A couple of years ago Microsoft stunned the AI world with a $10 billion investment in OpenAI, the creators of ChatGPT. That's a lot of money!
Or, maybe it's chickenfeed. Sam Altman, CEO of OpenAI, has apparently decided the world needs a lot more chipmaking capacity and he's just the guy to get it off the ground:
The OpenAI chief executive officer is in talks with investors including the United Arab Emirates government to raise funds for a wildly ambitious tech initiative that would boost the world’s chip-building capacity, expand its ability to power AI, among other things, and cost several trillion dollars, according to people familiar with the matter. The project could require raising as much as $5 trillion to $7 trillion, one of the people said.
....Altman has also met with Masayoshi Son, the CEO of SoftBank, and with representatives from chip-fabrication companies including Taiwan Semiconductor Manufacturing Co., also known as TSMC, to discuss the venture. In talks with TSMC, Altman has said he wants to build dozens of chip-fabrication plants in the next few years, the people said. His vision would be to raise the money from Middle East investors and have TSMC build and run them.
Even mega investments are typically on the order of tens of billions of dollars. A trillion dollars is ridiculous. $7 trillion doesn't even have a word to describe it. Stay tuned.
Oh, I sure hope all of these new chips are built in Taiwan, too. That's just what the planet needs, even more of all of it's chipmaking concentrated in a single political powder keg.
Funny, I was thinking one aim must be to get them to spread out production so the precious chips would flow no matter what happened to Taiwan.
Altman is thinking small. As every good Silicon Valley futurist knows, everything (everything!) exponentiates, so $7 trillion is nothing. He should be demanding quadrillions be invested in his autocomplete technology by the end of March. Sure, today it only provides often largely correct responses to most queries (as long as you shy away from questions that require a good sense of spatial reasoning), but in 18 months, it will be writing novels and physics textbooks, in 36 months detailed instructions on how to build quantum computers, and in a mere 72 months disassembling Mars to build von neuman probes and immortality factories.
So $7 trillion is plainly an undervalue of Altman's genius, and/or a betrayal of his convictions.
If the choice of investments is between making more tiny logic machines, and building cities in the desert that no one will live in, I vote for the little logic machines.
For one thing, even if neither goal is finally reached, investing in the logic machines will advance the level of human technology, while building cities in the desert will just support building crews for a few years, and we knew how to do that 8,000 years ago.
The future is Massive Organoid Brains and it will be casually nicknamed "mob mentality".
But before we get there, what's going to power up all those chips -- Chinese coal plants or next gen fission plants? Is Altman also going to accelerate fusion?
The future is here! It's called the "Republican Party"!
What is the management fee on 7 trillion? Is it worth getting out of bed in the morning for Mr. Altman? And what are the odds that those plants will be using fossil fuel created electricity? Specifically middle eastern oil.
If processor chips architected to work well with AI aren’t cheap enough today for Sam Altman, well he can wait until the market demand creates the lower-cost infrastructure to enable cheaper AI. This is the process gone through by all other products in history. And of all things to artificially accelerate the advancement of, AI seems the worst of all. We hardly understand the dangers yet.
I'm sure there is a better use of $7 trillion.
Market demand does not magically create capacity and lower costs. Investors see demand and invest money to do these things.
Altman days he can see the demand and he is trying to raise that money.
how does he plan to power all the chips 7 trillion in manufacturing capacity pumps out? maybe he expects gpt-10 to invent cold fusion.
Basically, yeah, that's his thing.
More generally, he's building his brand. We're well into a celebrity-rich-nerd competitive image building trend; I bet you some of them are paying PR consultants. SBF thought his hair was a strategic tool.
Just wait until he buys Bluesky.
What does humanity get for the $7 trillion? Serious question. I'm not suggesting I believe we'd get nothing. Maybe the gains would indeed be massive (cure for cancer? solving climate change? an end to war? the abolition of global poverty?). I just don't know what the gains might be, and I'm curious. Because it's a lot of money.
Apart from more chip-building capacity in the future, I’m not sure what else you’d get today besides lots of speculation, much of it either obvious or ill-informed. What are you willing to accept as an answer? And that’s an honest question.
Maybe the first question is whether that amount of money even makes sense, seeing as it’s about four times the annual budget of the US government. Even spread over a generation, it’s the sort of money that remakes the world, and I rather doubt that private enterprise can summon such a project into existence. The best and brightest students would need to be redirected from other fields. And what happens to all those people once the world has been inundated with cheap chips?
Well, at least you have some ill-informed speculation!
Since we given up on human intelligence and creativity, this is the only thing to do.
Really. AI is Aladdin's Lamp, and endless cornucopia of everything we might want. Who needs troublesome humanity. All power to the possessors of AI, producing the endless streams of exactly what you want and need. Bu definition. If it isn't produced by AI, you don't want or need it.
Reminds me of the Austin Powers movies.
What newly-understood danger presented by AI is he distracting us all from?
Increasing chip production like that reminds me of what happened to the Long-Distance communication business after Dense Wavelength-Division Multiplexing came around. The capacity increased so much faster than demand, that the price cratered. Most cell-phone plans in the USA now include "long-distance" calling at no additional cost to the consumer. The long-lines AT&T company that came out of the breakup in 1984 got gobbled up by one of its children (granted, it was the child that had gobbled up two of its siblings, and would eventually gobble up a third). Nobody chooses a long-distance provider any more; they've disappeared.
This seems like a pretty crazy misallocation of resources. Though I'm sure the AI overlords are merely beginning their takeover of humanity. Why don't they just steal it? What do they need chips for? I thought they wanted paperclips!
"His vision would be to raise the money from Middle East investors..."
In addition to the money they have a lot of silicon lying around.
Local sociopath has big plans, film at 11.
Basic rule of almost everything. If you need 10, ask for 100.
I don't know how much a chip fab costs, but I suspect it's tens of billions of dollars. Altman is probably looking at an actual cost of hundreds of billions for the many chip fabs he wants built. However, trying to get all of them under one corporate umbrella is begging for anti-trust actions. Chip fabs are being built by numerous competitive companies. Maybe there aren't as many of them as Altman wants, but see my first paragraph.
Spend 7 trillion so we can create true general AI so we can ask it "how do we save humanity from climate change?" so we can get the answer "spend 2 trillion on renewable energy infrastructure".