Jonathan's Kitchen
According to MIT Institute professor Daron Acemoglu, only a quarter of tasks that could use ai are likely to become profitable in the next decade.
Talking about Goldman Sachs' bags podcastAcemoglu says that even with large advances in ai, the impact of which will not be seen for several years.
That would mean ai would impact less than 5% of all tasks and boost U.S. productivity by just 0.5% and GDP growth by 0.9% cumulatively over the next decade, according to Goldman.
“The current architecture of large language models has proven to be more impressive than many people would have predicted, but I think it still takes a huge leap of faith to say that with just this next-word prediction architecture we're going to get something as smart as, you know, Hal in 2001: A Space Odyssey,” he said.
“There could be very severe limits on where we can go with the current (LLM) architecture,” Acemoglu said.
He's also skeptical that ai can get where it needs to go faster simply by adding more GPU power.
Higher quality data, not just higher capacity, will be needed, and it is unclear where that data will come from, he added.
To get performance out of $1 trillion companies like Nvidia (NASDAQ:NVDA), Microsoft (MSFT), Alphabet (GOOG) (GOOGL), Meta (META), amazon (AMZN) and Super Microcomputer (NASDAQ:SMCI) will spend on capital expenditures on ai in the coming years, ai will have to solve complex problems, said Jim Covello, head of global equity research at Goldman Sachs, in the episode.
“We’ve been doing this for a couple of years and there’s not a single thing that it’s being used for that’s cost effective right now,” he said. “I think there’s an incredible lack of understanding of what the technology can do today. The problems it can solve are not big problems. There’s no cognitive reasoning in this.”
