site banner
ZorbaTHut  2yr ago (text post)   4788 thread views

Small-Scale Question Sunday for September 04, 2022

Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?

This is your opportunity to ask questions. No question too simple or too silly.

Culture war topics are accepted, and proposals for a better intro post are appreciated.

26
Top Bottom Old Controversial Comments
Jump in the discussion.

No email address required.

Now that Stable Diffusion has been public for a week - what will be the next field to be revolutionized by AI?

(And if your answer is "writing" or "music", I'd like to hear what field you think will be next after those. Those are obvious candidates because AI systems are already in use in those fields and/or will be shortly, but due to structural differences between those fields and the visual arts, I'm skeptical that AI will have the same seismic impact there that it's currently having in art.)

Finance, if it hasn't already behind the scenes.

There's a lot of intermediaries who currently get paid pretty handsomely for a job that is, at core, just channeling money from one account to another and explaining what they did and why to a human. And 'money' just means a digital entry on a ledger for most purposes, now.

I see no reason why an 'investment/financial advisor' can't be completely replaced by a bot that listens to the customer's situation and goals, and based on its learning from a dataset of billions of similar situations, spits out recommendations for how to invest or otherwise distribute one's money to achieve that goal.

Same for stock brokers. Same for financial analysts. Same for Tax advisors, even (see my point about law, below).

Factors vitiating against this: Regulations and distrust of AIs to handle one's money.

I know that banks and credit card companies are already using AI to detect fraud and handle customer service. The question is when they'll allow/be allowed to give the AI the ability to access customer accounts directly.

Also: Law. At least the transactional side. There are already HUGE databases of highly structured information about every single topic that is relevant to the practice of law, and legal writing is, by it's nature, very predictable and rigidly formal such that any AI should easily be able to produce human-passing work that can match all but the most learned and innovative jurists for quality.

I have to assume we are mere months away from some company announcing that they've trained an AI to draft and analyze contracts and similar legal documents, AND to draft motions complete with legal citations based on a description of the desired motion and outcome.

This kills legal assistant and paralegal jobs instantly. It also carves a big gaping whole out of available attorney jobs.

Acting. Once we have a Stable Diffusion for movies (Any guesses on when this happens?), few will hire actors. We will likely have an explosion in content as anyone who can write can make a film, and lots of book get quickly made into movies. I can't wait for the Worm movie.

Scams. Imagine an AI that calls your grandmother claiming to need money, sounding exactly like you, using voice recognition and GPT-N (fine-tuned on previous successful scam calls and prompted by a selection of your own social media information) to reply.

It'd work just as well on non-English speakers too, so nations that have up to now been more or less immune to Indian/Nigerian scammers due to the language barrier will now get targeted just as easily — and they don't have any sort of resistance from being exposed to the current "weak" versions of the scams either.

For a few months many of my neighbors repeatedly got scam phone calls with my name and number on the Caller ID. I supposed in a few years these calls will also have my voice.

Speaking of phone scams, does anyone know of a way to block an entire area code minus a whitelist? I get so many spam calls (or group texts; those are the worst because of the other victims' subsequent "remove me from this list" mass-replies) purporting to come from my home region, where everyone who'd have a reason to contact me is already in my contacts, and blocking every other available number on those sequential group-texts one by one gets tiring. (Though I expect I've taken an appreciable bite out of all the nearest few thousand numbers already, anyway...)

Unfortunately I think sales will be sooner than most people think. The AI to automatically get rid of your accent in real time on a phone call already exists - from there you only need to be able to accurately feed phrases/words to the AI and it will be a better salesperson than most could hope to be. An AI also doesn't get discouraged, which from what I have seen is the #1 reason people leave sales.

I think music will be differentiated into different groups. People love concerts, and being able to physically play an instrument has always been, to my mind at least, much more impressive than being an artist. It's 'cooler' if you will, to play guitar than to wield a paintbrush.

Writing is, well, the writing is already on the wall. GPT-3 is incredible, and scaling shows no signs of stopping anytime soon. (1) I'm extremely confident that many lower-quality blogs are now entirely GPT-3, and instead of hiring ghostwriters, which is a shockingly common practice, writers will just use GPT-4 or 5 to offload their work. I'm neutral on whether or not this is a good thing.

Ironically I now think taxi drivers, who I once thought would be the first to go, will probably last pretty long. Mistakes are far costlier when driving than when creating a piece of art, or a piece of writing. One of the ways I view different jobs is how serious the immediate consequences are when you fail.

1 - https://www.gwern.net/Scaling-hypothesis

I think sales will be offloaded on the lowest end - if you do direct marketing, yes. Phone calls, maybe. The interpersonal relationships people still work on in business-to-business sales I don't see being co-opted by AI at any point before AI co-opts all human roles. The role of 'maintain a relationship (partially in realspace) with a human being in order to understand them and sell them things' can certainly be facilitated by AI, but it seems implausible that role can be turned over to machines before we build something fully general and capable of replacing human beings entirely.

I expect less new fields getting revolutionized and more currently AI friendly fields getting huge upgrades.

Imagine call center AI that actually sounds fully human or AI assistants that sound fully human.

However, sticking to the spirit of your question, my answer would be mathematics. With Mathematicians only existing to reevaluate the solutions given by AI to confirm they would work. Mathematics would become far more an engineering field than a person coming up with a solution on their own field.

I think this is misguided because it's the opposite of what currently happen, with theorem provers doing so much of the evaluation work relative to the 'creative' work. I can definitely see AI expanding the search space, though, with mathematicians working with the machine to find more novel or interesting results as a consequence. Much like art, I think AI are at the present time both a job-destroyer for the bottom end of the market (want your fursona fucking a famous politician? No longer do you have to pay $50, you can just get the machine to do it!) such as commissions but will ultimately enable people who understand art (colour, composition, etc, and consequently how to more reliably get the machine to do the work you want) to create more interesting and varied things at the top end.

Academic mathematicians are towards the top end of what you'd consider 'stem jobs' IQ-wise, so I'd anticipate a similar effect there.

Theorem provers do all of the evaluation work ... for those specific results which have been painstakingly translated into the theorem prover's language, which to a first order approximation is zero percent of the new and interesting results.

Training a transformer AI on MetaMath (or on Coq results, whatever large database has both complete proofs and adequately verbose comments), combined with the verifier itself, might be enough at this point to create a "math-paper to formal-proof" translator. Skimming through comments I see a lot of links back to the papers which originally published each theorem, which certainly ought to qualify as "adequately verbose" even if the database comments themselves are fairly terse.

Doing creative work would of course be more interesting ... if we could only define what's "interesting". 378+135=513 is a theorem among infinite others, but nobody cares about it. We tend to like math if it eventually has endpoints with real-world applications, and it's a bit hard to put that into an evaluable loss function. We also tend to like theorems if they're more general, and if they're short to state but long to prove, and if they're on the shortest path to proving other theorems, and maybe there's something to those criteria that could be quantified well enough to point a Neural Net Monte Carlo Tree Search in the right direction?

AI is more likely to replace technical jobs than manual jobs though. It's easier to program an AI to figure out a trillion parameters of data then to teach it to walk. Welcome to the upside down where the safest job will be factory worker.

Said tongue in cheek.

To a degree, yes. People who have a high-level understanding of their field, however, are those best placed to use new AI tools. Likewise, statisticians didn't disappear because we built better tools for statistical analysis, rather the demand for statistical education has never been higher. The tools are still used by someone and we tend to see the lowest rung automated away and smaller numbers of usually better educated employees getting productivity increases. Usually what this looks like is a lot of the lowest-skill (or those with a very narrow skillset) employees lose their jobs - the invention of the mechanical (and later electronic) calculator removed the need for human calculators, but engineers and mathematicians are still a thing.

Architecture. Since you can run the output through a layer that will check that the structure will not collapse, you will be able to ask the AI to design whatever you want and it will spit out a viable blueprint. Want a tower on your McMansion? Just use cad+img2cad to add it to the existing structure.

I'm here for procedurally-generated, tailored buildings.

Ironically, perhaps, I expect you'll still need human contractors and workers involved to do most of the actual construction.

What is this place?

This website is a place for people who want to move past shady thinking and test their ideas in a court of people who don't all share the same biases. Our goal is to optimize for light, not heat; this is a group effort, and all commentators are asked to do their part.

The weekly Culture War threads host the most controversial topics and are the most visible aspect of The Motte. However, many other topics are appropriate here. We encourage people to post anything related to science, politics, or philosophy; if in doubt, post!

Check out The Vault for an archive of old quality posts. You are encouraged to crosspost these elsewhere.


Why are you called The Motte?

A motte is a stone keep on a raised earthwork common in early medieval fortifications. More pertinently, it's an element in a rhetorical move called a "Motte-and-Bailey", originally identified by philosopher Nicholas Shackel. It describes the tendency in discourse for people to move from a controversial but high value claim to a defensible but less exciting one upon any resistance to the former. He likens this to the medieval fortification, where a desirable land (the bailey) is abandoned when in danger for the more easily defended motte. In Shackel's words, "The Motte represents the defensible but undesired propositions to which one retreats when hard pressed."

On The Motte, always attempt to remain inside your defensible territory, even if you are not being pressed.


New post guidelines

If you're posting something that isn't related to the culture war, we encourage you to post a thread for it. A submission statement is highly appreciated, but isn't necessary for text posts or links to largely-text posts such as blogs or news articles; if we're unsure of the value of your post, we might remove it until you add a submission statement. A submission statement is required for non-text sources (videos, podcasts, images).

Culture war posts go in the culture war thread; all links must either include a submission statement or significant commentary. Bare links without those will be removed.

If in doubt, please post it!


Rules


Recommended Posts And Communities

Recommended Realtime Chats