaps in all financial history. When this massive credit bubble blows, as they all do, so will associated nonsense narratives.
Worth noting, even if AGI were to be realized this financial mania will collapse because it is based on decades of monetary and fiscal distortions, meaning choices that will prove unsound, which must be wrung from the system at some point. All such cycles are fundamentally the same.
"This is dumber than WeWork" -Ed Zitron https://www.youtube.com/watch?v=N_3X6qF2tT4
Ed Zitron & Gary Marcus - the real GenAI heroes.
The magnificent 7 wasted such an opportunity.
Three years ago they should have bought each other's stock.
They'd be worth several times as much today.
Thank you, Gordon. Of all my comments these past few years I find this one to be my meatiest.
I also appreciate that now I know enough to check out your substack.