Hallucinations My Ass!

In my initial LLM "AI" rant "Bullshit All The Way Down", I stated:

I am offended by the current use of the term "AI". I understand, this is The Thing now, so, no one cares.

Given that I am an old man - more particularly an old white (mostly German) boomer male 5th generation US citizen - it is my job to be offended by things. But, I am politically ultra-liberal, so the domain of things that offend me is probably not that which is standard for my generation. Where have all the old hippies gone?

When I was writing my 2nd "AI" Bullshit post "Hallucinations? No, Bullshit!", the "hallucinations" term was offending me - it was really sticking in my craw. This post is attempting to determine why.

Here is the 1st paragraph of the wikipedia article on "Hallucination (artificial intelligence)":

In the field of artificial intelligence (AI), a hallucination or artificial hallucination (also called bullshitting, confabulation or delusion) is a response generated by AI that contains false or misleading information presented as fact. This term draws a loose analogy with human psychology, where hallucination typically involves false percepts. However, there is a key difference: AI hallucination is associated with erroneous responses rather than perceptual experiences.

OK, cogent points here:

1. "Bullshitting" - yes, FTW!
2. "loose analogy" - quite the understatement.
3. "AI hallucination is associated with erroneous responses rather than perceptual experiences." Bingo!

Let's repeat that: AI hallucinations are "erroneous responses" aka wrong answers aka incorrect answers aka false answers aka a chatbot giving detailed suicide instructions to a troubled child.

So I think what was sticking in my craw was the fact that this term, "AI hallucination", was probably invented by some marketing person. The term conflates an LLM model with a brain, presumably a human brain - though I'm guessing all mammals hallucinate, maybe all vertebrates, maybe anything with a nervous system. An LLM model is none of these things. Repeat, an LLM model is none of these things. There is no brain. There is a statistical bullshit generator.

But, hey, techbros need another unicorn, let's pretend that LLMs are just like the human brain! And these MFs are just like Leon, who's been claiming "full self-driving mode in the next year" for the last 12 years, claiming "AGI (Artificial General Intelligence) is coming any day now" aka RSN - Real Soon Now (Buckaroo Bonzai) - aka marketing hype to keep the stock price up.

No f#cking way. I think AGI is doable and is a worthy goal, but, as I said in my 1st Bullshit post, LLM is just 1 tool among many it will take to achieve AGI.

On their own, I think it is very possible that LLMs are a complete dead-end on the road to AGI. Chatbots have been around since ELIZA in 1966. Almost 60 years later, LLMs are certainly far more capable than ELIZA, but they still can be absolutely relied on to spew complete and utter crap some percentage of the time.

But, so what! Techbros are raking in the $$$! So, per capitalism, it's all A-OK! It's great! Maybe Leon will give them some government agencies to destroy! Capitalism, FTL! :-<

More on this soon ...