I’m dying right now at the downvotes I’m getting. I’ve been to two major AI conferences this year–are you dorks even in the field? XD
Look, AI is just the name of a broad group of strategies loosely categorized by the sorts of difficulties of problems we are working on (domain size, observability, etc.).
Understanding that this tech is part of artificial intelligence is just understanding the taxonomy of AI related subjects in computer science. That’s it. It’s not an admission that anything is actually smart, especially in a human-like manner.
(just to be clear, you understood I was agreeing, right? 😁)
But nevermind downvotes. The whole voting doesn’t make sense to begin with as people are just too different to have a agreeable understanding on what it means and ladzle the topics too.
I don’t compare millions of pieces of data and choose the one that is statistically correct. I analyse a question and answer based off of my understanding of the actual subject.
LLMs don’t understand shit, they text match and statistically poop out “correct” answers.
LLMs just act on tokens and those are their limit in resolution.
That’s why they’re so bad when it comes to rhyming or counting letters.
They’re not reasoning. And there is a difference between reasoning and giving it the statistical most probable next token - which doesn’t have a meaning whatsoever
No, LLMs are deterministic (or probabilistic, if you want to nitpick). We as a species don’t have the ability to perform nondeterministic computation - many people theorize it may not even be possible.
https://www.anthropic.com/research/global-workspace
At a certain level it does seem were using a non deterministic statistical model to at least mimic true intelligence, though honestly nobody knows at this time the boundaries or complete differences. We’re still on the bleeding edge of development. Also it’s worth noting that most “ai systems” anyone would interact with are actually layers of models, not just a bare llm anymore. Tokenizers, linkers, etc.
Thats the thing people don’t get. Everyone keeps saying “LLMs are not AI”, but forget that there is constant repeated layering of models, fliters, linkers, etc… by developers to make them more efficient. At some point the integration of so many subsystems into one main system, will generate something that appears for all intents and purposes, to be AGI (or at least mimic it to a point where its hard to tell).
Yeah. We’re far far from the days of it just being a simple little language transformer. I don’t know if it’ll ever have the spark of consciousness, but it’s certainly beginning to have the emulation of it.
You can’t compare both. Stark did build an AI.
Claude is simply a statisticaly accurate text generation model… marketed as AI.
It’s like comparing a carburator with oranges.
Most of the major AI conferences havw similar forms of AI being published. It isn’t marketed as such, it quite literally falls under the umbrella.
As much as it pains me, I was going to say that. It sadly is AI by definition.
It’s not what I would call AI (by quite some distance) and I long denied it because of that but my definition is not correct.
I’m dying right now at the downvotes I’m getting. I’ve been to two major AI conferences this year–are you dorks even in the field? XD
Look, AI is just the name of a broad group of strategies loosely categorized by the sorts of difficulties of problems we are working on (domain size, observability, etc.).
Understanding that this tech is part of artificial intelligence is just understanding the taxonomy of AI related subjects in computer science. That’s it. It’s not an admission that anything is actually smart, especially in a human-like manner.
(just to be clear, you understood I was agreeing, right? 😁)
But nevermind downvotes. The whole voting doesn’t make sense to begin with as people are just too different to have a agreeable understanding on what it means and ladzle the topics too.
I knew you were agreeing. I had just come back to find a bunch of downvotes to whether AI is actually AI. XD
What makes actual human intelligence that different from a statistically accurate text/movement generation model?
I don’t compare millions of pieces of data and choose the one that is statistically correct. I analyse a question and answer based off of my understanding of the actual subject.
LLMs don’t understand shit, they text match and statistically poop out “correct” answers.
Didn’t you describe just how our brain is working?
Wasn’t that the point?
To show current LLM models aren’t the same system as our brain
No, I meant
What does “understanding” mean here? Isn’t it analysing millions pieces of data and giving answer that is statistically correct?
Nah, and it isn’t what LLMs are doing now
LLMs just act on tokens and those are their limit in resolution.
That’s why they’re so bad when it comes to rhyming or counting letters.
They’re not reasoning. And there is a difference between reasoning and giving it the statistical most probable next token - which doesn’t have a meaning whatsoever
The statistically accurate arithmetic average dice throw shows a solid 3.5
A dice that always shows 3.5 is bad.
So determinism? Llms are not deterministic.
No, LLMs are deterministic (or probabilistic, if you want to nitpick). We as a species don’t have the ability to perform nondeterministic computation - many people theorize it may not even be possible.
We’re not even sure human thought is nondeterministic to be fair!
Is there a single reputable psychologist out there who believes that this is how humans operate?
Note, I think that non-human intelligence is entirely possible, just not this particular technology.
https://www.anthropic.com/research/global-workspace At a certain level it does seem were using a non deterministic statistical model to at least mimic true intelligence, though honestly nobody knows at this time the boundaries or complete differences. We’re still on the bleeding edge of development. Also it’s worth noting that most “ai systems” anyone would interact with are actually layers of models, not just a bare llm anymore. Tokenizers, linkers, etc.
Thats the thing people don’t get. Everyone keeps saying “LLMs are not AI”, but forget that there is constant repeated layering of models, fliters, linkers, etc… by developers to make them more efficient. At some point the integration of so many subsystems into one main system, will generate something that appears for all intents and purposes, to be AGI (or at least mimic it to a point where its hard to tell).
Yeah. We’re far far from the days of it just being a simple little language transformer. I don’t know if it’ll ever have the spark of consciousness, but it’s certainly beginning to have the emulation of it.
Now it sounds like I’m talking to an LLM.
You didn’t answer my question.
Didn’t know I was supposed to answer your question like a fucking llm. Figured this was a comment section for discussion. 🤷♂️
You failed there. XD
Nah, you’re just not worth it. Enjoy your assumptions and being a dick to people online.
You can be intelligent without being good at writing. However AI relies on text
Good point. What do we rely on?
Not everyone thinks with words. You’re grasping at straws.
No, that was a genuine question. Not words, not images, or any other feelings. So I wonder how is our information(memories) stored and processed.
Ask animals.