Why you should apply to the Hockey Beat

[Trotsky]

believes AGI will arrive within the next five years.
https://arcprize.org/arc-agi

Every year since 1960.  cf. cold fusion.

It's not real, my friend.  It's an illusion.  War's over; Hubey dropped the big one.  Even MIT eventually admitted it.

https://components.one/posts/building-heideggerian-ai

I'll read this stuff at some point but just FYI your form of argumentation is not persuasive. You're citing as gospel theory that was written decades before the major advances of the past 5-10 years.

Well, I don't mean to.  The point is the fundamental approach of AI is based on a misreading of what consciousness and intelligence are.  What changed in the last 5-10 years was a shit ton of money got behind it so now we're all going to find out quickly what a handful of pointy headed philosophers figured out 50 years ago.  

Which is good, actually.  AI is really cool in a lot of ways.  It just isn't "I."  It's math.  Math is great, I keep a lot of my stuff in it.  But it is a model.  The world is going to find out that the model of human cognition represented by math for the last 500 years is a model, not the actual thing.  The map, not the territory.

Descartes wanted to understand houses, so he drew a crude picture of a house.  Now he had a picture, so he could do all sorts of things with it: measure it, break it into pieces.  Centuries passed, and we got better and better at pictures.  A pretty weird guy named Frege developed a beautiful method of reducing all the components of houses to a syntax.  An infinitely weirder guy named Wittgenstein had Deep Thoughts about this and announced he had drawn the final, inexhaustible picture of The House and no others would be necessary, then 20 years later recanted, scratched it all out, and announced houses were really just whatever we agreed at any given time was a "house."  And then finally we had computers that could support tremendous CAD images of houses.  And with that library finally "AI" is able to generate new images of houses which never existed before.  That is remarkable and can be used for all sorts of purposes.

BUT YOU CAN'T FUCKING LIVE IN ONE!  It's still just an image.  It's not a house.  It doesn't keep the rain off your head.  It's in your head, it's not in the world.  That's the difference.  You want to tell me about the great new advances in imagery, and yes there have been.  But it's still all just imagery.

"Ceci n'est pas une pipe"?

Indeed.  Whole damn  chapter about it here.

This stuff isn't new.  It's just new to the Masters of the Universe, so puffed up by their own grants.

[ugarte]

AI is a statistically-weighted revision of this.  It isn't intelligence.  It isn't consciousness.  It doesn't have the self-awareness of a bumble bee.  And it will never have any of these.  It isn't undeveloped cognition, it is utterly a zero value.  It's a probabilistic calculator.

It's a blank canvas for our anthroporphism.

To the extent these things are true of AI, they are also true of humans. What we call human ingenuity, self awareness, or consciousness is really just us regurgitating data, similar to what an AI does.

Your post is a common retort about the limits of AI, but I think it's cope about a lack of human exceptionalism. Whatever limitations AI has now, it probably won't have them in 5-10 years (or sooner).

Nope.  This debate was fought and won 60 years ago.  STEMmies just never got the news.

They are playing with dice.  What's in our heads isn't dice, it's what happens when dice realize oh fuck me I'm gonna die that sucks!

The unexamined ontological assumptions underpinning computer science and the hard sciences are as obsolete as luminiferous aether, and have been for nearly as long.  As always, the awkward little boys with the super advanced toys who can't get laid (yes, Pat Churchland, fine, and one girl) but can destroy the world are working with dark age philosophy.  But they're well paid and get A's, so, hey, what's the diff right?  

Read a book, dweebs.  There are more things in heaven and Earth.

Interesting discussion of this in today's New York Times.

Gary Marcus is a notorious AI pessimist. For an alternative perspective I'd suggest Francois Chollet. Unlikely Gary, Francois is on the ground working in AI on a day to day basis. He designed the ARC-AGI test  to evaluate whether AI can acquire new skills outside of its training data. Recently, AI passed this test for the first time. Francois (who is not a used car salesman-style AI promoter like some people in the field) thus believes AGI will arrive within the next five years.
https://arcprize.org/arc-agi

AI that solves some neat tasks will exist. aGi is, depending on how you define it "a new form of intelligence" or "not intelligence in the way it is manifest in people." the processing of training data in ai is wholly unlike the way humans "train" their interpretive tools. the two paths are not recognizably similar and they will not - probably can not - converge. there will be overlap in mechanical skills that computers will be better at. there will be the illusion of convergence, like a chatbot that can convince you to be brave or send money or kill yourself with empathy. but it won't ever be the same thing. a chatbot that learns to mimic the desperation of someone pleading with their girlfriend not to leave will never actually care if the girlfriend leaves.

[Trotsky]

I think Bearlover is saying that's us, too.  We are just strings of impulses that evolved competitively.  Caring is an illusion and what we are is a text processor that happened upon outputs that gave survival advantages.

He's not saying AI is conscious, but that we aren't.

That's an untestable thesis, because whenever we report a case of care he can respond "and that's just what the evil genie fools you into thinking."  

Now, Descartes responded "AHA!  I have you, because you just admitted there is a YOU that is THINKING.  Bwahaha!  Checkmate!"  But that's dumb.  My refutation of the Churchlandian assertion that will is an illusion is phenomenological:



Our lived reality includes our awareness, consciousness, will, and care.  All we are is that lived reality.  Therefore, it's real.  It is the psychological equivalent of a Durkheimian social fact.  AI lacks that because it lacks an "inside."

Camus made the argument that you see this dramatically when you are confronted with a crisis.  You switch out of auto-pilot and are suddenly directly confronted with the world and recognize your will is a real thing, because right now you are whining that you are NOT HAPPY.  

The proof that we are aware is bitching.

[Beeeej]

I think Bearlover is saying that's us, too.  We are just strings of impulses that evolved competitively.  Caring is an illusion and what we are is a text processor that happened upon outputs that gave survival advantages.

He's not saying AI is conscious, but that we aren't.

That's an untestable thesis, because whenever we report a case of care he can respond "and that's just what the evil genie fools you into thinking."  

Now, Descartes responded "AHA!  I have you, because you just admitted there is a YOU that is THINKING.  Bwahaha!  Checkmate!"  But that's dumb.  My refutation of the Churchlandian assertion that will is an illusion is phenomenological:



Our lived reality includes our awareness, consciousness, will, and care.  All we are is that lived reality.  Therefore, it's real.  It is the psychological equivalent of a Durkheimian social fact.  AI lacks that because it lacks an "inside."

Camus made the argument that you see this dramatically when you are confronted with a crisis.  You switch out of auto-pilot and are suddenly directly confronted with the world and recognize your will is a real thing, because right now you are whining that you are NOT HAPPY.  

The proof that we are aware is feeling like a sniveling New Yorker.

These are the sorts of debates you really don't see on the Ferris State hockey discussion board.

[ursusminor]

I was debating on posting this here for a few days, but I might as well.

This article, quoting the Princeton Review, states that it was calculated that RPI is the second happiest college in the US! https://resume.io/blog/the-happiest-schools-in-the-us-uk-and-australia

Had I just seen that, I would have dismissed it as AI, but at the moment, it is on the front page of rpi.edu ::yark::

[Trotsky]

What do you think of that?  I gotta say, when the RPI student experience comes to mind, based on alumni I have known, "happy" isn't the first word association I'd have made.

[scoop85]

What do you think of that?  I gotta say, when the RPI student experience comes to mind, based on alumni I have known, "happy" isn't the first word association I'd have made.

I know several RPI alums and current students, and I concur with that assessment.

[Trotsky]

Having said which, it's one of the most fiercely proud group of alumni I know.  They really loved the education they got.  It just wasn't a joyful experience.

OTOH, I know a lot of Cornellians who were really happy even in media res.

[marty]

What do you think of that?  I gotta say, when the RPI student experience comes to mind, based on alumni I have known, "happy" isn't the first word association I'd have made.

I know several RPI alums and current students, and I concur with that assessment.

There is a reason there were "Tute, Love It or Leave It" tankards and glasses well into the 70's.  Maybe the Love It crowd is more vocal than the rest.

[adamw]

my turn to say I agree a billion percent with Trotsky (I know there's no a billion percent, don't at me). I've been having these same conversations with my wife and her AI-loving friends. I win these arguments every time -- at least, in my limited mind.

[adamw]

Well, I don't mean to.  The point is the fundamental approach of AI is based on a misreading of what consciousness and intelligence are.  What changed in the last 5-10 years was a shit ton of money got behind it so now we're all going to find out quickly what a handful of pointy headed philosophers figured out 50 years ago.  

Which is good, actually.  AI is really cool in a lot of ways.  It just isn't "I."  It's math.  Math is great, I keep a lot of my stuff in it.  But it is a model.  The world is going to find out that the model of human cognition represented by math for the last 500 years is a model, not the actual thing.  The map, not the territory.

Descartes wanted to understand houses, so he drew a crude picture of a house.  Now he had a picture, so he could do all sorts of things with it: measure it, break it into pieces.  Centuries passed, and we got better and better at pictures.  A pretty weird guy named Frege developed a beautiful method of reducing all the components of houses to a syntax.  An infinitely weirder guy named Wittgenstein had Deep Thoughts about this and announced he had drawn the final, inexhaustible picture of The House and no others would be necessary, then 20 years later recanted, scratched it all out, and announced houses were really just whatever we agreed at any given time was a "house."  And then finally we had computers that could support tremendous CAD images of houses.  And with that library finally "AI" is able to generate new images of houses which never existed before.  That is remarkable and can be used for all sorts of purposes.

BUT YOU CAN'T FUCKING LIVE IN ONE!  It's still just an image.  It's not a house.  It doesn't keep the rain off your head.  It's in your head, it's not in the world.  That's the difference.  You want to tell me about the great new advances in imagery, and yes there have been.  But it's still all just imagery.

Here is what ChatGPT itself has to say about Trotsky's critique ....

1. AI as Math, Not "I"
The author argues that AI is just "math" -- models and statistical systems -- not true intelligence or consciousness.
AI's perspective: That's essentially true. Modern AI (especially deep learning) is built on statistical pattern recognition and optimization, not on conscious experience or self-awareness.
Philosophical angle: This reflects the map vs. territory problem: AI produces symbolic or mathematical "maps" of the world but does not inhabit the territory of lived experience. This echoes Alfred Korzybski's idea ("the map is not the territory") and critiques of representationalism.

2. Historical Framing: From Descartes to Wittgenstein
The house metaphor illustrates the progression of thought:
Descartes: first crude pictures -- early attempts to model cognition.
Frege: formal logic and syntax -- symbolic representation.
Wittgenstein (early): language as a logical picture of the world -- the "Tractatus" view.
Wittgenstein (later): language as use and agreement -- the "Philosophical Investigations" shift.
AI today: powerful computational images/models -- generative systems.
AI would agree that this is an apt genealogy. In fact, many in cognitive science argue that current AI reflects a long history of reducing thought to symbols, structures, and formal systems.

3. The "You Can't Live in It" Argument
The analogy says: an AI model is like a picture of a house. You can admire it, manipulate it, and generate endless variations -- but you can't actually live in it.
AI's perspective: Correct. Models are abstractions. They simulate or generate representations, but they don't embody physical reality or consciousness.
Counterpoint from AI researchers: While you can't "live in" a CAD drawing, you can use it to design a real, livable house. Similarly, AI outputs -- even if "just math" -- can shape the world when coupled with action (robots, decision systems, medical tools).

4. Money and Acceleration
The critique that "a shit ton of money" made AI explode is accurate historically:
The last decade saw unprecedented funding in AI research, leading to scaling laws -- the realization that bigger models with more compute/data can achieve surprising abilities.
AI researchers would acknowledge that while the philosophical critique isn't new, the practical consequences of AI systems becoming widely deployed are unprecedented.

5. AI's Likely Reply
If AI itself were to respond to this critique, it might say:
"You're right that I am not consciousness. I am a tool that works with patterns and representations. But representations can be useful -- even world-changing. A map may not be the territory, but without maps, you'd get lost."

Bottom line: The critique is valid: AI is not conscious, it is representational and mathematical. But that doesn't negate its power -- just as architectural drawings aren't houses, yet without them, we wouldn't build skyscrapers. The tension lies in remembering the limits of AI models while appreciating their generative and practical potential.
[\\quote]

[scoop85]

Kudos to ace reporter Jane McNally for getting the win in her first career start in goal for the Cornell field hockey team, which defeated Colgate today 2-1.

[Weder]

Kudos to ace reporter Jane McNally for getting the win in her first career start in goal for the Cornell field hockey team, which defeated Colgate today 2-1.

Kinda cool to see a field hockey field in Barton. (Are they putting in and taking it out for every game, or is it just there for the duration of the season?)
EDIT: Oh, it looks like they're still hoping to be able to play at the new facility at some point this season.

[Old Red]

All I want from AI is the answer posited in Asimov's "The Last Question"

[Trotsky]

All I want from AI is the answer posited in Asimov's "The Last Question"

Insufficient data for a meaningful answer.