How people think AI is going to kill them: terminator robots.
How AI is actually going to kill them: by destroying their habitat and drinking all their water.
@aral In any case that means these LLMs are being used for something they were never meant to do and aren't ready for prime time.
@FinalOverdrive @aral I know. The AI being used and developed is brute force..... an approach adopted years ago to use reams of data rather than reasoning.... and we are seeing that scale up.
@PaddyRoggins @aral Hence the power consumption. We know, but the world doesn't and won't unless we shout it from the roof tops
@aral will be used to justify crimes against minorities as well, after all AI said it was best from some opaque utalitarian–inspired reasoning disguising racism, seksism, ...
@aral you need to make another one where its badly paid humans under the hood
@JohannaMakesGames @aral the main problem is well described by Ed Zitron. Humans under the hood are very bad but are not the main badness about ai.
@JohannaMakesGames @aral
Doesn't AI stand for "Actually, Indians"?
*driven by millions of users who don't care about artist's rights.
@aral the logic of capital
@aral i.e. by terminator robots owners
@aral the greatest trick AI ever pulled was creating the cultural perception that it would be the „slaving machines“ that kills us, instead of the indifferent wills of the money-men who built the machines…
And of course, now they keep recycling the same fears with classism and racism.
And so many sci-fi writers have been blatantly uninformed and ignorant Ito these issues throughout.
@aral And by polluting all the information they need to prevent it from happening.
@aral also by giving wrong advice (like this: https://www.forbes.com/sites/mattnovak/2023/08/12/supermarket-ai-gives-horrifying-recipes-for-poison-sandwiches-and-deadly-chlorine-gas/ ), which will be blindly followed.
And "assisting" internet searches, so we know we should eat:
- petrol in our spaghetti sauce
- glue in our pizza cheese
- rocks
- those mushrooms that melt down your liver
@aral I’m prepared to bet people have already been killed by AI giving them really bad advice. But your story is so much worse.
@billbennett Like people who trust IA for medoacl or electrical repair advice?
I won't be suprised if it has already caused deaths…
@aral Also terminator robots on top of that if you live in Middle East or Africa.
@aral And by inducing people to hate and murder each other by feeding them misinformation as "news" or in search results. Can we rename AI as AS: artificial stupidity!
Edit: someone has very helpfully pointed out to me that AS is an acronym for Asperger`s syndrome, and I wish to clarify that this did not occur to me at the time I wrote this, and I humbly apologise to anyone who might be affected by what was meant to be a tongue in cheek reference to the "intelligence" of LLMs.
@MxVerda @aral Thank you for pointing this out; my bad. I have no desire to suggest an acronym that would disparage any group, but merely to express my frustration that LLMs are regarded as a form of intelligence. The suggestion was made tongue in cheek, but I appreciate that does not excuse the implications that might be made, and for that I humbly apologise.
@HuK If you say so, bot.
@aral
But first AI will take away all viable employment
@aral so funny how people think like that
Don't count out the killer robots. They are coming along nicely. Just because they will be deployed by humans and not by SkyNet and won't be self-replicating doesn't make them not a threat.
All those "human in the loop" systems are being developed with the knowledge that it would be more profitable to take the human out of the loop at some point. Palantir drools about it.
@TheServitor @aral quantum computing will allow real reasoning and less clunky hands will enable that. Will be interesting if the algorithm architects will attempt to harness the power of quantum computing and infuse artificial emotion which is actually how industry is run. We are emotionally driven not data driven no matter how much rhetoric is to the contrary
@aral The luddites ruined the correct machines
@aral don’t forget job elimination
@aral Why not both? Terminator robots / automated targeting systems that drink all the water?
@aral I don't know. Mass suicide and artificial negligence might win this one long before anyone notices all the water is gone. Does anyone actually drink that boring stuff anyway?
And sometimes it's hidden. I was looking up details on making a tincture recently and an otherwise reasonable looking article said to "use a high ABV alcohol such as isopropyl"
@aral My current boss is using AI regularly, I'm currently concerned with the climate expense and not the LLM
@aral
And here I thought AI was going to save us from the oncoming global cholera epidemic by boiling all our water for us.
@aral and they still won't get *general* AI out of in it. Just hallucinating piece-of-shit LLMs that can only churn out spam and incomprehensible text..
At the expense of what little climate stability remains
@aral
Or by advising them to eat poisonous mushrooms
https://www.vox.com/24141648/ai-ebook-grift-mushroom-foraging-mycological-society
@aral That’s the 20-40 year plan.
For the average person, they’re more likely to have their life threatened by an AI:
- Rejecting their medical needs - organ recipient, insurance (US)
- Deciding that cost cutting measures on the product factory floor are worth the risk to safety relative to likely blowback
- Rejecting their job application
- Devaluing the only work they are able to do, being disabled
- Stealing time in productivity’s name from others who might have seen your pain
@aral ... or denying them treatment, or profiling them as a criminal.
@aral There's also the people who'll die to it but not by anything bombastic but because it decided to cut some bureaucratic thread that unravels someone's life entirely.
Making a massive misinformation generator means it'll misinform in small and big ways, some very noticeable but others will be subtle enough to cause some real damage.
@aral That just sounds like Capitalism with extra steps
@aral
Remember when we thought search engines would use more all the electricity and water?
Then it was social media.
It’s not that the technology is unproblematic — glorified predictive text of dubious origin being wildly and widely misused to support a fantasist tech bubble — but there is a pattern here.
@justanotheramy Almost as if there might be detrimental cumulative effects of the ever-increasing resource demands of Big Tech’s (capitalism’s) hype and bust cycles?
@aral or moral panics are recycled when they’re proven distractions.
OpenAI’s Superalignment team was pushing AI as existential threat when distraction was needed from IP… complications…, labour abuses, and bias.
Now that grounding is creating new issues like impersonation risk and exploit vulnerabilities, suddenly we’re supposed to be looking away at the water?
It’s too convenient and too recurring.
@aral If you take any emergent technology and project usage trends consistently forwards in the same direction, it will inevitably create the appearance of impending catastrophe.
But new tech doesn’t continue infinitely upwards, even in a mad bubble.
@justanotheramy true. But it doesn’t have to be infinite. It just has to accelerate current brittleness to the point of fracture. That we’re actually choosing that is pretty bonkers
@aral and denying them life-saving health care (bureaucracy saying, "the AI did it. i was just following its orders")
@aral A thousand years from now aliens visit Earth and sift through the remnants of humanity...
"Our archaeologists have discovered that apparently the downfall of their civilisation started with something they referred to as 'Clippy'."
@aral The waste heat from AI could be used to drive desalination operations in coastal nations. It won't be.
@aral Why not both?
@aral there is a push toward using AI to "automate away" jobs involving critical thinking like contract negotiation and decion making. The problem is that humans are much more driven by emotion than logic. Ever noticed that many people who claim to be "data driven" are actually balls of emotion.
@aral Exactly! I had a discussion on AI with a colleague and when I said I see an overall danger in AI, without being specific, he just threw in the argument„yeah killer robots are terrible but we can regulate them, see AI is just a tool like any other“…
It’s like you say. People have been effectively gaslighted that THAT is the real danger.
@aral By absorbing all their energy. AI consomes copious amounts of energy, the equivalent of a small nation, I've heard it reported? Any reports to the contrary?