It has been more than just hyperscaling. First of all, the invention of transformers would likely be significantly delayed without the hype around CNNs in the first AI wave in 2014. OpenAI wouldn‘t have been founded and their early contributions (like Soft Actor-Critic RL) could have taken longer to be explored.

While I agree that the transformer architecture itself hasn‘t advanced far since 2018 apart from scaling, its success has significantly contributed to self-learning policies.

RLHF, Direct Policy Optimization, and in particular DeepSeek‘s GRPO are huge milestones for Reinforcement Learning which arguably is the most promising trajectory for actual intelligence. Those are a direct consequence of the money pumped into AI and the appeal it has to many smart and talented people around the world

Go broke

This is not that. They're all hoping to be the next Google or FaceBook. They know damned well most are going to lose. The gamble is that they won't be the one holding the bag when the bubble pops.

This is as high stakes as tech gets today.

This is no revelation. THEY KNOW. The play is obvious.

Not one these investors wants to risk missing out on being the next Google or FaceBook or Twitter or Amazon. They know damned well the vast majority will fail. They're gambling on not being the one left holding the bag.

AI is here to stay, will continue to improve, and there will be a killer app, probably a dozen. My money is on life sciences, particularly medicine.

Unimaginable amounts of money spent just to provide a free service to help improve the human race by sharing knowledge. Such marvellous gentlemen.

Spend it all and fail

Well, in the Soviet example everything was government.

And governments seem to be so excited by the prospects of this "AI" so it's pretty clear that it's still their desire most of all.

EDIT: On telegraph and Panama you are right (btw, it's bloody weird that where it sounds like canal in my language it's usually channel in English, but in the particular case of Panama it's not), but they might perceive this as a similarly important direction. Remember how in 20s and 30s "colonization of space" was dreamed about with new settlements supporting new power bases, mining for resources and growing on Mars and Venus, FTL travel to Sirius, all that. There are some very cool things in Soviet stagnation - those pictures of the future lived longer than in the West against scientific knowledge. So, back to the subject, - "AI" they want to reach is the thing that will allow to generate knowledge and designs like a production line makes chocolate bars. If that is made, the value of intelligent individuals will be tremendously reduced, or so they think. At least of the individuals on the "autistic" side, but not on the "psychopathic" side, because the latter will run things. It's literally a "quantity vs quality" evolutionary battle inside human kinds of diversity, all the distractions around us and the legal mechanisms being fuzzied and undone also fit here. So - for the record, I think quality is on our side even if I'm distracted right now, and sheer quantity thrown at the task doesn't solve complexity of such magnitude, it's a fundamental problem.

I'm more about separation of addressing data and data model from addressing services and service model for storing and processing it, to make those uniform, because in uniformity lies efficiency and redundancy and ability to switch service models, and uniformity inside proprietary services is already achieved, so in this case uniformity works for the people.

I mean, that's probably what you meant, I'm being this specific to fight my own distractions and fuzziness of thought.

All for it to fail and implode on its own weight.

Imagine what we could have achieved globally if we had spent all that money on a different cause.

We could have managed to establish a colony on Mars, or perhaps we could have even finished developing Star Citizen.

I think Meta’s AI initiative doesn’t run on investor money since they do share buybacks instead of selling more shares to keep afloat. Meta makes more than a hundred billion of revenue from selling ads on Facebook and Instagram. So Meta’s AI program runs on boomers clicking on ads that have been generated with AI.

It's porn.

It's always porn that decides if a technology lives or dies.

This research is cooking us alive right now and for what? So machines can do all the creative things while we fight for scraps? I‘d rather the overly rich spend it on something harmless but silly. At least the average joe can make a living producing luxury items. As grim as it sounds but that‘s preferable to what‘s coming.

They're different, and I think this one has the capability of being more devastating.

The dot-com bubble was really broad. Hundreds or thousands of companies, all without vowels in their names trying to break new ground. A wild west style gold rush. When it popped a lot of small companies went bankrupt.

This is a handful of companies with billions of capital buying GPUs from NVidia to be make the largest hungriest machine they can. All in the pursuit of being first to create "AGI". If one of them succeeds, the others are toast and multiple 500+B dollar companies will collapse in on themselves. If none of it works, the same thing happens and it takes a large chunk out of $4T Nvidia too.

It's spent on NVidia GPUs. Jensen Huang just buys leather jackets from what I can tell.

Love that the pic associated with that link is Mark "Metaverse" Zuckerberg. A hallmark of successful dubious ventures, if any.

Let's be honest here, in reality, it would just made 5 people turbo rich while the rest stayed the same. Maybe make 4 more ships in star citizen, but that's it.

Bill Gates is only leaving $10M to each of his kids (I say "only"...). The rest he is bequeathing to charity. presumably his charity, but i don't know.

I believe that we are not yet in the end stage of AI. LLMs are certainly useful, but they cannot solve the most important problems of mankind.

More research is required to solve e.g.
a) Sustainable Energy Supply
b) Imbalanced demographies of industrialized countries
c) Treatment of several diseases

Like it or not, AI that can do research for us, or even increase efficiency of human researchers, is the most promising trajectory for accelerating progress on these important problems.

Right now, AI has not exceeded this scope. Yeah, AI can generate quite realistic fake videos. But propaganda has been possible before (look at China, Russia or Nazi Germany - even TikTok without any AI is dangerous enough to severely threaten democracies).

As a researcher in the domain, let me tell you that no one who seriously knows about video generation etc. is afraid of the current state of AI

I'm sure that silicon valley executives visualise a future where they own the machines that produce all intellectual property and do most jobs. They see a return to feudalism where they are the lords.

I think this greater vision is about as likely to be realised as it is that Elon Musk will invent full self driving, or robots that aren't obviously remotely operated, or a tesla roadster, or a battery powered articulated lorry with thermo nuclear explosion proof glass, or building a rocket to get the US back to the moon before the Chinese in what is clearly a new space race/pissing match. Or a hyperloop, or ever getting anywhere near to building a colony on Mars, or, or, or.

But I don't think it is just a case of AGI or bust. LLM's augmented with ai agents have a very real potential to replace a capitalism-destabilising percentage of white collar jobs without AGI.

Just like the dot com bubble popping didn't kill the web, I do think it is unlikely that any possible current AI bubble popping will kill capital's push to automate jobs away.

(And as far as I can see the AI bubble is the result of massive capital expenditure rather than rampant speculation, so because I am pretty confident in the 'value' to capital of LLM's + AI Agent's, I don't really see it as the same kind of a bubble as the dotcom bubble.)