To push something to the limit requires a lot of funding if public never got overexcited about some tech many really cool things would have never being tried. Also LLM are pretty useful even as is. It sure made me more productive.
I just imagine the world in which the industry defined by deterministic nature and facts has the bravery to call spade a spade. LLM's have a function. Machine learning also. But calling LLM's Intelligence and pushing the hype to overdrive?
Re:Does current AI represent a dead end?
by godelski on Friday December 27, 2024 @12:36PM
(#42524173)
Over hype is what led to the last AI winter. Because they created a railroad and didn't diversify. For some reason we're doing it again.
Re:Does current AI represent a dead end?
by omolobo on Friday December 27, 2024 @12:22PM
(#42524021)
You are putting it too lightly. From the author:
> When all its techniques are based on testing, AI safety is an intellectually dishonest enterprise.
Or, like the old Internet adage says, "never go full retard."
Re:Does current AI represent a dead end?
by lotsofpulp on Friday December 27, 2024 @12:30PM
(#42524099)
FYI, while that quote was popular to use online, it is sourced from the 2008 movie Tropic Thunder:
Re:Does current AI represent a dead end?
by omolobo on Friday December 27, 2024 @12:46PM
(#42524258)
I am aware, good scene.
Re:Does current AI represent a dead end?
by peter_retief on Friday December 27, 2024 @12:26PM
(#42524065)
In retrospect it seems obvious now.
For some. For others with fewer stars in their eyes, it was obvious from the beginning.
Re:Does current AI represent a dead end?
by dingnuts on Friday December 27, 2024 @12:35PM
(#42524154)
the launch of ChatGPT had an amount of hype that was downright confusing for someone who had previously downloaded and fine tuned GPT2. Everyone who hadn't used a language model said it was revolutionary but it was obviously evolutionary
and I'm not sure the progress is linear, it might be logarithmic.
genAI in its current state has some uses.. but I fear that mostly ChatGPT is hallucinating false information of all kinds into the minds of uninformed people who think GPT is actually intelligence.
Everyone who actually works on this stuff, and didn't have ulterior motives in hyping it up to (over)sell it, have been identifying themselves as such and providing context for the hype since the beginning.
The furthest they got before the hype machine took over was introducing the term "stochastic parrot" to popular discourse.
Aside of the anger that this public stupidity produced in me, I always knew that this day will come.
Maybe next time someone will have the balls not to call a text-generator's with inherent hallucination Intelligence? Who knows. Miracles can happen.:)