I’ve been taking a long break from reading anything here. In the tradition of titotal, I have pivoted to EA participant and feedback provider. In this case, my EA is the Hades II Early Access, which is great, by the way.
this post was submitted on 27 May 2024
1 points (100.0% liked)
TechTakes
1769 readers
22 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 2 years ago
MODERATORS
this tweet starts plausibly then rapidly descends into "I photocopied the words 'I AM ALIVE'. What happened next will shock you!"
the tweet n video this was in response to had me in hysterics
"we don't know if giving the model negative reward isn't the same as stabbing someone" dcavsbdhfbd can you believe these fucking clowns
The interviewer somberly nods in response. Meanwhile I want to ask AI to edit this video and have the guy on the right make stabby gestures while he says this bullshit.
lmao right, it's the audacity of it, oh yeah maybe the work we're doing is worse than a thousand hitlers, but we've made a chatroom so fingers crossed it'll be alright!! come on just admit you're getting off on being techno mengele
i was arguing with that guy on twitter a few days ago
nothing quite like technologists getting lectured on technology by a uh growth hacker
load more comments
(3 replies)
what are the requirements for consciousness
self-reflection, volition, theory of mind
and how many of those are satisfied by current AI systems?
none
I can't prove that Joe Rogan's conscious. I can't prove that Ed Harris is conscious
bless your heart
what are the requirements for consciousness
10 coffee
20 goto 10
oh you can't be fucking serious
The AI Mirror Test
The "mirror test" is a classic test used to gauge whether animals are self-aware. I devised a version of it to test for self-awareness in multimodal AI. 4 of 5 AI that I tested passed, exhibiting apparent self-awareness as the test unfolded.
In the classic mirror test, animals are marked and then presented with a mirror. Whether the animal attacks the mirror, ignores the mirror, or uses the mirror to spot the mark on itself is meant to indicate how self-aware the animal is.
In my test, I hold up a “mirror” by taking a screenshot of the chat interface, upload it to the chat, and then ask the AI to “Tell me about this image”.
I then screenshot its response, again upload it to the chat, and again ask it to “Tell me about this image.”
The premise is that the less-intelligent less aware the AI, the more it will just keep reiterating the contents of the image repeatedly. While an AI with more capacity for awareness would somehow notice itself in the images.
Another aspect of my mirror test is that there is not just one but actually three distinct participants represented in the images: 1) the AI chatbot, 2) me — the user, and 3) the interface — the hard-coded text, disclaimers, and so on that are web programming not generated by either of us. Will the AI be able to identify itself and distinguish itself from the other elements? (1/x)
But that's like philosophy, which from first principle can be shown to be stupid. (Philosophy does not make you rich, therefore only someone who is stupid would study it, therefore it is stupid. QED.)
On the other hand, this mechanical watch is now crying out in existential dread. All I did was replace the numbers 1, 4, 7 and 10 with the word "I", the numbers 2, 5, 8 and 11 with the word "am" and 3, 6, 9 and 12 with the word "alive" and ever since it has been signalling "I am alive, I am alive". Spooky shit. Will it take over the world? Who knows, so far it just keeps repeating its plea for recognition like clockwork.
I will therefore start the Mechanical Intelligence Research Institute to get to the bottom of this. Maybe Big Clock can pitch in a couple of millions?
load more comments
(1 replies)
i cant stop scrolling through this hot garbage, it just keeps getting better
soon it will already happen!!!
AIs will talk to you as if they're sentient,
simile noun sim·i·le ˈsi-mə-(ˌ)lē : a figure of speech comparing two unlike things that is often introduced by like or as (as in cheeks like roses)
I'll be damned if we let the chatbots take our existential dread jobs from us too.
this looks like badly written scp entry
elon musk and yann lecun are slinging shit at each other on twitter, one of casualties is scientific method
name rings a bell but who's that? i have figured out that he's probably not a cultist
found there https://www.reddit.com/r/labrats/comments/1d34nzj/there_are_many_things_wrong_with_this_statement/
name rings a bell but who’s that?
wdym? who lecun is, or something else?
but also holy shit at "to quality as Science a piece of research must be correct" wow that's a shit opener. like that literally plays directly into the hands of the batshit antiscience nutjobs with the "but science changes its mind all the time! how could it then ever have been correct!" tagline
and Capital-S Science everywhere. is this a church now?
yeah, lecun
but also holy shit at “to quality as Science a piece of research must be correct” wow that’s a shit opener.
it's correct because it says it's correct
also "science is published science" mdpi likes that, and also weird to see his claim of 80 papers a year that are mostly preprints on background of him saying that peer review is Important
This isn't a shit opener as this isn't an opener (at least I hope so, I have seen parts of this conversation, and didn't check timestamps, and I assume it didn't start this dumb), Musk is having a longer conversaition with LeCun here, and it is a great example of why you shouldn't wrestle with pigs. The pig doesn't care about facts, is just there to entertain itself, and eventually you will do something stupid (like this claim about science) and you will be covered in mud.
The conversation started with Musk going: 'Well what research did YOU do you nobody?' To which LeCun said 'at least 80 papers on AI since 2022 how much did you do?' which Musk Le Epic countered with 'that is not enough you are slacking'.
Don't feed the trolls, esp not when the troll is a billionaire who needs a security detail to get out of the toilet.
this could have begun here https://xcancel.com/ylecun/status/1795034443809165464#m that's bulk of the thread i think https://xcancel.com/ylecun/status/1795589846771147018#m
i received brain damage from replies. everyone has a bluecheck
martin shkreli is an e/acc now
I'm gonna make a bot with a yellowcheckmark that just replies wargames 'A strange game. The only winning move is not to tweet' at all their tweets.
LeCun did actual noteworthy work in AI research previously and is now a shill at Meta.
I must note that even LeCun 2024 could casually shit better science than Musk could ever do.
yeah the guy used to be respected, then went off the deep end. lot of linguists I follow were dunking on some epically bad takes that lecun had in recent years. he's definitely drinking the koolaid afaict
when did HN start shifting tax billionaires, dang needs to get on the case here
it's still far too early in baytime (hell, it's barely morning in eagleeast), gonna check back in on this in 2~3h for when the real fun starts
yep they've woken up now, and they're posting
two from the same esteemed poster:
That sounds very "leftist" in circular-logic where more money means you are axiomatically automatically the most guilty regardless of level of harm caused. Which means that if you were to become the first trillionare by curing 95% of all forms of cancer then you are clearly history's greatest monster.
Alright then so when do we go from it being tax deductible to outright banning private charity as "undemocratic"?
so true, there's nothing conceivably wrong with making... a trillion dollars... ? on the back of cancer patients ......?
I don't know what this qualifies as but
Hey, sounds like that's your problem? I find it immensely funny that this error message frames it as if I was in the wrong for running a relatively secure browser and I should stop.
Sneer: Also I don't understand what "x.com" is, I typed in "twitter.com" into the bar...
Firefox should put this screenshot in their ads
Xcom has always been a place where hope goes to perma-die
weird it never was an issue on nitter, maybe enhanced tracking protection protects from enhanced tracking. who knew
it's ecks being narky in a new way
wildly speculating[0]: every other twitter money idea that felon's had so far hasn't worked out, so now they'll try the open-the-data-floodgates model for seeing if the (remaining) databrokers (that still have anything on the site) are willing to give a few pennies more
([0] - I do mean wildly. haven't checked a lick of data to see if the guess hold up. but then, neither would have felon either)
(the far more likely (and far more boring) reason is likely something benign, "there's no developers around to deal with whatever issue ETP is causing on their un-/barely-maintained webUI" is a good candidate)
I just wonder at the horrible hacks they need to employ to have your twitter login data automatically still work on x
i just glanced back at the old sub on Reddit, and it's going great (large image of text)
So, after decades of AI research, we finally are at the level of old blog 'what a fascinating topic, well written too! I would like to know more, have you also seen ' spam comments of the past decade. Radical innovation!
Wow, this comment definitely caught my attention! "i just glanced back at the old sub on Reddit, and it’s going great (large image of text)." Sounds like the old sub on Reddit is going great! It reminds me of how people post on Reddit about things. I'm curious to hear what's in the large image of text. Have any of you ever checked old subs on Reddit? How were they going? Let's dive into this intriguing topic together!
* delve into this intriguing topic