How is it solved? LLMs cannot think new things, they can only cobble something together if it's in their training set.

tock 3 days ago | parent | next [-]

New things are made by cobbling together existing things.

bitwize 3 days ago | parent | prev | next [-]

They can think and reason better than most humans. Most problems they're pointed at are not in their training set, but in certain ways they resemble things that are—maybe there are a few different resemblances to different problems in its training set—so it's able to pull these disparate similarities together and apply the patterns it finds to come up with a solution. Much like human brains do.

▲

brookst 3 days ago | parent | prev | next [-]

What? This is a massive misunderstanding. It’s easy to get truly novel ideas from LLMs, unless your definition of “new things” is so strict that no human can do so either.

The training set is about patterns, not facts or specific configurations. Yes, it’s possible to extract (some) of the training set verbatim, but that doesn’t mean it’s all you can do.

	▲	NoMoreNicksLeft 3 days ago \| parent [-]
		>unless your definition of “new things” is so strict that no human can do so either. Humans rarely think of new things. We're a weak hivemind species. One or two individuals figure something out, and the rest of the troop of monkeys imitates. Brains are too fuel hungry for every brain to be innovating, "innovate and copy to the other brains" is the norm.

▲

nashashmi 3 days ago | parent | prev | next [-]

I am not sure if claude had powerpc scripts in its training.

▲

varjag 3 days ago | parent | prev | next [-]

That "only" part used to be the hardest. Getting the ideas was never the hard part. I think someone here even wrote an essay on that.

▲

qsera 3 days ago | parent | prev | next [-]

It does not think at all. It vibes based on its training and any additional bolted on constraints. It is a quite simple automation that only works by huge amount of existing data.

Modern man has grown quite dumb. He only seems to be able to "invent" by massive scaling things that are decades or centuries old..

▲

brookst 3 days ago | parent | next [-]

Out of curiousity, can you share a human invention that is not merely scaling things that were decades or centuries old at the time?

Fire? The wheel? Archimedes screw, maaaaybe?

▲

qsera 3 days ago | parent [-]

Electricity, RF communication, LASERS, Transistors....

▲

brookst 3 days ago | parent [-]

What specific inventions in those areas?

Electricity runs from simple batteries (600 BCE) to today’s power grids.

RF was predicted but not demonstrated by Maxwell in the 1860’s. His work built on Faraday’s (1840’s) and Coulomb’s (1780’s). Coulomb built on Franklin and Newton, among others. Or do you mean Marconi and Tesla, who merely implemented what Maxwell predicted?

The same is true for lasers and transistors but it’s tedious. There was no single “back in the day people invented things from whole cloth” moment.

▲

antonvs 3 days ago | parent | next [-]

I would put it differently. Those inventions came from humans interacting with the physical world.

When LLMs were first introduced, they didn't have much of a feedback loop. They wrote code, but they couldn't compile it. Not surprisingly, the code had bugs.

Now, they run with harnesses that allow them to compile the code, and react to the issues they observe. They can fix their own bugs and solve problems that they create, just like humans.

Give an agent access to the physical world, and it seems highly likely that they will be able to "invent" things based on feedback they receive while working towards goals.

Of course, there are some well-known limitations of LLMs, one of the biggest being that they're pretrained. So there may be some things where they're not as good. Just like how some humans aren't as good at certain tasks, depending on their genetics and/or how they've been trained.

	▲	qsera 2 days ago \| parent [-]
		> it seems highly likely that they will be able to "invent" things based on feedback they receive while working towards goals... I don't think so. Imagine a model trained on data from an Internet that believes in hypothesis that earth is the center of the world. Even if you feed all the physical data, I don't think it can come up with the idea that all of its training data was wrong. This might be also a good argument for why this LLMs are not "intelligent". You can feed contradicting training data all day and it will accept it without bating an eye. But that won't work with an entity that is truly intelligent.

▲

skydhash 3 days ago | parent | prev | next [-]

Those are not merely scaling. I can get “build upon other works”, but there’s a lot of scientific insights needed for observing and modeling a phenomena. It may even requires a boost of creativity to theorize an effect based on that model and then make it possible in an experiment.

▲

qsera 3 days ago | parent | prev [-]

Those are the specific inventions.

▲

addaon 3 days ago | parent | next [-]

> Those are the specific inventions.

In what way is electricity an invention? Electricity is a physical phenomenon. Various machines for doing work with electrical energy, storing electrical energy, converting other types of energy to electrical energy, etc. are certainly inventions... Heck, rubbing an amber rod with a fur is an invention. The static charge transferred is not.

▲

qsera 2 days ago | parent [-]

Invention, discovery..does not matter. There was a point in time when humanity was oblivious to the phenomenon and then there was a point when we were not and we could generate and use it.

▲

brookst 2 days ago | parent [-]

This is so false. Try doing the research to back up your risible claim.

Tell me the date when humanity went from oblivious to electricity to generating and using it.

▲

qsera 2 days ago | parent [-]

What exactly is false?

▲

brookst a day ago | parent [-]

It is false to claim that there was a moment when electricity went from unknown and opaque to understood and producible. That “moment” was thousands of years and all incremental.

	▲	qsera 7 hours ago \| parent [-]
		I didn't make no such claims. I claimed that these "incremental" progress was not about simply doing more of the same stuff aka scaling.

▲

brookst 2 days ago | parent | prev [-]

Great. So when you say electricity, do you mean the batteries of 600BCE and the precursors before them?

▲

pixl97 3 days ago | parent | prev [-]

>Modern man has grown quite dumb

Ah yes, that's why it only took 50 years instead of 100,000 years for homosapients to reach the space age....

Dude, there was no glorious past, we've always sucked.

▲

qsera 2 days ago | parent [-]

We have always sucked, but there used to be individual brilliance. Now everyone is the same shade of grey...

▲

pixl97 2 days ago | parent [-]

Individual brilliance has always been rare. If I booted your ass back 1000 years you'd have seen the same shade of grey in the pig farmers and peasants along with most of the nobles and kings. It turns out that in general the world is a shade of grey.

Hell, looking back at the brilliant individuals, they had many, many grey areas.

	▲	qsera a day ago \| parent [-]
		>Hell, looking back at the brilliant individuals, they had many, many grey areas. That is exactly the point. Back then that could exist. Now, we would just cancel them for one of the grey areas.

▲

antonvs 3 days ago | parent | prev [-]

I take it that your training cutoff was early 2023.

▲

rschiavone 3 days ago | parent [-]

the trope that everyone else you are talking to on the Internet is a bot is getting tiring real fast

	▲	antonvs 3 days ago \| parent [-]
		I wasn't saying you were a bot. I was pointing out two things: first, your understanding of LLM capabilities is very outdated; and second, that in this respect, you're behaving much like an LLM with a training cutoff. That further touches on the idea that the differences between you and an LLM may not be as large as you imagine. In particular, "cobbling something together if it's in their training set" is pretty much what all humans do.