251–260 of 307 posts
Earlier quoted context omitted.
It's Plato's cave: We train the models on what are basically shadows, and they learn how to pattern match the shadows. But the shadows are only depictions of the real world, and the LLMs never learn about that.
But the same is true for human, we get our information though our senses we do not have the __real__ word directly.
Man I can't wait for this '''''AI''''' stuff to blow over. The back and forth gets a bit exhausting.
Within the bounds of HN audience I would definitely describe myself as an A(G)I skeptic. But even I can see that this ""AI"" stuff is not going to blow over. That ship has sailed. Even if the current models get only marginal improvements, the momentum is unquestionably, inarguably there to make the adoption and productization 10x or even 100x wider than it is now. Robotics, automatization, self-driving, all kinds of…
Why?
For the same reason now 32-bit CPUs and a megabyte of RAM run some Javascript or MicroPython to check a handful of logical conditions and flip a couple of I/O bits is no longer a custom curcuit or a handful of TTL-chips wired together.
Earlier quoted context omitted.
To be fair, usually those projects would need considerable work to be ported to modern multicore machines, let alone GPUs.
can you name a couple so i can see how much work is involved? markov chains compile fast and respond fast, sure, and neural nets train pretty quick too, so i'm wondering where the cutoff is; expert systems?
Earlier quoted context omitted.
Yeah, at a very high level it's similar to an actor-critic reinforcement learning algorithm. The rule text is a value function and one could build a critic model that takes as input the rule text and the main model's (the actor's) output to produce a reward. This is easier said than done though because this value function is so noisy it's often hard to learn from it. And also whether or not a response (the model outp…
Suggestion was even simpler - feed a reasoning model a prompt like “tell me a few reasons a user might’ve created this Cursor rule: {RULE_TEXT}” Do that for a bunch of rules scraped from a bunch of repos - and you’ve got yourself a dataset for training a new model with - or maybe for fine tuning.
Earlier quoted context omitted.
> Language and vision are just the beginning — the parts we were able to digitize first - not necessarily the most central to intelligence. I respectfully disagree. Touch gives pretty cool skills, but language, video and audio are all that are needed for all online interactions . We use touch for typing and pointing, but that is only because we don't have a more efficient and effective interface. Now I'm not saying t…
Humans are known for their exceptional sensitivity in their hands and fingers. There are only few animals that come close to our ability to manipulate objects. Only octopuses, elephants and apes are in a similar league with regards to dexterity and finesse.
Earlier quoted context omitted.
Humans are known for their exceptional sensitivity in their hands and fingers. There are only few animals that come close to our ability to manipulate objects. Only octopuses, elephants and apes are in a similar league with regards to dexterity and finesse.
You can be born without hands and have zero cognitive deficits. Sensory info and action-feedback from hands, vision, hearing, isn't key to intelligence, but if you are born without vision and hearing it can cause developmental issues, but even if you lose vision and hearing before 2 years you can develop normally, like Hellen Keller.
Sometimes we get confused by the difference between technological and scientific progress. When science makes progress it unlocks new S-curves that progress at an incredible pace until you get into the diminishing returns region. People complain of slowing progress but it was always slow, you just didn’t notice that nothing new was happening during the exponential take off of the S-curve, just furious optimization.
The crypto mind cannot comprehend
Being Right and being Successful are not the same thing.
That's not super relevant in my mind. It's because they're showing fruit now that will allow research to move forward. And the success, as we know, draws a lot of eyeballs, dollars, resources.
If this path was going to hit a wall, we will hit it more quickly now. If another way is required to move forward, we are more likely to find it now.
Earlier quoted context omitted.
> That does not mean there isn't an AI system that can do X. You are holding the burden of proof here...
One of my supervisors used to say: "Don't tell me it's impossible, tell me _you_ could not do it." A true c_nt move that ends every discussion.
Earlier quoted context omitted.
What you're describing sounds like agentic tool usage. Have you kept up with the latest developments on that? it's already solved depending on how strict you define your criteria above
My understanding is that you need to provide and configure task-specific tools. You can’t combine the AI with just a general-purpose computer and have the AI figure out on its own how to make use of it to achieve with reliability and precision whatever task it is given. In other words, the current tool usage isn’t general-purpose in the way the LLM itself is, and also the LLM doesn’t reason about its own capabilities…