Liz Neeley: The hallucinations should be disqualifying on their own. But the staggering exploitation of water, energy, and creative labor is something I am begging my friends in academia to foreground in their conversations about AI. Let’s get specific about what we are doing, why, and at what cost.

The hallucinations should be disqualifying on their own. But the staggering exploitation of water, energy, and creative labor is something I am begging my friends in academia to foreground in their conversations about AI. Let’s get specific about what we are doing, why, and at what cost.

Use generative "AI" to spitball, hypothesize, overcome the tyranny of the blank page? I mean if the environmental costs weren't astronomical & the training corpra weren't largely stolen, then yeah, sure. Use it for facts? Knowledge? To fully *Replace* thought, feeling, & creativity? Absolutely not.

We also should not call them hallucinations, but call them what they are: mistakes, falsehoods

Yes. I think this paper is useful here - link.springer.com/article/10.1...

ChatGPT is bullshit - Ethics and Information Technologylink.springer.com Recently, there has been considerable interest in large language models: machine learning systems which produce human-like text and dialogue. Applications of these systems have been plagued by persist...

This is it exactly. It is bullshit. Not even a mistake because it is not an occasional anomaly from general correctness. It is pure confabulation in psychological terms. Nonsense that tries not to sound like nonsense.

It's also a very DELIBERATE consequence of the currently fashionable definition or 'goal' of Artificial Intelligence. The original Turing Test was a crude metric for 'emulation'; present iterations of it distil the principle of 'sampling' intelligent-seeming activities into PURIFIED DECEPTION.

(I keep mentioning that time a computer beat a grandmaster at Chess, and they STOPPED calling chess computers 'AI' because it became obvious that chessplaying ISN'T a high-intelligence activity - instead, great mental discipline is required in order to 'dumb down and math up' for Chess...)

I also think the Turing test is just bad test of intelligence, and it should not be used or taken seriously

With hindsight, it's clear that Turing's 'con-test' is a filter for ability-to-deceive, and has nothing to do with 'intelligent activity'. And visual illusions reveal that the task of 'fooling a human observer' (eg, into thinking a human is observed, rather than a machine) is a very 'low bar'.

dont denigrate my wholesome healthy american drug trips by comparing them to an ai shitting and pissing all over itself

Exactly

No - they are clearly 'falsehoods', but it's equally clear that they AREN'T 'mistakes'. Such programs are DESIGNED to behave in this way, just as a previous generation of so-called 'AI' was designed to play chess in the belief that doing so 'must be' an intelligent activity.

They are BS, here's the paper bsky.app/profile/manj...

Can we just admire the title (and sentiment) of this new journal article for a minute link.springer.com/article/10.1...

ChatGPT is bullshit - Ethics and Information Technologylink.springer.com Recently, there has been considerable interest in large language models: machine learning systems which produce human-like text and dialogue. Applications of these systems have been plagued by persist...

Those words still imply consciousness, though, which isn't accurate

Disagree a statement can be wrong or false regardless of the consciousness of the generator. If you ask a calculator what is 1 + 2 and it says, 7, it is wrong there was no conciousness

That's not at all what I'm saying. I'm saying that words like "mistake" and "falsehood" imply that *the person or article using those words* considers the AI to have consciousness. A tree falling in a forest can't be "mistaken" about it because it's not conscious.

Every single talk and paper.

Many of my colleagues in CS are torn about AI. They recognize the costs and risks, and they feel professionally and morally obligated to meet them through research and teaching. Unfortunately, this means growing in AI rather than rejecting the trend, which would be an abdication. It’s a catch 22.

Personally, I loathe the way entire research communities in CS and beyond have been co-opted by this morally bankrupting tech. However, after a year of grieving the situation and pondering what it will do to my field, I’ve accepted that I cannot just pretend it’s not happening or wish it away.

I wish that while other departments are hiring AI like crazy we would take the opportunity to instead hire great people in other areas that are getting underplaced in this market.

Unfortunately, hiring people isn't something Investors like to see at this time.

I agree. It’s actually an incredible opportunity to pick up top scholars in other sub-fields who are currently being passed over for mediocre AI/ML researchers worldwide.

Agreed- starting my freshmen seminar next fall with this discussion - wish me luck!

Absolutely. And my statement that "AI is a worse environmental disaster than Spotify" *absolutely* landed with my 19th year old undergraduates.

link.springer.com/article/10.1...

ChatGPT is bullshit - Ethics and Information Technologylink.springer.com Recently, there has been considerable interest in large language models: machine learning systems which produce human-like text and dialogue. Applications of these systems have been plagued by persist...

Post