singularity2001 a day ago

they're still bounded by training data and how it relates to the real world.

Yes and no. They can extrapolate and build upon the training data, as was the case with the last dozens of math proofs

buzzin_ 20 hours ago | parent | next [-]

Of course they can't just talk to themselves and improve the knowledge of the world. They are just stochastics parrots that give you an average answer.

LLMs giving you something novel would be like if you let a model play chess against itself and become the best player in the world this way. Totally impossible.

/sacrasm

▲

HNisCIS a day ago | parent | prev [-]

Yes, but the point being for the LLM to be useful to us it has to do something relevant for us and we have to define what is relevant. Even if you view them as fully human level or beyond (in terms of agency) there's still some purpose that we have to help provide them with.

To put it differently, if you have some idealized model in front of you that can do anything a team of humans can do, what do you say to it? It's still just a model that takes an input and provides an output.

▲

rlt a day ago | parent [-]

I don't understand. You will ask it for something you want, and instead of needing to pay a the team of people with expertise, it will just do it for you.

Someone will still need to ask for something (or maybe it will just run autonomously making things it thinks we'll like) but it won't require any expertise in the field.

▲

lanstin 16 hours ago | parent [-]

I can’t help but believe that reality will keep the interactions more useful when coming from a person with expertise in the field, e At least for technical questions.

For things like sickness, old age and death, generally the lacks are not knowledge but the consistent will of the group. Like global warming, there is no real profound knowledge gap except how to get everyone to agree to mitigating the carbon burn. Perhaps it is likely LLMs would be able to persuade the group to stop burning the carbon before it’s [more] uncomfortable, but the small group of people committed to accelerating the carbon burn can also pay for LLM persuasiveness.

▲

ecshafer 12 hours ago | parent [-]

Global Warming does have a knowledge gap of sorts. Yes there is a viable solution to Global Warming, drastically cut back carbon and methane emissions. Where the knowledge gap is: drastically cut back on carbon and methane emissions, whilst maintaining or even improving quality of life in developed countries, having no economic issues, maintain industry, and keep developing countries developing. That is the hard part.

	▲	lanstin 11 hours ago \| parent [-]
		There might be a few industries where a technical fix is needed, but having to replace all our infrastructure is always a huge decades long win for the economy. Solar and batteries are made from abundant materials and cheaper than anything else, and almost all the sticky technical issues have good solutions with large amounts of cheap energy. Fuck we could desalinate and pump sea water back onto Greenland and Antarctic but instead we are arguing that windmills cause cancer or whatever the fuck the oil companies are paying morons to be persuaded by. Before the morons took over in the US, banks and insurance companies were salivating to fund electrification and selling 5 billion electric vehicles. The leadership there apparently has the moral fiber of a petty crook instead of a vanquishing hero, so they are ditching a lot of that, but none of these are prima facie cases where GenAI / LLM will help, unless you think "everyone having a personal physics tutor" will transform the world the way my friends and I thought "everyone being able to chat with a physicist" would transform the world 35 years ago.