You don't need to believe that AI is intelligent or "superintelligent" to notice that giving it near infinite compute and letting it loose on the internet is quite likely to have bad results, especially for the internet, possibly more widely
Remote
Tom Walker
@tomw@mastodon.social
mastodon 4.8.0-nightly.2026-10-06Web developer. Sometimes designer. I like code, words and mass movements. Jag lär mig #svenska.
Interested in/may post about #html #css #wordpress #openweb #decentralisation #socialism #protest
2021 Followers
787 Following
50 Posts
Joined January 03, 2017
Birdsite:
Pronouns:
He/him
I'm very much a sceptic when it comes to the idea of AI ending the world yadda yadda.
But if 'superintelligence' could potentially end humanity, couldn't it also just-as-potentially invent a utopian postcapitalist Star Trek-ish future? Doesn't that possibility deserve at least a little bit of the thought and debate that is being spent on the apocalyptic one?
Believing that an AI is conscious or sentient requires you to believe that perfectly comprehensible and replicable mathematics, performed sufficiently quickly, can somehow summon consciousness.
The *impression* of consciousness emerges from the input data: the entire web, all obtainable human literature to this point, etc. It's an echo of that vast corpus – nothing more.
Does AI have risks? Undoubtedly. Lots of them – from fuelling cyberattacks to military uses to the financial bubble.
But to believe specifically in *extinction* risk as the leading threat is clownery, to be honest.
"AI takeover" is a sci-fi scenario and "AI kills us all" even more so. There are much more realistic risks right in front of our faces.
AND, if we're going to talk about risks to humanity, we should look at the actual science around climate change instead of the sci-fi of AI.
We can just unplug the computer.
Well, hmm, OK. Maybe we'll have to unplug lots of computers.
Maybe all the computers.
We missed some computers and they turned the other computers back on.
We're going to have to destroy all the computers.
RE: @kestral@masto.hackers.town
But have we considered that getting the fascists to use Linux might slow them down
I set up a DMARC rua=mailto: for a domain, and uh. Why would anyone want this? I'm now getting spammed to heck with reports from 100 different Yahoo domains
The two emerging "camps" are both wrong about AI.
The extinction people think their sci-fi religion is coming true.
The dismissive people think it's all just a hype and a bubble.
AI is clearly dangerous. Not existentially so, I think! But still dangerous, most specifically to anything that relies on closely-guarded knowledge/"security through obscurity", which includes most computer systems plus many kinds of real-world weapons.
It spreads state-actor level capabilities beyond state actors.
Replying to
You can draw a line correlating "internet becomes bad" and "decline in use of the term information superhighway"
Replying to
@puppygames @khaleer @eobet @digikata @djlink No, Matrix is the thing that happens sometimes in open source where there is an existing "alternative", it sucks massively, but its existence means that no one else tries. See also, GIMP
I'm trying to use Squarespace - didn't choose it obviously, helping a community group with 'fix our website stuff'
And what the fuck?
What the FUCK?
How is this so UTTERLY UNUSABLE and yet so expensive?
It just does. not. function. at a basic level. Making a page and adding some text is like pulling teeth.
It makes Wordpress look like a flying car. I'd rather use Microsoft Frontpage.
"Are LLMs conscious?" is such a dumb question being asked by nominally "smart" people.
It's like inventing planes and then asking "is a plane a bird?"
No. Like, duh. Obviously it is not. It is a man-made thing that is also able to fly.
Replying to
@RMKrug@hachyderm.io @geospacedman@mastodon.social AI imagery is a signifier that a similar level of low effort went into the rest.
"High effort presentation, low effort imagery" is a category that exists in theory but very rarely in practice.
@campuscodi@mastodon.social Is this responding to an actual sideloading coercion problem, or just an imagined one? Most common scams seem like they are already possible without the need to get the user to install a malicious app
It's weird that "nobody ever scans QR codes" was a running joke for *years* but then QR codes suddenly became mainstream and normal
Heartbreaking:
Someone used ChatGPT to correct me about a fact...
... and they were RIGHT.
I think this is the first time that's ever happened
Replying to
@grumpygamer@mastodon.gamedev.place So much in the tech world is lottery winners who think they are geniuses
Boosted by @ferrix@mastodon.online
Replying to on mastodon.social
@bagder@mastodon.social Canary in the coalmine – the whole apparatus of formal bug-bounty programs is surely doomed since it is an incentive to spam and the effort bar to produce plausible spam has been lowered so much
"This is an official email..."
Oh wow, an official email? Wow wow. I can't believe I got an official email. Amazing.
Dear crying big-tech leader,
If you repeatedly tell a significant proportion of the world's population that your aim is to make them unemployed,
what did you THINK was going to happen?
WHY does all AI "graphic design" feature that arched text thing? Every single time!
No real designer does that almost ever.
It's an instant giveaway.
If you needed it spelled out, here it is spelled out.
And remember this when they have some disastrous security incident a few months from now.
Replying to
@ashley it is extraordinarily frustrating, and I think worse when you know what you're trying to do, but it just won't let you do it!
Replying to
@ErikUden@mastodon.de @chbmeyer@digitalcourage.social I think you're way underestimating the possibility that someone there saw the post (probably on Reddit etc), saw your relatively unusual name, found the account and banned it
Replying to
@baldur@toot.cafe The dot-com bubble popping did not collapse the software job market... at least in the medium-term. And I hear a lot of people got cheap hardware out of it. And chairs
Replying to
@RMKrug@hachyderm.io @geospacedman@mastodon.social An incredibly lo-fi image can still be made with care. Someone's questionable colour choices on a map don't necessarily take away from their message.
But using AI shows such a lack of thought, such a lack of care, that it is overwhelmingly likely that the "science" is sloppy or BS.
You can say it's not proven, as if we have to hold aesthetic critiques to some burden of proof, but it hasn't steered me wrong yet.
Replying to
There is a persistent ethical issue in "tech" spaces where people act like technology is neutral and no one wants to judge the people building Killbot 9000, because after all it is merely "tech" and does not carry any inherent moral position
One under-discussed difference between (what I'll call for shorthand) "the average person who might be reading this" and the population at large is that, whatever you think of AI, *you know what AI 'writing' sounds like.*
So: when someone sends you an undeclared AI-written message, you know what they did.
This leaves us with a gradually increasing number of social minefields to navigate (or at least messages to ignore/dismiss) in a way that is not currently the case for most people.
Replying to
@Gargron@mastodon.social @sixohsix@icosahedron.website starts as the second but then becomes the third
Replying to
@creepy_owlet@mastodon.online @RMKrug@hachyderm.io @geospacedman@mastodon.social An "IT specialist" is absolutely aware of the problems, or incompetent in their field
Replying to
@nullagent@partyon.xyz @jack@mastodon.sdf.org People shouldn't delete things like this from their phone based on random (and it seems disputed) social media posts. This is the slightly smarter equivalent of "delete System32"
Replying to
@bornach@fosstodon.org @stefan@front-end.social Yeah, my second sentence would point out that it is the most likely or most "average" next word being predicted (yes I know there is actually some randomness injected). So in other words the writing produced is, by definition, mediocre
Replying to
@ldodds They really will propose anything other than paying for the really basic maintenance that will fix this stupid issue
Replying to
@gytisrepecka@social.gyt.is @nullagent@partyon.xyz @jack@mastodon.sdf.org If you read the rest of the thread there are *loads* of caveats to that. My point is don't delete things with names like "system safety core" because you saw a viral social media post.
OK, is it just me or have you noticed that Mr Blue Sky turns into a Katamari Damacy level track halfway through
Replying to
@stefan@front-end.social This is good and it's basically the same as the one-line explanation I usually give for "how AI works". The vast majority of people do not know this and generally believe that AI models "simulate thinking" in some sense.
After AI destroys the internet, let's go all in on bringing back CD-ROMs
Replying to
@osma@mas.to When fake federation hits legal reality, you notice that it is possible for random regimes to send a demand to the one real central authority, and that's enough for game over
A lot of chat about whether AI will "kill us" is either about AI that eventually decides that humans are useless, or ignores us like ants, or decides to turn us into paperclips (or whatever) in the incidental pursuit of some other goal.
What's almost missing from the discussion is that militaries are putting AI in drones, right now, and sending them out into warzones.
You don't need AI to "go rogue" if people are *intentionally putting it in autonomous weapons*.
Part of me is annoyed by AI but part of me also feels like it's not that deep really. The widespread use of microwave meals has not ended cooking or restaurants. You can embrace pot noodles or think they are awful but regular noodles were not displaced. I get a little pain somewhere when I see an AI takeaway menu with that stupid paint-swoosh text but it doesn't feel like graphic design is over forever
Replying to
@brewsterkahle@mastodon.archive.org It's a signoff blob (as seen at the end of some magazine articles) but because there is a character that is a circle with an R in it, they have used that
Replying to
@computer@facts.computer Computers have been a weird journey from "outsider uncool" to "outsider cool" to "mainstream cool" to "about as interesting as a vaccum cleaner"
The Mainstream has discovered x-risk chat and doesn't really know what to make of it
Replying to
@cwebber@social.coop But they're going to implement decentralisation real soon now, for sure!
(A dump truck full of money arrives)
Uh... so uh, we've all had an incredible journey...