Remix.run Logo
▲ ChicagoBoy11 a day ago

But to extender your analogy: We do understand how Claude does what it does and how it is built, so how is this different?

▲layer8 a day ago | parent | next [-]

I would dispute that we completely understand how LLMs do what they do. An analogy would be that we meet an alien race that gives us their proof of the Riemann hypothesis translated to Lean. Plus of all other provable mathematical conjectures we come up with. And we might kind-of understand how those aliens do math. It would still be a bit sad that we'd be relegated to consuming what the aliens give us, and never be able to ascend to the level of the alien mathematicians

▲ptoo a day ago | parent | next [-]

We didn't meet a hypothetical alien race though. We created it. We are giving birth to something that will soon surpass us in every conceivable way.

▲layer8 a day ago | parent [-]

That doesn’t make much of a difference. It’s alien, and for the most part we didn’t design it to do the things it turned out to be able to do by happenstance. We might as well have grown the aliens in a test tube by some unexpected gene manipulation.

▲CamperBob2 20 hours ago | parent [-]

Over half the cells in our bodies are not part of us. The lines we draw are seemingly never respected by nature.

▲Turneyboy 11 hours ago | parent | prev [-]

An analogy here is evolution. We understand what it does on a basic level. There is natural selection and species that are well adapted to their environment thrive.

This absolutely wouldn't allow us to predict what species will emerge and that something like humans would be the dominant species.

Similarly, we understand the mechanics of how LLMs work, but that does not make it predictible in any way.

▲Philpax a day ago | parent | prev | next [-]

We definitely don't understand how Claude does what it does! We know how to grow a Claude, but we don't understand what goes on inside a Claude.

▲ruszki a day ago | parent | prev | next [-]

I remember when about 15 years ago, I wanted to make my neural network home work based on proven math. I had to train a model to learn basic number recognition. The math said 100000 neuron is what's needed, so I typed in that, and my laptop immediately froze completely, and died a little later. I learned on that day that we know empirically that a few 10s are enough.

We have the same thing with LLMs. We don't know how they work. We don't understand them. We have only empirical, statistical indications, nothing is proven. We know how the small parts behave, we really understand those, and also the different algorithms. But they whys? No. We're getting better, but still far away.

▲hbs18 a day ago | parent [-]

If we don't know how LLMs work, how did we create them? They didn't just spawn suddenly on someone's hard drive.

▲lawandjustice 21 hours ago | parent | next [-]

What does 'know' mean? And what does 'mean' mean?

▲ruszki 20 hours ago | parent | prev [-]

Trial and error. A ton of trial and error, and a lot of guesses. More wrong than good guesses.

Basically every science field is based on this.

▲barbazoo a day ago | parent | prev | next [-]

And presumably mathematicians understand the output, the proof, whatever it is that came out of the process. I don't feel like humans have to be the one doing all the manual math and figuring out all combinations.

▲streetfighter64 a day ago | parent | prev [-]

I mean I understand how cars work but that doesn't mean I'd feel the same accomplishment of driving 42 km as running a marathon.