āAll of these short-term risks require urgent and forceful attention from governments and international organisations ā¦ā
So, weāre fucked then.
2 Likes
Usually Iād be optimistic or cynical but this guy knows better than most.
1 Like
Iām sure heās right when he says we need governments to act to protect us. Unfortunately we need pretty much all of them to do that. This lot
are working very hard on AI. But theyāre more interested in controlling their own people and screwing everyone else. If Hinton knows that they will, in fact, come to the table though then great.
This has strong Oppenheimer āI have become death, destroyer of worldsā vibes. I think I feel quite sorry for him.
Agreed - in other interviews he says the potential for medicine in particular is one of his favorite aspects of AI. I do get the sense he made a genie and itās not going back in the bottle.
Thatās generous. Heās had an interesting time and a good deal of freedom to do what he wants. Like Oppenheimer, heās so bright he must have known how it was likely to pan out. Thatās how it is when you invent something potent. Ten minutes after someone mastered fire there would be someone else saying āGood servant, bad masterā. And heās got the Nobel Prize.
OK, he lives in a world where the likes of Xi, Putin, Kim, Modi, Netanyahu, Trump etc etc are in charge. But we all do.
2 Likes
2 Likes
Iām still interested in whether AI can do abductive reasoning. Four to five years ago Larson said it couldnāt, and that no-one even knew how to start enabling it to do so.
Now the internet says it can. The piece reads OK until you get to the Conclusion which contains such a howler of a non sequitur that they might as well have written
If Newcastle United spend the money theyāve got for Alexander Isak wisely, that may lead to them winning the premiership in 2-3 years time. This guarantees that AI systems are more open, equitable, and compliant with moral norms in addition to improving their capabilities.
It takes AI to screw up that badly. Well done ChatGPT.
3 Likes
I could well imagine him doing his work, thinking about things like the medical research vectors etc. and feeling like heāll be leaving this incredibly positive dent in the world, only to have VC-backed tech-bro wankers piss it all up the wall producing Studio Ghibli style artwork with off-by-one digit errors, but also stealing all limelight. The whole AI field gets a super shitty rep, and the corners where it is genuinely beneficial get tarnished by the association.
As recent winnersā spending patterns show, it will take a lot more than the value of the Isak transfer to win the premiership 
So it fails the plausibility test for a start.
1 Like
True of a lot of IT Iām afraid. e.g. My hotmail account now wants to send a verification code to my other (nominated) e-mail address every time I try to log in. Fuck knows why - weāve got along just fine for 20 years without this - but there you go. That other address is also a hotmail one. In the early days of Outlook MS wouldnāt let you open two hotmail accounts simultaneously*. In the end they had to cave, so now you can. So, you try to open A@hotmail.com. They block you, saying āWeāre sending a verification code to B@hotmail.comā. You open B, copy the code and paste it, as requested, into the A login screen. The IDIOTS then recognise that B@hotmail.com is already open, so they change your A@hotmail.com login to B@hotmail.com and use the verification code to open an account which a) you donāt want and b) is already open anyway. Then they wonder why people tell them that their work looks like it was specād and implemented by distracted teenagers and, if it was checked at all, that was done by pre-schoolers.
*The whole point of Outlook was that it allowed them to own the whole of you. If you had two accounts then you could keep some of the stuff you did separate from other stuff you did, as in āprivateā. And we canāt be having that can we ?
1 Like
I donāt think ChatGPT has got as far as plausibility.
We live in a world where Donald Trump has been elected POTUS.
Twice.
I doubt plausibility is even on the first round wish-list!
3 Likes
PICNIC
How so ? I am completing the login request to A@hotmail.com. When I type in the code MS have sent to B@hotmail.com someone changes the account that is being logged into from A@hotmail.com to B@hotmail.com. Are you saying Iām doing that ? Because if I am Iām doing it without touching the keyboard or the mouse.
I have a sneaking suspicion that MS are doing it. But if you know better then do explain.
Alas, it has for a lot of people. Thatās less praise for ChatGPT, and more intense criticism of quite a lot of folks.
2 Likes
Use the authenticator app
Thatās the really scary thing about AI. Iām not so worried about SkyNet i.e. the AI harming us of its own volition. Iām pretty scared of some starry-eyed but clueless bean-counter replacing humans with the next PostOffice Horizon, only this time it will be assessing benefits claims or deciding who gets the next transplantable kidney.
EDIT: Dunno how I managed to reply to you Chris - this one was meant for Jon.
Yeah. There are ways round the problem. Itās easy enough to launch a different browser and open B@hotmail.com in that.
My point is that people who are paid money to write this stuff release it with a problem that they expect their customers to work their way around. The whole āIf Ford built cars like Microsoft writes software ā¦ā debate is as old as the hills, I guess. If you want to use it, and these days thatās absolutely a necessity, then you just have to accept that it will be buggy.