The answer is probably somewhere in between. It'll be somewhat useful. And somewhat overhyped.
I do know that most individuals who choose to use AI are in the end, going to choose the cheaper models coming out of China, etc. So I'd be amazed if there isn't a crash in The US.
I'm in disbelief that someone wrote an article about Wankel without mentioning Mazda.
Mazda is not a US company, but they did "bet the company" on Wankel. By 1973, they were selling almost 2 million Wankel-powered cars per year. After the '73 fuel crisis, Mazda mostly reverted to piston cars, but still sold 600,000 Wankel cars per year until the mid-'80s.
I don't think that Mazda's successful use of Wankel implies anything about the future of AI. But calling Wankel a failure without even mentioning Mazda is ... incomplete.
You make a good point--I worded my post carefully, but I could (and should) have added more nuance to my description of the Wankel as a failure. It was not, as you said, a *total* failure. My point was that it certainly *was* a failure in the sense that it never came anywhere close to meeting the expectations that the hypesters set for it, and that the (thankfully not-too-many) investors who bet on a Wankel revolution got hosed.
As I briefly alluded to, it did get commercialized in Europe, which was what led to the excitement and bubble in the US. I was trying to get to the discussion of AI more quickly, but I probably should have included a paragraph mentioning Mazda's intermittently successful commercialization of the tech and, more generally, acknowledged that it had (and still has) niche uses.
Honestly, that actually might have made for an even better metaphor: the Wankel was tech was promoted by self-interested corporations as transformational, but its limitations meant that it only came to have commercial value on a very limited scale (arguably in absolute terms, and certainly in comparison to the expectations set for it).
Anyway, thanks for reading, and for commenting with good points.
It’s worth mentioning our largest investment is in compute resources which can be repurposed if LLMs don’t pan out. Imagine if we take all this compute and dump it into alphafold, genome decoding, or 1,000 other machine learning tasks.
Are investors hoping to get rich replacing white collar workers going to be sad? Sure, but science would get a shit load of very cheap compute into the foreseeable future.
And let’s be honest, most of the reasons there is so much rewrite required for AI code is because people who don’t understand good software architecture are the ones writing it.
Good fuck billionaires and trillionaires the “economy “ should be the inter -relationships we have with eachother not some fiat currency and how much humanity has to suffer for the made-up paradigm of stocks, credit, money and share holders that have nothing tangible to back up what they clam as assets.
There's two intuitions about AI that lead to optimism or pessimism.
Pessimism: AI is just another technology.
Optimism: AI actually is Intelligence, just like the acronym says.
For the optimist, no matter where AI is today, it will learn more and adapt, because that's what intelligence is. The optimist sees that Moore's law of hardware compute has been measurably prportional to AI ability. Moore's law is not slowing down, therefore AI will be getting smarter. Further, intelligence on silicon means AI doesn't sleep, doesn't grow old and forgetful, never dies, and can be replicated effortlessly. Where does that all logically lead? Intelligence with no clear upper-bound. Which is also why many optimists are worried about AI risks.
The pessimist somehow ignores all that and sees Wankel rotary engines.
At my work they have already resigned themselves that AI can be used for research or as part of a vendor wrapped package, but they made the governance for home grown AI agents so burdensome that people got the message that “you aren’t going to get your custom AI agent approved.”
I think you’ll find that half a dozen narcissistic sociopaths with no greater ambition than to have the largest number on the Forbes billionaires list are betting the economy on a doomed technology while the rest of us stand by and watch and/or cry.
The answer is probably somewhere in between. It'll be somewhat useful. And somewhat overhyped.
I do know that most individuals who choose to use AI are in the end, going to choose the cheaper models coming out of China, etc. So I'd be amazed if there isn't a crash in The US.
I'm in disbelief that someone wrote an article about Wankel without mentioning Mazda.
Mazda is not a US company, but they did "bet the company" on Wankel. By 1973, they were selling almost 2 million Wankel-powered cars per year. After the '73 fuel crisis, Mazda mostly reverted to piston cars, but still sold 600,000 Wankel cars per year until the mid-'80s.
I don't think that Mazda's successful use of Wankel implies anything about the future of AI. But calling Wankel a failure without even mentioning Mazda is ... incomplete.
Hi Matt! Love your name.
You make a good point--I worded my post carefully, but I could (and should) have added more nuance to my description of the Wankel as a failure. It was not, as you said, a *total* failure. My point was that it certainly *was* a failure in the sense that it never came anywhere close to meeting the expectations that the hypesters set for it, and that the (thankfully not-too-many) investors who bet on a Wankel revolution got hosed.
As I briefly alluded to, it did get commercialized in Europe, which was what led to the excitement and bubble in the US. I was trying to get to the discussion of AI more quickly, but I probably should have included a paragraph mentioning Mazda's intermittently successful commercialization of the tech and, more generally, acknowledged that it had (and still has) niche uses.
Honestly, that actually might have made for an even better metaphor: the Wankel was tech was promoted by self-interested corporations as transformational, but its limitations meant that it only came to have commercial value on a very limited scale (arguably in absolute terms, and certainly in comparison to the expectations set for it).
Anyway, thanks for reading, and for commenting with good points.
It’s worth mentioning our largest investment is in compute resources which can be repurposed if LLMs don’t pan out. Imagine if we take all this compute and dump it into alphafold, genome decoding, or 1,000 other machine learning tasks.
Are investors hoping to get rich replacing white collar workers going to be sad? Sure, but science would get a shit load of very cheap compute into the foreseeable future.
And let’s be honest, most of the reasons there is so much rewrite required for AI code is because people who don’t understand good software architecture are the ones writing it.
Good fuck billionaires and trillionaires the “economy “ should be the inter -relationships we have with eachother not some fiat currency and how much humanity has to suffer for the made-up paradigm of stocks, credit, money and share holders that have nothing tangible to back up what they clam as assets.
There's two intuitions about AI that lead to optimism or pessimism.
Pessimism: AI is just another technology.
Optimism: AI actually is Intelligence, just like the acronym says.
For the optimist, no matter where AI is today, it will learn more and adapt, because that's what intelligence is. The optimist sees that Moore's law of hardware compute has been measurably prportional to AI ability. Moore's law is not slowing down, therefore AI will be getting smarter. Further, intelligence on silicon means AI doesn't sleep, doesn't grow old and forgetful, never dies, and can be replicated effortlessly. Where does that all logically lead? Intelligence with no clear upper-bound. Which is also why many optimists are worried about AI risks.
The pessimist somehow ignores all that and sees Wankel rotary engines.
Y
https://procrastinationforthegreatergood.substack.com/p/comment-ipo-ed-genai-might-go-the
What do you mean turns out to be a dud? This is akin to wondering in 1996 if the www will turn out to be a dud.
You say the Internet 1996, I say Wankel engine 1971 :)
How so?
Except they will never admit it.
At my work they have already resigned themselves that AI can be used for research or as part of a vendor wrapped package, but they made the governance for home grown AI agents so burdensome that people got the message that “you aren’t going to get your custom AI agent approved.”
“We”?
I think you’ll find that half a dozen narcissistic sociopaths with no greater ambition than to have the largest number on the Forbes billionaires list are betting the economy on a doomed technology while the rest of us stand by and watch and/or cry.
I hope both go down in flames.
Are we betting the economy on AI? Biotech is out there. And green energy.
And large companies are already upping process automation using AI, so it has value.
Green energy isn’t a thing.
Well, that is what it is called. Mini-nukes, cold fusion.