[fasterik]

I love computers too, but it doesn't resonate with me when people call AI "snake oil." The comparison suggests that the thing doesn't do what it's marketed to do. AI does more or less what it's marketed to do, sometimes badly.

I still write code by hand. But LLMs have been a legitimately useful tool when I've wanted to dig into a new field like computer graphics, theoretical physics, or numerical analysis. Or even just asking the LLM to write a piece of code and learning from its output. I think it makes me a better programmer because I can bootstrap the knowledge needed for a new project much faster and spend more time programming.

[drchickensalad]

> The comparison suggests that the thing doesn't do what it's marketed to do

In my opinion you should interpret the usage of "AI" here to mean "the entire business/management/financial/bubble ecosystem surrounding LLMs". The snake oil is much more how LLMs are being weaponized and utilized rather than a specific technical assessment (although that often is an issue too)

[indymike]

> The snake oil is much more how LLMs are being weaponized and utilized

Calling something "snake oil" implies that the product only, at best, has a placebo effect (amplified, no doubt, by the sales presentation). LLMs seem to be much more useful than a placebo.

[datsci_est_2015]

I think you just simply repeated the mistake that the comment responding to you was pointing out, i.e. conflating “LLM” the technology with “AI” the hysterical product hype bubble phenomenon.

[patates]

I think then the AI-hype should be disregarded as snake-oil, not the tech itself. I know it sounds pretentious, but I need to say that words do matter.

[fasterik]

My prediction is that it will go the same way as the dot com bubble. The hypesters and fraudsters will eventually collide with objective reality, but the technology will persist and society at large will benefit from the infrastructure and the increased access to knowledge.

[Shellban]

Assuming that we recover from the damage being done now. As one example, a friend of mine has remarked that large corporations will benefit from the current AI-induced reality of no one being able to afford their own hardware, and keep prices that way to enforce a renters model on computers.

[thewebguyd]

That's my fear, and unfortunately I think its likely to happen. I feel like we will settle down a little bit, but into a new higher "normal" baseline that still largely makes it unaffordable for most.

There's also still the risk of the creation of a new economic underclass, if both a) hardware remains too expensive for local inference and b) subscription or pay-per-token based inference also remains expensive or increases in price, then individuals will largely be locked out of the benefits that having access to AI could bring, leaving it purely in the hands of larger companies. People will only get to use and experience these tools through their employer, for the benefit of their employer.

[teo_zero]

> People will only get to use and experience these tools through their employer

Like in the era of mainframes, before the personal computer was a thing. I wonder if universities will play the same role, though.

[abc123abc123]

This is not correct. Prices will adjuts when the crash comes. At worst, the market and prices will just open peoples eyes up to the reality that 99% of the daily software we use, runs perfectly fine on a 5 years old computer.

The idea that you _must_ have 128 GB RAM and 1 TB ssd in your computer is juts absurd.

Remember... we reached the moon with the compute power of a pocket calculator, and there is no eternal law that says that everything has to be written in javascript.

[thewebguyd]

> The idea that you _must_ have 128 GB RAM and 1 TB ssd in your computer is juts absurd.

I don't think anyone thinks that? Unless you are trying to do local inference with bigger models, which isn't everyone's cup of tea but I do think is an important capability.

Outside of that use case though, prices are still ridiculous. 32GB of DDR5 will run you close to $500 right now vs. $80 before the AI build out, and 32GB is what I'd consider comfortable so long as everything keeps being electron and web wrappers (assuming Windows & macOS here for the general population, obviously you can get by with less on a Linux desktop for the most part).

Personally, I'd love to see more software do more with less, and go back to performant native apps but in reality I just don't see that happening, except maybe over here in Apple/macOS land which has always had a decent culture of native apps. For now, the incentives aren't there (for commercial software).

I hope you are right on prices, but I don't share the optimism. We've seen time and time again that once prices go up, they don't always go back down even when the supply crunch is eliminated. Manufacturers realize people adapt to whatever "new normal" prices are at and refuse to lower them. Same thing happened post-COVID supply shocks. Companies have zero incentive to start a price war, even after supply shocks are over.

I think we are stuck with ~$500 RAM for a long time.

[protocolture]

>large corporations will benefit from the current AI-induced reality of no one being able to afford their own hardware, and keep prices that way to enforce a renters model on computers.

Depends how long the RAM correction takes. It is interesting how RAM prices have stifled the creation of cheap laptops capable of running big models. But at the same time, this seems like a second order effect and not the intention.

[api]

It’s a demand spike.

[echelon]

> As one example, a friend of mine has remarked that large corporations will benefit from the current AI-induced reality of no one being able to afford their own hardware, and keep prices that way to enforce a renters model on computers.

Choose one:

- You spend 30 hours writing a program to manage data for your hobby. You write it on your personal computer.

- You spend one hour generating a program to manage data for your hobby. You have to lease an H200 behind an API to do it.

Which one will you choose?

I know which one I'm choosing.

[anon7725]

I choose A.

I know that many others choose A as well.

A wonderful service known as the web has connected people who choose A with others who choose A and of course with a great many who don’t need to make a choice and benefit from the work of others.

I mourn a world in which few will choose A, because for many to choose B seems to lock us all, tragedy of the commons style, into a worse world.

[account42]

And let's not forget that choosing B is only possible at all because many people chose A before you.

[dgellow]

I’m choosing the first option. I’m not sure which one you chose?

[sdenton4]

Of course, you can use Option B to write the program, and then run it on your own machine...

[bigstrat2003]

Option A is fun, whereas option B is miserable. Option A is also cheaper. It's a pretty clear win for option A here.

[Hugsbox]

You're making a really strong case for option A, whereas B sounds really depressing

[gyomu]

> society at large will benefit from the infrastructure

Data centers as infrastructure are very different from DSL rollout though. Much, much more expensive to maintain, with a much much shorter timespan.

If the bubble pops and data centers get shut down because there’s no one to pay the bills, there won’t be much left 5-10 years later in terms of infrastructure.

[m0llusk]

Maybe we could repurpose old processors to power toaster ovens.

[bigbuppo]

I mean, a 125kW server stuffed into 5U will definitely work nicely for toast.

[api]

The same thing happened around PCs, gaming, the Internet, the web, and cryptocurrency. It's a hit driven industry that loves hype.

[AgentME]

LLMs remind me of being a kid again being in wonder of all the possible things that could be done with a computer that haven't been figured out yet. The internet was relatively new and everyone had their own ideas of what that would enable. Fast forward to a few years ago and it was easy to believe that a lot of the low-hanging fruit of things an individual could do with the internet, apps, 3d graphics, etc, had been decently picked over and that things were stabilizing. Now I have no idea again what computing will look like in 5 years and it's exciting.

[dosisking]

I have a pretty good idea of what computing will look like in 5 years and it's pretty depressing.

[knownjorbist]

You have convinced yourself of having a pretty good idea, not to be confused with having anything close to a crystal ball's view.

[keybored]

Nestlé can give us our daily allowance of baby formula and [AI corporation winner] can give us our daily allowance of free compute. These years will bring interesting times!

[ux266478]

The fats in Chinese Water Snakes are rich in omega 3s and do have genuine benefits to consumption. The problem with snake oil wasn't that it was useless. The problem was with hucksters selling it as a cure-all for everything from cancer to syphilis. The metaphor is pretty apt IMO.

[tptacek]

Yes, I got that Google summary too. But "snake oil" patent medicine didn't contain snake oil.

"Snake oil" refers to something sold as a medicament that has no beneficial effect.

[tcmart14]

Exactly what I was thinking. It's not that snake oil sales people sold totally useless stuff, its just that the stuff they sold did not deliver the value that was promised. Another example that is still going on today. There is a community of people that swear the ingesting silver prevents all kinds of things, even so far as a cure for cancer. It's snake oil, but it doesn't mean it doesn't have any medicinal purposes. Silver does have anti-microbial properties and can be used topically to manage infections.

[zem]

the problem, ironically, is that hucksters were selling other oils as "snake oil" when they didn't have the same omega 3s. the bad reputation was due to fake snake oil.

[mid-kid]

The snake oil is how the people at the top scream "in x years we won't need programmers" and end up proving themselves wrong time and time again. It's a real technology and it can do a lot, but it's being sold like snake oil while we're still figuring out what it's actually useful for and how to leverage it properly.

[tptacek]

Snake oil implies that it does nothing, not that it doesn't do everything it's boosters claim it does. Snake oils were medicines sold as cure-alls with no active ingredients.

[girvo]

I wonder what the better pithy phrase would be then for "thing that is obviously useful, but is being hyped beyond it's (current) ability by those with a vested interest in doing so"

[tptacek]

You could reasonably call it "overhyped". People will disagree with you, but that's fine; you won't be making a falsifiable claim.

[akerl_]

That seems pretty par for the course for every major advancement in technology. So maybe just “capitalism”?

It’s hard for me to think of any piece of new tech that hasn’t been over hyped by the people selling it.

[girvo]

For sure. I'd argue this cycle is operating at a different scale though

[lostmsu]

> "in x years we won't need programmers" and end up proving themselves wrong time and time again

This is how it looks in your head, maybe. But in reality since Sonnet 3.5 - when the whole "no need programmers" started - no "years" have passed. Sonnet 3.5 came out on June 20, 2024. We are still 5 days away from the lowest possible "years". So even if you quoted them literally, they could not have possibly proved themselves wrong yet even once, let alone "time and time again".

[mid-kid]

It was just an example of the type of shit they say to sell it, and then walk back from.

https://fortune.com/2026/05/26/sam-altman-dario-amodei-walki...

[lostmsu]

Yeah, and you hallucinated that example.

The link isn't any stronger either.

[otabdeveloper4]

The moronic "no need programmers" hype cycle happens every 15 years. We've all been here many times before.

[tines]

> I think it makes me a better programmer because I can bootstrap the knowledge needed for a new project much faster

faster != better

[JCTheDenthog]

For me, the painful banging my head against the wall to figure something out is usually the most rewarding sort of learning experience.

[specproc]

AI is snake oil. It sells you a slot machine in the guise of a colleague.

Oh, not using it right? Not the right model? Insert coin to continue.

Snake oil, total snake oil.

[holoduke]

I just reversed engineered large parts of my 2011 car odb comms. Was able to hook a stm32 board to the car communication and have full control over a lot of stuff so that I can build my own instrument cluster from a lcd screen. It literally took me one evening to get the first proof of concept working. I never touched stm32 stuff before.

[dgellow]

you did none of this. Did you personally learn something from that experience? Other than how to use your agent

[holoduke]

Of course I learned from it. I mean the reverse engineering part which is basically try and error is something I rather skip. The remaining things like wiring the hardware is still there. The boring stuff is what the LLM can do for me. I still find the process to get stuff working challenging and interesting. It's not only about the end result. It's just a different approach than the old school low level one

[specproc]

Good for you, son. I've just spent the whole f*king day screaming at an agent on a deadline.

[rolandog]

This right here. People simp for LLM companies as if their experience of using the out-of-pocket top-of-the-line "team of PhD's" paid models will be what is deployed when trying to contact your bank, insurance, etc. No,... once tech companies stop playing the "no/some revenue until we own the world" VC game, we'll all be stuck trying to talk to GlueSnifferGPT when reporting an emergency.

[holoduke]

That will never happen. You know that. It will only become more and more. there is no return. My advise: don't stick with those blocker ideas.

[rolandog]

Not sure I follow; more and more what? And what do you mean by blocker ideas?

[tptacek]

This doesn't even make sense. Maybe if you fleshed it out?

[totallymike]

It makes perfect sense to me. Type in a prompt like “how can I make the cheese on my pizza stringier” and maybe it’ll tell you to use different cheeses, but maybe it’ll tell you to add glue.

If you don’t like the answer, don’t worry, they’re building more data centers in poor neighborhoods so you can keep submitting the prompt until you get a better one.

[tptacek]

What year is it, that you're getting frontier model responses saying to put glue in your pizza?

[totallymike]

I trust that you can use your reading comprehension skills to understand that by referring to a famous example of LLMs producing garbage, I’m simply using it to illustrate the phenomenon at large, rather than to suggest that I am still struggling to find glue-free ways to make my pizza stringier.

If you still need help breaking down what I meant in the previous post, feel free to ask. Sentences can be tricky.

On a related note, just because Gemini doesn’t tell you to put glue in your pizza anymore, by no means implies that this particular problem is solved: https://www.404media.co/it-is-trivially-easy-to-use-reddit-t...

[tptacek]

I'm sorry, I could have said this more plainly: no, no frontier model for the past several years has told you to put glue in your pizza.

[specproc]

Snake oil is something not medicine sold as medicine.

AI is something not a colleague (a slot machine), sold as a colleague.

[tptacek]

Right, that's what you already said. What I don't understand is the "slot machine" analogy you're making. In what sense is AI a "slot machine"? Are you talking about the stocks of AI companies?

[specproc]

Variable intermittent reward. Pull the lever, get something nice or something bad. I've had far too much bad today and I'm furious.

[spamizbad]

I would say the claim that AI is going to replace most white collar work a very snake-oily term. The technology behind it however is very compelling and interesting.

[tptacek]

I don't know enough about most white collar work to make any predictions. But I know a lot about software development and information technology because I've been a professional since 1995. The claims being made about AI's impact on that profession do not seem at all snake-oily to me.

[dosisking]

> I love computers too, but it doesn't resonate with me when people call AI "snake oil."

Not to worry, a fool and his money will soon be parted.

I think the main problem is that there is no definition for "AI".

And from your use case, I don't see any difference between that and a search engine.

[godelski]

  > The comparison suggests that the thing doesn't do what it's marketed to do.

Because it doesn't.

What AI is being sold as is incredibly different than what it actually does. I love AI. I spent years in grad school researching it because I loved it so much (it was never about the money to me). But what it is and what it can do is so different from what it is being sold as.

Snake Oil is an apt comparison because it is being sold as a cure-all. Medical problems? AI. Financial problems? AI. Scientific research? AI. <Insert problem>: AI... It isn't that ML[0] can't help with these problems (it can!), it is that "AI" is being sold as a solution to these problems. As if humans will be obsolete in 6months[1].

LLMs are a fantastic example. We (lossy) compressed the entire internet and build a human language interface into it. That's some real Sci-Fi shit right there. That's an incredible achievement with a lot of utility! But how is it sold? If you call it what it is people will act like you're diminishing its status. We've exaggerated the accomplishments so far out of proportion that we can't even recognize big of an advancement that these machines actually were. LLMs were a huge step forward, but even a giant is small when you compare it to a titan.

So yeah, I do think it is being sold as Snake Oil. And that's been my fear for quite some time (you can dig up my history if you're that passionate). But that's also what we've done with every major tech recently. Hell, even cryptocurrency has real value. The thing that killed it was all the hype built around it when the tech was just in its infancy. Do we really want to do the same thing to AI? It certainly has more utility to it than cryptocurrencies. But it doesn't matter how good the actual product is if people are sold on something else. What matters is how the actual product matches to peoples' expectations. There is such a thing as "overselling", and we're certainly doing that as a community. I know it is an exciting field and there's lots of exciting technology, but we can't promise the moon if we can't deliver.

  [0] It wasn't long ago that "AI" was a red flag and "ML" was seen as less likely to be bullshit. 
  [1] I'm still waiting on my self-driving car...

[bigbuppo]

And why was AI seen as a red flag?

[totallymike]

It was a red flag because artificial intelligence doesn’t exist, and anyone claiming to use it or work on it is either lying or delusional in thinking they could accomplish it.

Nowadays people just say “AI” when they mean “LLM,” which is an unrelated thing entirely, but people want people who use it.

[bigstrat2003]

> I love computers too, but it doesn't resonate with me when people call AI "snake oil." The comparison suggests that the thing doesn't do what it's marketed to do.

Well yeah, because it doesn't. AI is being claimed to be a magical genius intelligence which will solve everything forever, but in reality LLMs are still idiots you can't trust to not screw up without a tight leash. They can't even do the one thing they are supposed to be good at (programming) well, despite all the effort which has been focused on trying to make them good at it. They don't remotely do what they are marketed to do, not even close!

[overgard]

Isn't coding solved and we should all be out of a job by now according to Dario? Or what about AI 2027 -- we're only 6 months away! Time to build a bunker!! LLMs themselves aren't snake oil, they're just a useful technology, but all the marketing around them is FUD mixed with hype mixed with the most irritating people on the planet (the ones that aren't bots at least).

[lostmsu]

> Isn't coding solved and we should all be out of a job by now according to Dario?

You sound like you have a quote in mind.

[keybored]

And we’re back to the pitch.

[moron4hire]

We don't need the caveat that it's sometimes useful on every post about the problems of AI.

[echelon]

This is less an anti-AI post and more a post against the greed of the industry:

> But things feel different now. I can relate to what Chris Person said when he expressed his frustrations about how these slick conmen are using the technology I adore as tools for exploitation and disempowerment. The Internet, built by idealists on a foundation of openness and community, has become a mire of dark patterns and gardens with ever thicker walls, desperate to keep people within an ecosystem where their attention is the prized commodity. I’ve witnessed a nerdy space full of nerds be invaded by marketers, callous capitalists, and “brogrammers”—exaggerating the worst, most toxic, aspects of geek culture in their pursuit of money and power. I’ve poured hundreds of hours of work into open source projects only to have it all be scraped into a plagiarism machine and then aggressively sold back to me. It feels that the hope I had for the future technology could give us, the naïve and starry-eyed fantasies I fostered in my youth, has been eroded when faced with a reality where the thing I love can make a lot of money for people who don’t care for any of it.

You can simultaneously believe that AI is really cool and also that also a lot of companies are degrading the internet, society, and private ownership at large.