Don’t be fooled by the new name for the Windows AI. This is the return of an old friend.
2 Likes
Centrally Located Intelligence Providing Pleasing Ynteractions.
I should have known!
2 Likes
5 Likes
https://twitter.com/PlainSite/status/1662281423339008002?t=QSKhyFvPJuQAYkzBt9GVpA&s=19
https://twitter.com/PlainSite/status/1662282169702821890?t=Ny5Gh_EoPCAUq3JGe1RhyA&s=19
1 Like
The funny thing is this worked on four previous judges before getting caught. (Not really)
Chat gpt can’t even win at tic tac toe.
https://twitter.com/Noahpinion/status/1662866165436399617?t=oq4LfRpsrWWKFudln_F5eQ&s=19
I am utterly confused by this story. I get having it write the brief, seems so lazy not to cite check, but maybe just really in a rush.
However, what I really don’t understand is after being called out by the opposing attorney and judge, to go back to ChatGPT and get copies of the cases, without then at least confirming they are real. I guess maybe his law firm is trying to save money and not have WestLaw or Lexis, but you can usually just google a cite and get the text of most published cases.
- GPT-4 is dumber than an ant!
- How can you tell?
- Have you see how it plays tic tac toe??
1 Like
What if it’s just pretending to be stupid tho
1 Like
6 Likes
Very hard to imagine that this story is true…
2 Likes
Yeah a couple of seconds of critical thought reveals questions like wait, wouldn’t cutting communications with its operator mean it never got any go signals? The operator is in fact the critical instrument by which it is able to complete any missions at all.
Or who have the go signal for the strike on the operator?
How to outsmart (or get outsmarted by) your computer
But isn’t ai going to be very practical and pragmatic without any ethical concerns?
It may not be true, but I think one day it’s plausible.
There are huge ethical concerns with giving humans power over things. They are often wrong or driven by self interest, racism, religious delusions, lack of knowledge without awareness, vanity, etc.
The US elected Trump president. AI couldn’t do worse if it only made decisions based on flipping coins.
I think we need to get powerful AI online asap to counter the threat humans pose to humanity.
In line with the threat of humans to humans, the killer drone thing has been debunked (but who the hell really knows). The source says this was a “thought experiment” not a simulation, i.e. he made it up.
1 Like
I mean it’s a good hypothetical, but it’s not about AI alignment or anything. It’s about ensuring AI is a reliable killer of the things the controller wants killed. This is also a major concern with nonAI weapons systems that necessitates distribution of information and various command and control functions. Even then, sometimes it’s the terrorists, sometimes wedding parties.
At the very apex of nuclear war command and control systems human beings have placed Donald Trump and Putin. AI doesn’t need to be superhuman, it just needs to outperform angry fifth graders, because that’s who is running the show for humanity.
1 Like
I appreciate the Hinton interview for the biography and the history but the views on the future feel a little like this:
https://twitter.com/kareem_carr/status/1664664096850018304?s=20
2 Likes