Even if ML models hit an unexpected scaling problem, they're still very very useful. Extremely disruptive, too. Then again, it could be ages before we figure out how to put them to the best use.
Even if ML models hit an unexpected scaling problem, they're still very very useful. Extremely disruptive, too. Then again, it could be ages before we figure out how to put them to the best use.
But, if you want my take on it, there's a lot of hype and not much substance there.
There's a good article at https://garymarcus.substack.com/p/did-googleai-just-snooker-...
I'm on team "What a time to be alive".
Stunning trees in isolation are impressive and awe-inspiring, but note that the rest of the trees in the forest are made up of just the same stuff and composed in extremely similar ways!
ML shows not that computers can be smart, but that smarts is just a matter of statistics + abstraction. In this way no "intelligent" being is any more impressive, just more complex and fine-tuned.