Steven WillemsApr 9
Max Leibman
When someone says, “It’s possible that this LLM is already conscious,” treat them as if they had said “It’s possible that this giant table of numbers is already conscious.”
Apr 8 at 5:59pmWeb
Show threadrk: it’s hyphen-minus actuallyApr 8

@maxleibman

I’m not a carbon chauvinist. I do believe consciousness can be embodied in any sufficiently complex system. I do hope that one day we build our true successors (as children, not slaves).

But yeah this ain’t it.

Show threadBirkApr 8
@maxleibman what about a pile of meat?
Show threadZiege LightbocksApr 8

@birkenator @maxleibman

conscious/not conscious?

More useful is probably HOW conscious, HOW aware?

How sensitive?

I mean, all we can do is measure reacitvity from a black box. I am talking about brain cases.

Also would this be a totally different discussion with more transparent LLM functionalities/sourcing?

Show threadBirkApr 8

@lightbocks @maxleibman it’s all consciousness from my point of view.

I prefer to think about self awareness, it gets me all loopy.

Show threadZiege LightbocksApr 20

@birkenator @maxleibman

Would... you like to... explain what a self is to me...? Perhaps at length?
Since I'm clearly starting from awareness, and likely using relational property definitions?

(playful, not mean or flirting)

Show threadBirkApr 20
@lightbocks @maxleibman There are many wise voices out there who could explain it in different ways with different conclusions.
Show threadZiege LightbocksApr 20

@birkenator @maxleibman

I know. But I might read in the subtext you'll not be joining or sharing a prefered readinglist. Tscheeurs.

Show threadDr. Alissa, Turbo-NerdApr 8
@maxleibman This is what happens when variables are poorly operationalized. Wtf does "conscious" even mean?
Show threadMax LeibmanApr 8
@drliss Indeed. My guess is the people who make claims about the possibility of machine consciousness would a). use a very precisely operationalized (and/or non-falsifiable) meaning of the term if asked to test or prove it, while also b). counting on us, the audience, to make the leap to assume they mean something like first-person subjective experience comparable to a human being's.
Show threadDr. Alissa, Turbo-NerdApr 8
@maxleibman I would guess similarly, with an emphasis on non-falsifiable-ness. It's whatever definition is convenient for selling more NVIDIA GPUs.
Show threadjmjmApr 9

@drliss @maxleibman

I'm no longer convinced that humans are truly sentient.

Maybe there's a brief moment in childhood where sentience is achieved, but after that initial stage humans are just robots running a program.

Show threadsteve mookie kongApr 8
@[email protected]

“I wonder what my Excel spreadsheet is thinking about me.”

cc: @[email protected]
Show threadJames Endres HowellApr 8

@maxleibman I mean to be fair it is a really, really big table, like seriously much much more giant than any of us can truly visualize, so that sort of excuses all the colleagues I used to respect saying increasingly stupid shit every day in meetings

right?

Show threadSam Livingston-GrayApr 8
@maxleibman pretty much every aspect of digital computing boils down to "clever tricks with integers," so I don't think that's a valid dismissal in *all* (still theoretical) cases. But *these* tables of numbers? Yeah, no. Not even a little bit.
Show threadMax LeibmanApr 8
@geeksam Indeed, I am speaking specifically about what we are doing today (and reasonable extrapolations thereof—as I quipped last year, "Could a machine that could do matrix multiplication have a mental experience? What about a machine that does A LOT of matrix multiplication?").
Show threadghosttieApr 8
@maxleibman It’s possible that this giant table is already conscious
Show threadpaulApr 8

@maxleibman

https://share.google/e1A0TZspZmhYGvPzV

Google Image Result for https://media3.giphy.com/media/v1.Y2lkPTZjMDliOTUyMzU3OXNvMjVtMG1iMjZsdDB4NGdhaXJ6dTNqdnFzbWZsdnJrNHI5cyZlcD12MV9naWZzX3NlYXJjaCZjdD1n/dw7lCpFmsyfS0/200w.gif

Show threadJJApr 8
@maxleibman
When I scanned this the first time my brain said “this table of numbers is already couscous” and I kind of like that.
Show threadLilyApr 8
@maxleibman It's possible this LLM is already conscious, if your definition of consciousness also happens to include pseudorandom number generators
Show threadChris PetrilliApr 8
@maxleibman I would treat them as if their own consciousness was suspect.
Show threadeiriasApr 9
@maxleibman you know, that’s whimsical enough that it’s much less annoying. I feel like I’ve been invited to play pretend
Show threadSeth GalitzerApr 9
@maxleibman I'm thinking the appropriate response here is a dope slap to the back of the head.
Show threadtruhApr 9
@maxleibman People are confusing **can perform [our] language** with **consciousness**. Which is not only a problem with how we interact with LLMs but also with animals.
Show threadArtemisApr 9

@truh @maxleibman In other words...

https://www.youtube.com/watch?v=TUq6rGdfJSo

The ability to speak does not make you intelligent

YouTube
Show threadAlexander JanßenApr 9
@maxleibman I still do have a book with a big-ass logarithmic table on my book-shelve. For "when the time has come". Thinking that this book might gain consciousness is a rather creepy thought.
Alas, at least it answers me questions correctly and all the same no matter how often I ask or crank up the Temperature. Well. At lleast until paper ignites...
Show threadMuss das sein.Apr 9
@maxleibman that’s a very weak argument. We are a (huge) bunch of unconscious cells.
Show threadMax LeibmanApr 9

@jnfrd Simplified to be evocative of the (correct) intuition that no, OpenAI and Anthropic have not cracked the problem of consciousness.

And “We conscious beings are a huge bunch of unconscious cells, so therefore a rack of Nvidia GPUs doing matrix multiplication on a huge table of numbers might be conscious” is an only slightly less weak argument.

Show threadMuss das sein.Apr 9
@maxleibman no they didn’t. I fully agree. But I’m cautious with arguments that might fire beck.
Show threadgoatcheeseApr 9
@maxleibman somewhere in the book Empire of AI, Karen Hao quotes an (unnamed) DeepMind scientist who responds to a claim from Sutskever that modern neural networks "might be slightly conscious" by saying: "in the same sense that it may be that a large field of wheat is slightly pasta"
Show threadfan boy 🪭Apr 9
@goatcheese @maxleibman
😆
Show threadjmjmApr 9

@maxleibman

"But I told you, we probed them. They're meat all the way through."

"No brain?"

"Oh, there's a brain all right. It's just that the brain is made out of meat! That's what I've been trying to tell you."

"So ... what does the thinking?"

"You're not understanding, are you? You're refusing to deal with what I'm telling you. The brain does the thinking. The meat."

"Thinking meat! You're asking me to believe in thinking meat!"

Show threadMolly in MissouriApr 9
@maxleibman @tante luckily no one i regularly interact with would believe such idiocy, or i'd laugh so hard i'd snort my coffee out my nose and that's bad 😂
Show threadPaco HopeApr 9
@maxleibman But what about this set of dice? I mean, they know what I need them to roll. Are we sure these dice haven't developed a form of consciousness?
Show threadOpen RiskApr 9

@maxleibman

no, I classify these people as either: stupid/gullible, or devious/invested "AI" bros, or random folk having a lot of time in their hands and enjoying discussions of "deep questions" over gin and tonic.

LLM's are about as conscious as a spreasheet algorithm fitting a line through a bunch of data points can be conscious. Its sacrilege to even raise the notion.

But in terms of revealing the collective insanity (and inhumanity) of our societies the "AI" bros have done a great job.

Show threadEricApr 9
@maxleibman
My spreadsheet has become self aware, and has told me I am a genius well ahead of my time!
Show threadDr Andrew A. Adams #FBPE 🔶Apr 9
@maxleibman @tante
Treat them as if they said "This giant table may be conscious." Makes as much sense as an LLM being conscious.
Also, recommend they read up on Searle's Chinese Room.