Ask HN: How to deal with the AI-pocalypse as a regular guy?

LLMs are basically attention machines. Just like the steam engine did physical work, LLMs do mental work.

You can't get a steam engine to dig a hole though. It was used to pump water. Then carry things. Then locomotion. They were limited to rails. They did take jobs. Then other kinds of engines appeared. They did things steam engines didn't do so well.

But turning an engine to a bulldozer isn't easy and that created a lot of new jobs. Right now LLMs are doing the equivalent of pumping water. Heavy, boring, tedious things that humans don't really want to do. I hated doing data entry and cleaning and LLMs are great at that.

One thing many creatives won't admit is that a lot of their work is generic and repetitive, simply requiring lots of attention to get good output. Screenwriters were one of the first people I've observed utilizing LLMs, because they understood this. Someone hacked Copilot to start writing scripts. There's probably a lot of things it can do, perhaps in animation and logo design and stuff. And there's a lot it can't do. But they're shocked because it's doing the "soulful" stuff that people thought made them human, when it's actually just the advantage of having a huge training database of tropes.

Will art go obsolete? The first knowledge worker to lose his job to AI was Kasparov, and yet human chess is still more popular than ever.

Working freelance in web development is the opposite of job security, so I can understand why you feel concern about the impact of AI on your industry.

However, reading this, I do not get the impression that your background gives any kind of unique insight others have missed, more that you’ve just read a lot of opinions online.

Online isn't reality. Online, people don't talk about corporate politics and inertia and the million tiny factors that determine which jobs get "replaced" and which don't. I use quotation marks because most people mid-career or higher can see changes coming and either shift out into another role or go into management or solution architect roles where there is a need for human oversight (and tend to be roles that offer more money).

AI is an existential risk only if you choose to do the same job forever.

I have tried a variety of AI tools to simply Transcribe or Translate French to English from audio and video tracks.

One of my favourite French TV series with 35 episodes and there are no English subtitles for it.

With all the hype around AI I thought it would be a simple task using Open whisper, Open whisper faster and Vibe.

I Downloaded and tried all and every model, still no success.

I have extracted the audio with ffmpeg in different formats and changed the video formats but nothing helps

Absolutely useless

At my workplace they recently enabled Microsoft's Copilot which is basically OpenAI + RAG.

It's a joke. It is painful to watch how slow it types. About 30% of the time it doesn't work at all. I asked it how big a particle accelerator would have to be to have 1 PeV protons and it got the right equation for the radius of the ring but it told me the ring would be like 10^34 meters in radius which is about 10^24 times too many because it got the units totally wrong.

It is completely unable to help a person get 5kg of antimatter, make a nuclear weapon or a deadly virus that will kill everybody because... it can't. People are afraid it will try so it has been taught to claim that it won't do these things.

It is also completely unable to put a list of items (say 20 or more) in a specified order. For instance it falls down on "list isotopes that decay by emitting a positron in order of half-life" or "list US states in order of how many characters in its name", etc. It can't do it any more than it can help you get 5kg of antimatter, but it lies and says that it can, will attempt and get it wrong, apologize when you point this out, try again, get it wrong again, endlessly. If you like pushing bubbles around under a rug, gen A.I. is really for you.

Now yes, LLMs can solve difficult problems in information extraction (say relationship extraction) that we were lost at sea with 5 years ago. But intelligence, truth and such are all difficult philosophical concepts.

People who know better act insulted when I remind them that neural networks don't repeal the laws of computer science (e.g. Godel, Turing and all that.) They can't make P=NP, they can't solve the very serious problems of logic + arithmetic that Godel warns about, but they are very good at telling you something that will bypass your defenses (I think this comes out of picking the "most likely" word because it won't surprise you) and that you want to hear (what else is RLHF?)

One real danger of LLMs is that they are better at seduction than we are because they don't have a self to manifest itself and get in the way. I am very worried about LLMs taking over dating sites, being used to run "pig butchering" scams and similar sorts of evil.

See

https://www.amazon.com/G%C3%B6del-Escher-Bach-Eternal-Golden...

to get some idea of how LLMs could appear to be 95% of the way there but still be structurally wrong to solve real-life problems that involve logic, arithmetic and everything else.

Can''t you just replace every mention of AI with outsourcing in your post?

Literally everyone was freaking out in the 2000's about their job because of India.

Well, all the promoters promote these because most of them don’t understand it or has money incentives. These models consume HUGE energy to run and a lot of knowledge to coerce an useful result. You as a developer has the knowledge to ask these questions, but if you look at the people who are looking to cut costs(management), most of these people are not adequately knowledgeable at all. Look at any decent sized corp or even startups, most management have difficulty correctly fleshing out a proper and clear requirement. A product that is successful is like a garden, it needs tending and I don’t see these manager types having the knowledge to tend, else we would be all out of work several decades back. The thing is, everyone can use a hammer, but only the experienced ones can produce a crafted creative good. Everyone can probably pain a banana but only the real artists well practiced can produce an aesthetic portrait of a banana.

> advances of AI in the past years can imagine a not-too-distant future (likely less than 10 years) where literally everything we are doing currently will be obsolete

Based on that you really understand what is current AI is all about no more than a bbc journalist.

So don’t worry about your worries, there are far more elephants in the room than this one.

The cool thing about AI is that its all or nothing.

What we have right now is not really AI, its not really intelligence in the sense that the models can't reason. All the models out there are forward pass (and yes, even recurrent ones, which can essentially be unrolled). There is nothing on the horizon that indicates that we are even close to getting something that can reason, or are headed in the right direction.

The biggest benefit of current ML stuff is translation. Everything that ML does now is translation, and it will apply to dev more and more over time.

For example, Python and Node dominate the programming world because its way easier to translate ideas into code through abstraction layers. In the same way, the future is going to still require developers, but these developers are going to be coding in something that looks like python. You will be able to type plain english instructions on command line intermingled with traditional code, which will save you a lot of time because you won't have to learn specific syntax of some framework. The language will then be translated to some IR (most likely JS bytecode), which will then be translated by a specially trained model into efficient machine language instructions for a specific architecture.

There will probably be even a come back for FPGAs, because you there will probably be models that optimize architecture for your code, and there would be some use cases where processing at slower speed but in parallel with specific "virtualized hardware" configuration for optimality is actually faster.

And all that will mean is that dev will move faster. You will be able to write code faster, test the code and simulate production usage easier, and generally make things work in way shorter time with way less people. The thing thats preventing a lot of stuff being made now isn't really lack of ideas, its lack of funding.

Now, as for actual AI, i.e intelligence that can reason, if we somehow manage to build it, the cool thing about it is that you will be able to simply ask it a question "Im a regular guy, what is the best way for me to optimize my life that guarantees my current standard of living?", and it will answer this for you.

I'm in my 40s, main job is parenting & homemaking, previously taught high school physics, math, and other science, also did some engineering/materials science, and I also do part-time work for a local public utility. I like it being feasible to walk to work again.

As we layer on the abstractions I feel further and further removed from the land that sustains us. It's an existential threat. I don't knowingly choose to use any products of LLM/ML, and I've scaled back my computer use to: reading some blogs (low tech magazine & no tech magazine, plus sites linked from there); access to library ebooks; medical access for family; credit-unioning; watching a streamed show with my spouse for date night or something on PBS with the family (otherwise I don't bother with streaming, and I can imagine a time again when I don't even read, but a lot has to heal/reassemble for that to be tenable); and keeping up distance friendships via videogames (BG3 and Solium Infernum are all I play these days, and then almost never solo, after decades of "self-medicating" after a rough childhood).

I read mostly nonfiction now, towards being a knowledgeable member of the rural community I'm in, inspired in part by Jessica Carew Kraft's book Why We Need to be Wild. We're in this together, and the best place to start is with those within walking distance.

I want to second two other responses: first, the one about outsourcing to India. This was a very real threat, and practically no one wanted to get a CS degree for a while. But we learned that the hard part about writing software is not hacking out code, it's communicating with the people that are driving requirements. Being in the same culture and time zone turned out to be more important than being cheap. This part of the job is still hugely important and will continue to be until AI is good enough to read the minds of business people that can't properly communicate what they need.

The other is connection. I think long term this is all we have -- people will prefer people because they are people, even if that is not as good as AI. Artists won't get paid for putting paint on paper anymore, for the most part, they will make money because they are human and can build a human connection with their clients. Not everyone will like this -- artists who may be brilliant with a paintbrush but totally lack people skills will probably wash out. I think for the rest of us, things will go similarly but later. Build and reinforce your human relationships.

Finally, breathe a little. We have some time, probably. The world won't change overnight. Stay on your toes, keep your eyes open. There will be opportunities as well as risks.

The simple development tasks, like building a basic CRUD website will get easier. Companies will want something more sophisticated. Development of anything sophisticated by AI will take much longer. And, by then, everyone will want something new and more complex once again. All of this has happened before, and all of it will happen again.

I work in an industry already being shaped by LLMs. Today, it's eating away at the bottom of the market, but I assume it will eventually eat away at the top where I am positioned too.

I was talking to a friend about the book "psychology of money" and the part he shared was related to paying down your mortgage, even if it makes more financial sense to keep your money invested in the market with a higher expected return than the interest rate on your home loan.

Being disrupted by an LLM has created a ton of anxiety for me, and this conversion led to investing almost my entire NW into Nvidia.

My thinking was, if I am disrupted by AI, Nvidia will at the same time likely grow in value.

The amount invested is still only a fraction of my total expected earnings over the next 10 - 20 years if AI did not exist, but it has calmed me quite a bit about my impending unemployment.

This is a case where The Bitter Lesson [0] should make you feel better about the future. We are already butting up against the limitations of the current model and veering into fine-tuning and incremental improvement of what we've got (RAG, etc). The next step change will indeed leverage an order of magnitude more compute, but what we have now isn't "it" (AGI).

[0] http://www.incompleteideas.net/IncIdeas/BitterLesson.html

Build relationships.. demand for certain jobs waxes and wanes, but people will always be around.

The more connections you make and the more people respect you, the more likely you are to get a job when you need one.

Don't worry and live for the now.

It's not just you in the boat and whatever happens you can follow the crowd when/if it happens.

I personally believe it'll just be a new set of tools to reskill in, much like a calculator to an abacus.

Sure it would help to know how to build that calculator, but just changing the batteries and typing sums will be fine, and you can continue doing the theoretical physics equations the calculator helps with but cannot do itself.

That is with linear growth. Or logarithmic, depending on what you look at.

No one knows for sure, but some point out that transformers and satellite tech might hit a quality plateau after this short explosion enabled by nvidia. It doesn’t live neither up to hype nor up to initial speed in this year yet. We’ll see in the next generation.

It is certain that over the next 10 years, AI is going to do more than 50% of the tasks we are currently doing everyday as a front or backend devs. This will in turn reduce the need of devs too. But, this can't necessarily be a downside. AI has open up a big window of possibilities, not just for large companies but also for devs like us. I can literally think of 10 small AI products I could build quick after learnig some ML, exploring the vast amount of OSS repos and levaraging openai (potentially). You spent time as a chef and now are a web dev. I don't think that would have been as easy as it sounds like. Similar way, adapting to AI wouldn't too tricky too.

All of us have been implementing things through manual logic so far. For example, If I need to find patterns in some given data, I'd write some regex, extend it over time to cover more possible inputs and it would just work with the frequent updates to code. Think it in AI way. I'd build a model, train it with sample inputs and outputs and over time it'd be able to learn the patterns by itself without having the modify or extend the code. Sounds fascinating to me. People want things like these, that don't require a dedicated dev to maintain themselves, that run on itself, just like a business that runs on autopilot.

Try to stay strong and think how can you make something easier for someone else and build it. You don't necessarily have to work for someone else to be able to make a living. And even if you have to, don't think AI is not for you. Apply it (theoretically) on what you're doing, maybe you can find some ways to do something lot easier than what that is. Hope it helps.

I am worried too. But not as much for the AI itself, but the hoards of brainless humans (including quite a number in positions of at least some power) who are being seduced by the idea of AI without a hint of critical thought.

As a regular person, I would suggest that one method is to not over-intellectualize it, as much as promoter types may throw extreme possible future outcomes at the screens one look at.

Having thought a little about it, some thoughts below. If something helps, great. If not, I tried. What do you think of the following concepts?:

1. Big fat trends. There is repeatedly a "hype wave" of this or that, in technology. Companies jump on board. Old work gets done using the latest fad. There are some improvements and some detriments as a result. But overall "progress" marches on. So, you have crops of companies doing the same basic busy-work, which still requires integrations. It's a cash cow. Companies at the top may be vertically integrated, making shovels. Lots of people buying and using the shovels to help dig for the gold. The more things change, the more things stay the same, though. In some way, this might just go on this way. In some ways, big AI appears to be just a jobs program to keep people busy, so don't worry too much about it.

2. Limitations. Without being overly specific, the ML systems today can do a lot and almost surely will do more tomorrow. But there are still huge, obvious (to some) problems with these systems, that, currently, and a far as this observer can see into the future, will still require a human in the loop (to actually be as effective as purported). I guess this gets closer to the "politics" that you, the OP, state a valid reason (the pointlessness of it) to stay away from (perhaps for good reason). But there's something to be said for addressing the limitations of machines trained on a bunch of imperfect data. At some point it does seem to come down to ideas, and ideas can be inherently political. I think we have technological progression, and society gets complacent and needs to catch up, in order for people to really be able to make use of the tech advances. Some people have talked about this at length I think. As a chef turned web guy, I think you could pivot into this adjacent/more-human-centric area if eventually needed.

To elaborate on (2) broadly, consider what ML branded as AI is supposedly able to unlock. Then look at the websites of 5 or 6 top AI or AI-tangential companies, including maybe their blogs or hiring pages. And see JUST HOW LACKING in true intelligence those companies really are. (If you can't see it, that's okay.) The point is that the people that develop the AI really are not utilizing very fully AT ALL the larger field or true capabilities/applications of what they sell. If we're going to be all lofty about it, we are obligated to explorer here, too, IMO. I don't think it's that such companies and their employees are too close to the problem. I think it is that the problem space is that large and difficult. This is not meant to be a detractor so much as it shows how much opportunity is out there, and how much careful, sustained effort meaningful FORWARD progress will take.

3. Just stay focused on what is actually real and in front of you. What do the customers want to pay for? Offer that. In tech this seems to often require picking up certain skills along the way I guess, although often old things remain applicable, too. I think that will keep going for a long time.

4. There could theoretically be a burgeoning arts and architecture scene if we don't have to manufacture widgets ourselves so much. I guess we haven't really seen this play out a whole lot, maybe some (not talking NFTs). What is thought to throw this off track would be an asymmetry in art-funders (and art-interested) and art-makers. I don't have the answer here, maybe UBI is part of it. But I think this is an area that have businesses invent creative solutions. Because we shouldn't defund music class to push AI and enviro-tech; there has to be a better way, what is it?

Stay calm and meme on: When in doubt, a good meme can save the day.

First, I’m kind of surprised that low/no-code tools, as well as outsourcing, weren’t an existential threat well before AI. Like, anyone can create a website nowadays, and yet it’s never been a better time to be a web developer.

I think we underestimate the difficulty of our jobs, or perhaps we underestimate the chance that something better will take its place. Perhaps a bit of both.

The other thing that helps me sleep at night is that I really don’t believe that AGI is as near or inevitable as people make it out to be. I don’t necessarily think models will simply get better with more data and I don’t think it’s reasonable to expect that we’ll just somehow figure out intelligence. When I think about AI’s bitter lesson, that nothing has really worked better than scale and data, I take that as a sign that we really don’t understand intelligence. I don’t take it as a sign that “scale is all you need”. Maybe to a certain point, but for AGI? I don’t know man.

Of course, suppose I’m wrong and we create AI that’s smarter than your typical software developer and costs the equivalent of pennies an hour. In which case, well, game over, but I don’t think I’ll feel like much of a loser. Sure, there goes my job, but the economy also probably gets wrecked in the process for a variety of factors. Massive unemployment aside, there’s also the fact that paying for software is kind of stupid when you can just pay a competent AI to make it for you or use some open source alternative made by people pooling AI resources together. This is kind of a big deal considering that tech companies basically carry the stock market. I haven’t even touched on the damage done to the rest of the economy by the existence of intelligence-on-demand for cheap. Either the government considers drastic measures and a radical rethinking of economics, or they ride out the worst economic crisis in history and try to figure out which holy cow gets cut from the government budget now that tax revenue is in free fall. Despite how little faith I have in the government, I trust their self-preservation instincts will kick in and they’ll resort to taking certain socialist economic policies more seriously.

I suspect my job is towards the middle of the range in timing as to when it’s in the crosshairs. And by that point we’ve got a serious societal questions to be answered so not even sure personal perspective matters at that point. So in short not doing much aside from trying to be on the leading edge of applying the tech where I can. (As user, I do dabble in dev too though)

In the short term I’m far more concerned about the quality of info just absolutely crashing. Remember the time before google search that made all knowledge accessible? There is a real risk of a returning to that era. Just be virtue of real drowning in genAI and signal/noise collapsing. Even worse things like political misinformation going viral was not a thing back then. So it’s double bad in that sense

From the title, I honestly thought you were asking, “As a regular, sensible person, how do I put up with all these AI doomsayers?”

It’s a classic bubble. People are overpaying for chips (tulips) because other people are overpaying for chips. The big tech giants see the entertainment value in LLM and don’t want to be caught pants down. These investments are mostly being made on blind faith, not on the basis that any real value will come from it. There isn’t going to be an AI-pocalypse, but we are going to eliminate a lot of copy and paste cookie cutter shit even faster than before. Marketing and advertising feels like an area where that might happen. But the true creative marketing people, who aren’t just the same person at every car dealership or supermarket but truely creative innovators will not be affected. Apply the same logic throughout the marketplace. It’s all worthless shit work anyway. Big tech is uniquely positioned to take advantage of this and be the provider of choice, but it’ll be a race to the bottom in terms of pricing. It’ll end up being a waste of money and the shareholders will punish the ones who go too far. Apple had their chance today and all we see is UX and workflow stuff, not some new “intelligence”. Shit, half their play is just making buttons to send shit to ChatGPT. This shit is desperation and grasping at straws to make good on our egos, to try to make something out of nothing because we fell for something so many humans have falled for: if someone (or thing) is well spoken, they are trustworthy and valuable. Great religious orators have long used this to control the masses. Not to invoke Godwin’s Law, but look at Hitler. This mathematical parlor trick IS the level up. Computers can generate probablisticly realistic words, images and audio. That’s it. It’s already happened. The data wall is looming and we’re all going to go back to laboriously training ML like we have for 20+ years to get little incremental improvements. It’s a bubble. You’re fine.

See: https://en.m.wikipedia.org/wiki/Tulip_mania

People will still be needed to translate the instructions from upper management, i.e. "make it better, now" or "make the red thing do what the blue thing does" most of which are never going to be parsable with any kind of AI, so humans will form the abstraction layer even if they're writing less of the code.

What struck me yesterday was a post about a "$10,000 ChatGPT error" at a startup, where clearly the young devs had "got ChatGPT to do it" and refactor a load of stuff because the code was then in Python of some flavour, which I read between the lines and figured was likely the only coding they knew about from a bootcamp. But then there was a terrible error in the GPT-generated code which caused loads of hassle because nobody caught it, because nobody was really even reading the code upon which the company depended, because who cares, ChatGPT is doing everything now. So it did make me concerned that tech startups are going to shift from being funded with lazy ideas or marketese to being funded with the lazy ideas and marketese of people who cannot even code at all.

Hence it's a step closer towards idiocracy as much as it is our glowing AI future, when Silicon Valley produces neither silicon nor even software, because once the young generation can go "just get the AI to do it" nobody is going to spend time keeping their mental tools sharp by actually coding all day long. Understanding coding could become a rarer and rarer skill in some ways, so today's coders will be the dark wizards of a decade's time.

Also I don't think LLM's will ever reach the point where they can code deeply enough because a large amount of majorly complex material and massive projects are not in publicly accessible repositories. And what I mean too is that I think for all the hype, continuing quantum leaps in AI, and human-coded-at-great-pain quantum leaps in AI are still needed. I don't think that simply throwing more GPUs at the problem is going to keep delivering spectacular returns, rather, diminishing returns, and things might flatten out for a while. I don't think there's going to be some "cascade of free lunch" and AIs will improve themselves on their own, at least not in the immediate future. And that is also a recipe for a mess and chaos.

A good thing to do is to focus on doing what you really love, try reducing your social following closer to zero, and not buy into the hype so much. The AI-pocalypse could still be a long way off and there's no need to worry yourself all the way there. Humans want humans. If an AI could generate me perfect "new" Bob Dylan or Beatles songs, then they'd be interesting for 30 seconds and I'd go back to the real thing. An undead, disembodied AI therapist is also not going to replace an actual flesh-and-blood empathetic therapist any time soon. Much of what we find most profound and valuable does not reduce down to language, which is where AI first took a mis-step, right at the very idea of the Turing test. It's been stuck with that confusion ever since, by people who believe it without thinking philosophically – it reduces us to the only board that AI can ever win on, whereas we have so many more games. Including the "ignoring it because it's an AI" game.

One of the main things capitalism does for us is keep us busy all day. AI automation isn't going to remove this key feature of capitalism.

Simple, stop being scared at made up ghosts from your computer and start being very concerned about the very real threat that is ever increasing global temperatures and pollution. The only people that want you to be afraid about AI, are the people who are either selling AI or are paid by the people that sell AI. It legitimizes their grift and keeps the hype going. The "threat" of AI is nothing compared to what is in store for us once large areas of our planet become effectively uninhabitable.

> How are you dealing with this?

I help to make a podcast, called cybershow.uk

I hope it entertains, informs and emboldens people feeling existential anxiety as you describe.