The problem with AI as purely Large Language Models

[Ace-Garageguy]

 

[stavanzer]

"Are Hallucinations Popping the A.I. Bubble?"

Not on your life. It will never happen.

Too Many Young Techies have too much of their Ego invested in A.I. Uber Alles.

Too Many Investors (Corporate & Vulture) have too much Money, Ego, & Reputation invested in A.I.

Too Many People who are convinced that they are the Arbiters of the Future have too much of their Ego & Reputation invested in A.I.

This Bubble will have to Catastrophically Destroy Billions of Borrowed Dollars, and Several Reputations and Huge Numbers of Fragile Egos, before it goes away.

And when It finally does, the Backlash against A.I. and Tech in general will be epic. Enough to wipe away the bad, but much of the good, as well.

See: "Baby/Bathwater"

Silicon Valley, and It's many Disciples aren't nearly as smart as they think. They do not have nearly as much control over events as they think. And they have little knowledge of the world outside Universities, Cities, and Think Tanks. The last 10 years has proven that handing the keys to the world to these folks, has provided some great video games, some horrible side effects, and lots of Social Dysfunction.

Sometimes the old ways are wrong, or backwards. But, Changing them just for the sake of Change, is not working the way it was promised to us.

Remember "Chesterton's Gate"

[Ace-Garageguy]

Up until a few years ago, I read everything Ray Kurzweil wrote, and I believed the future of AI was in good hands if he was one of the people at the helm.

https://www.theguardian.com/technology/article/2024/jun/29/ray-kurzweil-google-ai-the-singularity-is-nearer

His mathematical modeling of neural networks based on mimicking the cascade of decision events that was at the time the explanation for human consciousness made a lot of sense...and something similar is alluded to in the video.

IBM's Watson and a chatbot app that's apparently fallen by the wayside also gave me a lot of hope.

Kurtzweil now works for Google, and I'm beginning to think he may have become lazy (or simply doesn't have the authority his title would imply), and the majority of the current generation of AI researchers have elected to take what looks to be an easier path, only getting halfway home with the hardware/firmware, and hoping "training" will let the machine get the rest of the way on its own...apparently an approach embraced by a "consensus" of people in the field, rather than opting to do the hard work.

Possibly much to their peril.

The appalling stupidity of some of the garbage Google's "AI Overview" vomits up is evidence aplenty that they've missed the mark by a mile.

But a massive problem is that the majority of people who accept Google as the font of all knowledge can't tell the difference between idiotic drivel and correct interpretation of all available information...which Watson excelled at.

Edited August 16, 2024 by Ace-Garageguy