We’ve all heard about the AI race but I don’t understand what it is that we’re (they’re) racing towards. NVIDIA has said that China will win. What will they get by winning and how will we know when the race is over and has been won?
Feel free to move the thread as necessary.
Moderating
While this is basically factual, it is going to very quickly lead into opinion and speculation which will take the thread well outside of FQ bounds.
Moving to IMHO (from FQ).
Factual information is of course still welcome, and while this forum change allows for speculation, let’s try to keep this thread as fact-based as possible. In other words, keep the speculation and opinion based on factual information and not just wild guesses.
At least early on, the pioneers of AI took pride in not having a goal:
“You’re supposed to have a problem to solve, not a technology in search of the solution,” CTO Greg Brockman told the Possible podcast in May. In OpenAI’s early days, he said, the leaders spent months “just writing down all the different ideas that we could work on for both GPT-3 and for GPT-4 … Maybe we could do a medical thing or a legal thing.”
Instead they decided to ignore the rule altogether—to great success. As it turns out, “every company, every individual, every business is a language business,” Brockman explained. “So if you can add a little bit of value in existing language workflows, then it will just be able to be adopted so broadly.”
This week, renowned venture capitalist Paul Graham offered up his take on artificial intelligence—while invoking some of the same language Brockman used.
“AI is the exact opposite of a solution in search of a problem,” he wrote on X. “It’s the solution to far more problems than its developers even knew existed.”
The AI “race” is a race for market dominance, which at this point is mostly a race between U.S. companies (with government funding and backing) and Chinese companies (again, also with funding and the backing of their government).
What they get for winning is market dominance, and the lion’s shares of the profits from such dominance, which at this point are expected to be huge.
Imagine that you are some company like McDonald’s, for example. What if you could replace all of your workers with cheap robots? Your operating costs would decrease dramatically, so there is a HUGE financial win here. What if you are the company that makes those robots? You’re going to get rich selling them to all of the companies that want cheaper labor.
What does this mean for the economy? Where will the people who are currently working those jobs find employment? That gets into a difficult area of discussion.
Another “win” is military dominance. Imagine having the best fleet of AI-driven fighter jets in the world, that you can send into battle with no risk to pilots since they are flown entirely by AI. Currently the military is working on the “dedicated wingman” concept, where drones follow a human leader, so the human element might not be completely out of the picture, at least not in the near future. But the war in the Ukraine and battles elsewhere have proven that drone technology has been a game changer in recent years, so whoever makes the more intelligent and capable drones will definitely have a huge advantage on the battlefield.
Medical AIs could replace doctors, drastically reducing health care costs while at the same time improving health care in areas where skilled medical professionals are scarce and currently overloaded.
Lots of possibilities, and lots of possible repercussions and potential down sides as well.
The theory is that AI will be as revolutionarily powerful as the steam engine in its time and will rapidly obsolete any company or country that doesn’t have the latest and best AIs driving everything.
A 2025-era business that did everything on paper with quill pens and used teams of horses to power their factories could not successfully compete here in 2025.
The belief is that the way current 2025 leading edge business and government and scientific research is run will seem equally quaint as quill pens and horses by 2050 when the AI-powered future has arrived. 2050 at the latest.
You’ll know China has won when the US economy abruptly collapses or our entire military is destroyed in a blink with no Chinese casualties.
Or at least that’s the hype. The truth may be a wee bit less decisive.
I guess that I understand what some of the potential use cases are for AI and I understand that at this point, nobody really knows what we’ll be able to do with it.
By way of analogy, did the US, via Google, win the search engine race? Is the AI race similar to that? Given that Baidu is the most popular search engine in China, if the situation is analogous, could China have their own AI and the US and the rest of the world have their own?
The whole thing reminds me of Trump saying that (paraphrasing) “The US needs to be a crypto superpower”, which I also don’t understand.
Beyond the immediate short term goal of market dominance, I think the longer term goal of AI research is the hope (and fear) that one day we may develop an “artificial general intelligence” that can exhibit human or superhuman ability across many disciplines. It usually also implies an ability of the AI to learn and master novel domains of knowledge on its own, without manual human iteration. Eventually this would lead to an AI singularity, a point in time at which AI can so effectively self-teach that it spirals into a positive feedback loop and soon becomes Arnold Schwarzenegger.
But most people don’t think LLMs (the current generation of statistically modeled chatbots) will lead us directly there. There’s disagreement about how far they CAN get us and how profitable and sustainable they’ll be after the current bubble. So for now the shorter term AI race is to just remain one of the surviving companies after the bubble collapses and the industry consolidates and such.
For now, we’re still in the “let’s throw as much compute and electricity at it as we can and see what comes out the other end” phase, though that’s already showing signs of slowing. China doesn’t have the compute power we do right now because Nvidia is embargoed there, but they have more renewable energy and a government that can direct resources anywhere it wants. Interesting times, I suppose, as long as you don’t mind an apocalypse or three.
I agree with @Reply that the current trend in LLMs won’t lead us to a true AI, and also agree that the current market is nearly 100% corporate hype and hysteria that is a bubble that’s going to burst causing major damage.
I mean, by most of my reading, the current crop of LLMs are losing money at a tremendous rate - just in operating costs much less the R&D, start up, and improvement costs. An example:
Estimating ChatGPT costs is a tricky proposition due to several unknown variables. We built a cost model indicating that ChatGPT costs $694,444 per day to operate in compute hardware costs. OpenAI requires ~3,617 HGX A100 servers (28,936 GPUs) to serve Chat GPT. We estimate the cost per query to be 0.36 cents.
Our model is built from the ground up on a per-inference basis, but it lines up with Sam Altman’s tweet and an interview he did recently. We assume that OpenAI used a GPT-3 dense model architecture with a size of 175 billion parameters, hidden dimension of 16k, sequence length of 4k, average tokens per response of 2k, 15 responses per user, 13 million daily active users, FLOPS utilization rates 2x higher than FasterTransformer at <2000ms latency, int8 quantization, 50% hardware utilization rates due to purely idle time, and $1 cost per GPU hour.
Now, optimists feel that efficiencies will reduce those costs over time. But IMHO the sheer cost of the electricity (absent a major change in our power generation and storage infrastructure) will kill off all but a few, and that’s a best case scenario.
Now, I could absolutely see China coming out ahead in that case - generally they’ve been able to build such infrastructure much more quickly (for better and MUCH worse) and have few protests when the government funded business ruins the local environment or causes brownouts.
And I won’t speak for where the USA will be in 4-10 years, currently we’re transitioning to just as callous and unjust society that would just love to use AI for domestic spying on the “wrong” sorts, to replace a workforce (even if it costs more!), and for weapons because said government doesn’t care about losses due to avoidable conflicts.
But I think the hopes of finding the next major profit/cost-saving from the current crop of AI research is more about greed and unfounded optimism in the next big thing.
With trump’s fear of electricity and magnets, I suspect we will soon see an executive order requiring all new AI installations to be steam powered.
All good answers. I’ll add another consideration:
AI is the combination of talent, compute, and data.
The realists are out to make as much money as possible out of the bubble before it bursts.
The techbros are out to create a world where everything can be done by AI, so they don’t have to have employees anymore, and the bulk of the population can be discarded.
The authoritarians want a society where AI can watch everyone every hour of the day and create a totally pervasive authoritarian state that will be impossible to overthrow.
I’ve always assumed that “winning” would be replacing most or all of your human employees (greedy bastards who want to eat and whatever) with much cheaper AI. Or else be the major supplier of AI in that circumstance.
Tangent here; the more serous game changer will be quantum computing which, if it ever becomes viable, will be of apocalyptic consequences, what with predictions such as the instantaneous breaking of every password in the world propounded.
Maybe AI will allow us to contact the entity which runs the program of which we are a part.
Of possible significance was the following statement from OpenAI followed by a swift retraction:
https://www.cnn.com/2025/11/06/tech/openai-backtracks-government-support-chip-investments
OpenAI, the world’s leader in AI and ChatGPT parent company, went into panic mode on Thursday over what it said was a very public misstatement.
Two top executives furiously backtracked from an earlier comment that suggested OpenAI might need government support to cover the $1.4 trillion in chips and data center infrastructure it’s committed to buying.
So an industry leader isn’t confident that they can support the expenditures to develop the product. Though Sam Altman expressed much faith in Capitalism, where if they failed, they’d be replaced by someone better, and then hinted that the US needs to support AI investment to maintain dominance.
Make of it what you will.
“The US cannot afford a Scam Gap!” basically.
We, the Oligarchs, of the United Grifts of America, in order to form a more perfect 0.1% hermetic society, do hereby …
Advanced AI will make advances in economics, military, medicine, science, technology, etc occur far faster than we are used to.
There was more innovation in medicine and STEM in the 20th century than in the 1900 years before the 20th century. People may dispute that, but its accurate. The world of 2000 was far different from the world of 1900 when it came to science, technology and medicine.
In theory, radical advances in AI will do the same. We may see radical advances in much shorter time frames than we are used to. The time frame to go from 6th gen fighter jets to 7th gen fighter jets may take 30 years w/o AI. With advanced AI, it may only take 5-10 years (possibly less). So a nation with an inside track on AI will have a massive military and economic advantage over nations with less advanced AI. Hence why NATO nations are limiting computer chip shipments to China. A modern NATO nation would decimate the Nazis in WW2 because modern military technology is so much more advanced. Even modern Russia, which is struggling to defeat Ukraine, would decimate the Nazis because modern Russia has such more advanced technology than we had in the 1940s. Nazi planes would be shot out of the sky before they were anywhere near in range of enemy planes, and Nazi tanks would be destroyed easily. It would be like when the US went to war with Iraq in 1991 and the US destroyed thousands of Iraqi tanks in a few weeks.
Economies used to grow by 0.1% a year before the industrial revolution. Now developing nations can grow at 10% a year and developed nations grow at 2-3% a year. Economic growth rates 20-100x faster than they occurred pre-industrial revolution are common.
A nation that has advanced AI will see its military, economic and political influence dramatically grow globally. That is why there is a race between China and the US.
Before the industrial revolution, human society was mostly limited to biological muscle. The industrial revolution gave us machine made muscle and as a result we experienced rapid advances in STEM, medicine, economics, etc. When cognition is done by machines rather than biological tissue (which it has been up until recently) we will see another dramatic increase in the quality, quantity and speed of advances in STEM, medicine, economics, etc.
Also even if we are in a stock bubble, even when the bubble pops, AI will still advance. Current AI may not, in the near term, be able to deliver the economic growth that people are expecting and AI stocks may tank. But that doesn’t change the fact that AI itself will keep advancing, the same way the internet kept advancing after the dot-com bubble burst.
Probably significantly slower for quite some time, since I suspect the term “AI” is going to be badly tarnished.
EDIT: Or perhaps I should say more badly tarnished, since I hear lots of people treat it like a swear word these days.
Once the bubble bursts, is there anything practical we’ll be able to do with all those trillions of dollars of computers in the AI data centers? A computer’s a computer, after all.