It's in the title. I've been waiting and hoping for some counterexamples, but I've concluded, and I think people need to be ready for the eventuality that the AI bubble is NOT going to burst.
That's a bold claim, but I'm prepared to back it up.
- Not adopting the AI paradigm is going to become increasingly costly
First of all, it's not going to not burst because AI is good or useful. In fact, right now, we see lots of companies pushing away from AI because it is unreliable and problematic.
But the alternative costs from the other side are creeping up. If you are a company looking to hire developers to write software, you need to provide development machines to those developers. A development machine that might have cost $2000 a couple years ago is well on the way to $6000-$7000 in the near future.
And that's small potatoes. Even if you buy those for your developers, who are still highly paid, mind you, the software they're going to be writing will NOT be for PCs - regular people are NOT going to spend that for a PC. Which means most of the software that's going to be developed will target iOS and Android... and that's it. Which will continue to put 30% of all development profits RIGHT back into the locked down AI nightmare ecosystem. It's a disgusting, negative feedback loop that's about to accelerate in ways that would've made the most nightmarish predictions 5 years ago seem conservative. You might not want to swim in the ocean of crap, but they are actively eating the pier under your feet, and using the bodies that hit the water to take it apart even faster. They are going to change the world in ways that will FORCE you into their system - you don't get a choice.
- Bubbles don't HAVE to burst anymore
Tesla hasn't put out a successful new product in 20 years, and it continues to barrel right along, with its useless hack CEO hanging on as the richest person in the world. The old rules do not apply, and the sooner we acknowledge it, the sooner we can prepare for the new normal.
- NOBODY who is responsible for enforcing anything like responsible economic activity will EVER allow the bubble to burst
WHO is going to allow the bubble to burst? The investors who would lose everything? The SEC that would collapse the entire economy by not turning a blind eye? The captured politicians that are being paid billions to be complicit?
Let's be clear... the invisible hand of the market, to the extent it ever existed, certainly does not now. The idea the market is fair and responds properly to economic activity requires all actors to act, if not in good faith, at least SELF-INTERESTEDLY against each other as checks and balances on each other to verify the veracity of real claims about the economy. Is ANYBODY deluded enough to think that's happening? Everyone who could potentially blow the whistle or pop the bubble knows each other and they all have guns pointed at each other's heads knowing they all go down together. It will NOT be allowed to pop, and if that means making up numbers out of whole cloth, it's going to happen. If people won't rise up when their pedophile president is murdering people in the streets, they're certainly not going to when OpenAI claims 1 trillion in profit out of nowhere. Add to the fact that small investors who MIGHT get skittish are SUCH a small and irrelevant piece of the pie in these economic transactions that even if they all pulled out, the machine could not be stopped, and you realize that there is no stopping this nightmare.
It's not going to pop. Barring a revolution, we are on an inexorable course to the most awful tech dystopia imaginable. And even revolution is unlikely to be enough. Most people are so addicted to corporate tech that they're more likely to link arms and defend the headquarters of their favorite social media than take up arms against it. Make no mistake... the end of consumer facing open hardware is not a temporary redirection of resources to a failing bubble - it is a complete shift in the entire paradigm of how people use technology, bringing it under the complete control of a very few. This is not the latest salvo in an ongoing battle - it is the final bomb that has ended the war, and there's nothing left but slow attrition until personal computing and the very concept of devices you own and control sit in the dustbin of history with cars you could work on yourself.
I want nothing more than to be wrong. I am not happy to doomsay here. But to pretend this is some kind of blip in a machine that's going to stabilize someday is to ignore every single bit of functioning pattern recognition I have.
I don't have a good conclusion. I guess - hug your families, hoard what tech you can, and maybe make offgrid plans now. Good luck to all of us.
I want the AI bubble to pop, but as time goes on, I increasingly believe it will not. At least not without being tied to another major collapse.
I work for a large software development company that initially was very hesitant about AI anything, but has since gone whole hog because everyone else is doing it. To run our stack locally and debug interactions between services requires a machine that costs about 50% more than it did a year or two ago and they are not in stock at the moment. Storage and memory costs are insane. We are partially offloading the development stack to the cloud now, so expect to pay more for the server you rent as you compete with us for resources.
In the hands of a competent developer, anthropic's models for instance, are pretty good in my experience. In the hands of an incompetent developer, they are a nightmare. Tuning base prompts has helped us stem the flow of garbage from weaker colleagues to a manageable level, and continuing that work seems to be improving things incrementally. I can foresee having a library of prompts for the whole organisation for each common thing we want to do. I expect that model owners will inevitably steal those prompts since they are in the position to observe them, and they'll use them to improve models to where they better intuit what you are asking for and use better patterns by default. I also expect tooling to change to assist models, probably to the point that they will be AI first.
But it's clear to me that everything is changing for better or worse, regardless of what anyone thinks of it. As for what corporate thinks, it remains to be seen if competence is valued, or if "close enough is good enough", will suffice.
I don't expect anthropic or anyone else in the AI space to stop iterating on models of any kind. I do expect anthropic's (and others in the space) goal to be to get the models to a point where not using them will be a net disadvantage when trying to compete in the software development space. Then they will up the price, and use fiddling to keep the price just below where it makes sense to hire another developer, in order to suppress those wages.
Moving on from there, I expect them and others in the AI space to be training models for all kinds of things, which will end up suppressing wages in other industries. Whether that is even possible remains to be seen, obviously, but it won't stop them from trying. Success (if any), or even the perception of it, in the software development space will spurn others to try the same thing but somewhere else.
They are right, bubbles do not have to burst anymore. The housing bubble is one example. The stock market basically only going up is subject to the same forces. Economist Gary Stevenson explains. TL;DW, once rich people reach a certain wealth threshold, they have nowhere to put profits except to reinvest, which begets more profits, and so on, until they outcompete you for those assets. Once they buy up as much as is available for one asset class, they move to another. This is why housing and the stock market keep going up and up. They will even invent new asset markets to pull more money from people: see what's happening to veterinarians for instance.
Who are shareholders? I've heard that 80% of the stock market is owned by the 1%. And 90% of the stock market is owned by the top 10%. Even if none of that is true, what is close to the truth? 50-60%? That's still too much, and how long would it take them to reach those higher numbers anyway? Besides, if the majority shareholders already have more money than they could ever spend, why do they even need to change their sentiment?
To preserve the status quo, preventing anything that stops his gravy train. The same train that is causing all of this.
Absolutely. Everyone's pension has been swept along into this system, and held for ransom and to beat us with at the same time.
The bubble is being maintained by not taxing those who have more money they can ever spend. They reinvest it. The profits they end up with don't come from nowhere, that money has to come from somewhere, and there's only so much money in the system. If you follow that path, you come to know how the monetary system works, and why we inevitably have to print money to keep the system going (it's a function of the system). Printing money causes inflation. The money funding this whole situation is essentially coming from all of us, stolen fractionally by inflation. The bubble won't stop until people have nothing left and/or are no longer able to function (societal collapse), or we tax billionaires and begin to fix things. We have a hard road ahead either way. Good thing it's so necessary to create Bigfoot blog videos, we couldn't live without them.