On the go? Check out Big Technology Podcast on your app of choice:
Spotify: https://spoti.fi/32aZGZx
Apple: https://apple.co/3AebxCK
Etc. https://pod.link/1522960417/

"And for that reason, I'm out."

The bubble will burst,  AI will not die but a lot of companies will die and thats a very good thing

OpenAI is becoming the next Metaverse. I've had my popcorn ready for the collapse.

"You just spent a trillion dollars, to be an app" Yep that sums it up pretty good.

If Cuban is voicing realistic opinions among the quagmire or bs, then the bubble has reached the top.

this video and these comments are gonna be so fun to look back on in the future

AI is not profitable, the industry is in complete denial to delay a crash

If it turns out "there can only be one" AI company, then the winner will still buy off the losers' data centers. However, what would be more destructive is if a revolutionary breakthrough yielded the same throughput as data centers but with models ran locally on laptops or PCs. That would render the infrastructure obsolete, resulting in a money pit.

Mark would know, he sold a worthless business for billions

The people in the comments so pressed about what he’s saying are the kind of people that have all money invested in crypto and get upset when you explain to them that crypto is not replacing the banks, it’s just an unregulated market propped up by speculation and illegal transactions.

If you think a company spending 1.4 trillion and generating 13 billion dollars in revenue isn’t going to crash, you don’t understand how this works

Apple was smart not to be too quick to jump on the AI bandwagon.

They're shitting away that money AND DO NOT BAIL BUSINESSES OUT WHEN THEY FAIL!

0:43 "They're [OpenAI] just shitting away that money" - Mark Cuban

⚠️AI companies are failing businesses ! And now they are making this scenario to cover the short comings! 😶‍🌫️

LLMs also ain't really the technology to do it. Inference might get "better" (each iteration is marginally improved at best, genuinely less effective at worst), but the reality is it's like trying to run a car with square wheels. You can throw horsepower at it and it'll go faster sure, you can make it more efficient, you could make the wheels slightly less square or use smaller squares, but it's not the right tool for the job. LLMs are very good at sifting through a lot of data, but people are trying to use them for everything without any limitations. We don't use RAM as a CPU because it's not designed for it. You could probably force it to work with horrible results, which is what LLMs are doing.

One of the first lines was important. The equipment in data centers will rapidly depreciate in value on one of 2 ways: Resale value due cheaper alternatives later or, upgradable costs.

So when a new center is built with newer hardware, its not simple to resell the old hardware unless they eventually make enough money through services or data selling

Sometimes it’s better to be an iterator than it is to be an inventor

The bathroom is dirty - make AI clean it.

A correctly timed short will be MARVELOUS