I'm getting seriously worried about my career in science due to the speed with which AI is advancing.Just 4 years ago it couldn't do elementary school math, now it's tackling millennium prize problems
>>17049447Why? It's gonna be of good use in science and people have to interpret/use it, people like you. I'd be more concerned as an accountant or some nonsense like "product manager". I'm sure AI that solves advanced problems and checks itself can be made to fill in excel sheets and check itself.
why would you be worriedAI might be an incredibly powerful tool but it's still a tool that you yourself can leverage to speed up your workflowand as much as the retard CEOs at the top wanna replace their workforce completely with AI, it always ends up backfiring on them
Is law safe? Worst case I can just do criminal/family law or some shit. That literally can't be done by ai
>>17049463there will be an inevitable day in court when someone decides to represent themselves pro se with permission to utilize an AI lawyer and actually win their case
My only cope remaining is that LLMs so far are excelling mostly at domains that are readily verifiable and there's too much humanslop in the literature of the physical sciences for it to make rapid progress without performing its own experiments first.
>>17049477it's more perfectly suited for areas where humans do not want to do actual grunt work, like calculate millions of different formulations, making tweaking a single value between each one
>>17049447Should have listened to Uncle Ted
>>17049477If it's possible to construct a digital rlhf training environment for a subject, the model can become superhuman on that subject.
>>17049463>Is law safe?No. This is a huge AI market involving customers that can afford expensive systems and numerous suppliers. And the results are impressive. 2 years ago we saw toys, last year it looked OK, this year it is impressive, and we have no idea what the future looks like. I cannot see we will hire any juniors for the foreseeable future, the old pyramid structure will turn into an obelisk.t.Patent attorney and physicist
>>17049447
>>17049463>Is law safe?Law is probably one of the least safe professions.
>>17049463Entry level law work is getting absolutely annihilated.>>17049617
>>17049574>t.Patent attorney and physicistI've seen this one before..
>>17049817There are at least two of us, plus at least one Examiner, probably from USPTO. And Physics is a common background among my colleagues.Examiners used to have great job security but that is changing too. While patent offices now use a lot of AI, they are also flooded with AI generated applications from applicants filing without using a patent attorney. So there have been a lot of recruitments in much of the Western world, possibly elsewhere too in order to process the increased volume. Once that transient backlog comes under control, they too might face redundancies.
>>17049499stop using the internet then, rofl
It honestly does feel like something big is about to happen. I thought AI would run out of steam by now but it's just ramping up
>>17050431You ain't seen nothing yet.
>>17050521I wanted to go to university for physics or engineering but it seems like a waste of time now. If it can solve millennium problems right now than what white collar work will even exist for the average person in four years from now? Maybe the top .1% will still be needed but everyone else seems doomed. Is tradework even an answer when contract work also slows down with the economy at large and everyone will be flooding into it after losing their jobs? If everyone is losing their jobs no matter what then shouldn't you study math and physics anyway because at least you won't be retarded?
>>17049574But the country needs to legally permit the use of ai in justice. If it's not accepted then it can't be used
>>17049447Your career in science was probably fucked anyway. However le AI stole those "solutions". Not that they are actual solutions. Both the Navier-Stokes and the Riemann things are mere progress that doesn't have much to do with the really hard part of the problem.
>>17049447Quite the opposite really. Think about the meta, knowledge is in itself a mind dependent property of information. Symbolic jargon become information when a mind indexes it to knowledge space. Einstein describes it best in, “Das Raum-, Ather- und Feldproblem der Physik”: >Concepts can acquire content only when they are connected with sensory experiences. No logical investigation can uncover this connection; it can only be experienced. And yet it is precisely this connection which determines the knowledge-value of the systems of concepts.>An archaeologist from a later culture finds a textbook on Euclidean geometry withoutdiagrams. He will find out how the words “point”, “line” and “plane” are used in the theorems. Hewill also discern how these are derived from each other. He will even be able to put forward newtheorems himself according to the discerned rules. But the formation of theorems will be an empty word-game for him so long as he cannot “have something in mind” for a point, line, plane, etc. Until this is the case, geometry will hold no actual content for him.>What does this “having something in mind for a line, point, intersection etc.” mean? It meansindicating the sensory experience content to which those words relate. This extralogical problem constitutes the essence-problem which the archaeologist will be able to solve only intuitively. Only in this sense can the question of the essence of an abstractly described entity be meaningfully posed in a reasonable way.We will always need a human second author to index math papers into our collective mind's knowledge space. Since the sheer output of math result is about to go up, we will need more human mathematicians around to keep up to date with cutting edge research. A human is needed more now than ever. This is the best time to be a mathematician.
>>17049447If it gets to that point society is just completely fucked and you've got bigger problems.
>>17050521Why would you take the word of someone with an incentive to lie?
>>17049463>literally the best possible use of LLM's>some dumbfuck on /sci/ thinks it's safe from themnever change you utter fucking retards kek
>>17049463>Is law safe?No, I'm a lawyer and I'm on suicide watch.You see, what makes an independent lawyer money is random people coming in to see you to handle their case because you know how to address the court, what forms to submit and even write a non retarded document describing their issue. Most people can't do that. A month ago Gemini wrote a big-ass appeal for me that normally would take me hours or days to compile in 15 minutes. Appeals are probably the hardest thing I usually do. I only had to do minor checks and corrections that took me another 15 minutes. It was then that reality finally broke through to me and I can no longer cope - I am fucked if the computer can do that.
>>17050907Oh and all those people coming in to see you because they have no idea how to handle their case and what document to submit to the court? LLM's will them exactly what to do and will write a better document than 90% of lawyers out there would
>>17050907>No, I'm a lawyer and I'm on suicide watch.Where every lawyer should be
>>17049463No field is safe. Some are less in danger than others, but everything will suffer in some way (except "owner of AI company").The safest thing is less a particular field, but getting into it sooner before automation catches up. So this means don't do a university degree right now. Because even for just a 2 year degree, AI will come a huge way by the time you finish and the field you're in may no longer be hiring.That said, if you are going to do some degree, some safer ones are the ones based in physical work. Veterinary science is a good one, even if they do invent robot vets the animals would get spooked by them.Otherwise a job that will always be legally required to be a human is okay. I have no doubt that AI will easily surpass human teachers if it hasn't already, but I can see a lot of pushback on the idea of having a classroom full of kids just talking to their individual TeachBots, from teachers obviously, but also parents and the general public. Even if it is legal, schools and universities may want to seem more prestigious by having quality human teachers. Self-driving cars have existed for ages but drivers are still required to be human basically everywhere. Judges will all be required to be human, and prosecutors/PDs/whatever. But a huge amount of the lawyers who got laid off from being a desk monkey will be competing for those positions.Other things that customers will just prefer a human for are a good one too. For example tour guides. Who knows how much tourism will crash when no one has a job anymore though.
Overvalued
>>17050907>A month ago Gemini wrote a big-ass appeal for me that normally would take me hours or days to compile in 15 minutes. Appeals are probably the hardest thing I usually do. I only had to do minor checks and corrections that took me another 15 minutes. It was then that reality finally broke through to me and I can no longer cope - I am fucked if the computer can do that.I thought LLM's still hallucinated too much to do that, or am i a year behind?
>>17050562The "use of AI" is wide in scope. A large part in legal proceedings is to find all relevant laws and case law and more, and that is already well established, even by legal organisations:https://www.epo.org/en/news-events/news/myepo-services-launch-groundbreaking-ai-powered-legal-search-tool>>17050886You have no idea how bad it already is. And it is getting worse.>>17050907And I have seen an AI draft a complete patent application with about 10 hours of human preparateion and follow up, a task that would normally take at least one week. I went through the entire application and could only find a handful mistakes that were utterliy insignificant. I am still in shock.What remains safe still are oral proceedings, making new legal theory (super hard) and interaction with the client to get the full picture of the invention (harder than many expect). There are also some new services coming up that we have great hope for bringing in new types of well paid work.>>17050928As long as you treat a LLM as a junior, it is in my experience safe. Ask for opinions and you tempt fate. LLM is very good for interpolation between known poimnts of facts but once you extrapolate you are going to see LDS grade hallucinations.
>>17050887Because for the past 4 years these people have been consistently right while people like you who think you see through the ruse have been consistently wrong
>>17050535Yea but I don't like spending 200 grand for four years just on a degree
>>17049447Be nonchalant & carefree when dealing with life. If Tech overlords win, it's gg for 99 percent of humanity. If they don't, your skills will stay useful. Either way, life is meaningless and existence is fake & gay
>>17051077You also have an incentive to lie (not admitting you got grifted).
>>17051077>Because for the past 4 years these people have been consistently rightYou're psychotically ill and everyone outside your padded cell knows they've been consistently and intentionally "wrong" about nearly every single thing they claimed.
AI is slowing down rapidly and it's unlikely mathematics will see further automation, let alone any physical sciences (especially biology and medicine).
>>17051900And this you base on exactly what? A forlorn hope for gainful employment?
>>17050928>I thought LLM's still hallucinated too much to do that, or am i a year behind?You're like six months behind. Most people have no idea just how fast LLMs are improving.
>>17052210My faith in Jesus Christ
