I am a product designer and literally every conversation I have now has everyone asking can we use an LLM here? They will be everywhere before long.
For example: buying drugs, getting fake passports, hiring hitmen, and selling drugs.
Never forget Jeremy Bentham proposed the Panopticon.
We understand things enough to know how to use them. Yes, this means the unknown parts can be used to scam people, but that’s true with anything.
The two necessary components seem to be novelty/hype and inscrutable inner workings.
Especially if it's technology about which I'm still somewhat sceptical.
LLMs hallucinate so you have to check everything they spit out. But their outputs sound convincing and people are lazy...
The great chance of LLMs is of course assistive technology, where human actors and LLMs collaborate to do tasks. I am afraid however that what will shape the impact of LLMs on humanity much more is a different thing: Throught history there was always a certain number of people a dictator had to be at good terms with in order to stay in power. My fear is, that this number will become smaller, because it will be much easier to give the realistic impression that you have the support.
Existing concepts of reality and truth will definitly be completely and utterly destroyed by LLMs, and even actual, real information will be tainted by the fact that it could be fake – we are already seeing today on a smaller scale what living in such a world feels like if we look how societies in a post-truth environment operate.
My prediction (and I'd love to be wrong on that) is that the negative use of LLMs will outweigh positive use significantly, because it favours use cases where you don't have to care about correctness.
<2050>
Did Donald Trump supporters dye their hair orange in support?
Yes they did... (news article supporting the idea, complete with photos and video (with audio))
I almost fear for any history that has been digitized.
History is written by the victors, and AI is looking like a victor at this point.
* It might be on the order of 1%, but it would be insane for a rational actor to ignore a 1% probability of self-destruction.
Always, if the next question is "how much?"
There's undeniably a lot of work being done in this area, but so much of it is just ill conceived shovelware in an attempt to make it big in the gold rush.
We get like a dozen "Show HN" posts a day where there's a service that basically prefixes an input with a prompt and bounces it off ChatGPT and returns the response.
There's also no doubt interesting products to be built using this new technology. But I don't think it's going to be made by just gluing 2-3 APIs together. If you can do that, someone else can do it too. That's not a business model. Being the first to do something means very little.
LLM to solve math problems, implement API requests, etc. = hard no
LLM to solve challenging constraint problems or statistical problems where we already have a good solution (A*, heuristics, really clever algorithms) = no
LLM to solve challenging problems where we don't have a good solution (medical diagnostics, legalese, manual translation of badly-formatted data) = yes... (people will disagree but if the LLM does these better than a human, it does these better than a human - and if not, the humans will almost definitely benefit from using an LLM)
LLM to do something with natural language, like implement a chatbot, explain something, cheap therapy = hard yes
The problem, however, is that your third point is where the vast majority of the potential usefulness of LLM lies, but simultaneously where integration with other systems is highly non-trivial, hence my "calm your proverbial tits" position on the issue.