Is AI actually helpful?

[onehitwanda]

everyone in tech was stupidly excited that you can now write backend services in JavaScript

I wasn't. I was very, very drunk though.

except this time it has gone more mainstream

The cryptocurrency boom of... what, six years ago, is another excellent example of a technology that was going to change everything; a small number of people made fortunes and a large number got absolutely rinsed.

AI is a ponzi scheme. There is some incredibly useful tooling and methodology that is and will come out of it, but the sheer volume of waste that is going on is... staggering. I'm pretty confident that in a year or two, 90% of AI firms will be pivoted or dead, there'll be two to three major AIAAS that exist to rinse corporate customers, and for everyone else we'll simply have a glut of content slop that people will develop the skills to ignore in the same way they currently ignore Grammarly ads on Youtube. Maybe we'll get some cool new things out of it, but the vast bulk of the current valuations are simply going to vanish like smoke and fairy gold. All that's going on right now is an attempt to cash out early so that the rubes can be left holding the bag when it all inevitably goes titsup. A new generation of Meths will buy remote islands and populate them with Stepford Wives; meanwhile, we plebs will be left to deal with the new societal problems that AIs will have created - an entire generation of children who believe what their AI tutor tells them, without learning to expend any effort on even a base level of skepticism. A Brave New World, though one that Aldous Huxley didn't envisage.

I wish I had the money to cut away and farm apples or herd yaks or something. Sadly, that's not my fate. So I'll be the crazy cat lady in the corner who's still typing away in a ssh terminal when everyone else is telling GladOS to make them a fucking sandwich.

Sigh. I'm tired, boss.

 

[nice90sguy]

I'm in my sixties, still actively coding and ML engineering daily after over 40 years.

With careful prompting, some of my day-to-day Python/bash work is sped up by around 25% (so I agree with the results in the article there).
For unfamiliar coding languages to me, like Rust, and for some CSS and Typescript, it's a godsend. But as always, careful prompting is required.
It's also very good at being a gopher for me when I don't want to trawl too many Arxiv papers and github repos. I've found some really useful stuff in much less time than if I'd have had to do the searches myself (and it seems to be able to get to quite a few subscription sites).

Ive tried many copilots, and most of them "know what I want to do" if instead of coding, I write a well-thought-out comment instead.

 

[ElectricBlue]

I'm in my sixties, still actively coding and ML engineering daily after over 40 years.

With careful prompting, some of my day-to-day Python/bash work is sped up by around 25% (so I agree with the results in the article there).
For unfamiliar coding languages to me, like Rust, and for some CSS and Typescript, it's a godsend. But as always, careful prompting is required.
It's also very good at being a gopher for me when I don't want to trawl too many Arxiv papers and github repos. I've found some really useful stuff in much less time than if I'd have had to do the searches myself (and it seems to be able to get to quite a few subscription sites).

Ive tried many copilots, and most of them "know what I want to do" if instead of coding, I write a well-thought-out comment instead.

Have you encountered hallucinations, though? If you get one, how on earth can you rely on any other content the tool produces? From what I've seen, it's fundamentally unreliable, because of that alone. Not what you want helping someone code an aircraft flight control system, for example...

 

[Jackie.Hikaru]

Have you encountered hallucinations, though? If you get one, how on earth can you rely on any other content the tool produces? From what I've seen, it's fundamentally unreliable, because of that alone. Not what you want helping someone code an aircraft flight control system, for example...

ChatGPT is always right! Except in my area of expertise… Weird

What unsettles me is that when it’s wrong it’s often almost right, yet it always uses a tone that’s very tuned to convincing you that it’s 100% right.
The problem is there are so many use cases where almost right can hurt people or get them killed…

I’m not necessarily against LLMs. I’m against how they’re being implemented, and I’m against this whole dot com bubble vibe we got going on, and I’m against Microsoft and google and the rest of those fuckers trying to shove it down my throat

 

[BobbyBrandt]

Eleven Labs has an AI audio program that is fantastic, however the company can't get out of its own way; the platform is based on credits and requires a lot of trial and error, so you burn up a lot of credits very easily.

More importantly, because of their rigid TOS, the software is not allowed to be used for erotic stories or porn, making it very useless for a site like this.

Also, for a variety of reasons, it's almost impossible to find a site that will allow AI audio to be uploaded there.

Literotica doesn't allow it and I completely get why, for both moral and practical reasons.

Amazon has had their Polly project live for a couple of years.

They recently integrated it into a "Virtual Voice" beta program which allows for e-books on KDP to be converted to audiobooks that are offered on Amazon and Audible.com. This provides exposure to a large audience of audiobook customers.

They will allow published e-books containing erotic content to become audiobooks, but they automatically insert an adult content disclaimer.

 

[ChuckWolf]

Amazon has had their Polly project live for a couple of years.

They recently integrated it into a "Virtual Voice" beta program which allows for e-books on KDP to be converted to audiobooks that are offered on Amazon and Audible.com. This provides exposure to a large audience of audiobook customers.

They will allow published e-books containing erotic content to become audiobooks, but they automatically insert an adult content disclaimer.

That's a better way of doing it.

 

[BobbyBrandt]

That's a better way of doing it.

And so far, it's free of cost to produce the audiobook

 

[EmilyMiller]

ChatGPT is always right! Except in my area of expertise… Weird

Haven’t you heard, expertise is out of fashion. People don’t like “elites” telling them stuff. But LLMs doing it is fine, obv.

 

[Jackie.Hikaru]

Haven’t you heard, expertise is out of fashion. People don’t like “elites” telling them stuff. But LLMs doing it is fine, obv.

Maybe they’ll listen to me again when I start throwing cats at them. That should work… right? @onehitwanda ? Should I start stockpiling cats??

 

[GuiltyCowboy]

All of our [interminable] debates about AI seem to assume some sort of benefit to the technology. Arguments instead talk about ethics and loss of human jobs. But what if AI doesn’t actually boost productivity?

In a wider sense, AI is absolutely not helpful right now because what the world really needs is a massive boost in the humanities: we need geniuses in behavior and human experience and community and humor and dignity.

Almost all the problems in the world right now require expertise in navigating us through messy social and political situations.

In this sense, self-driving cars and AI lesson-planners are completely fucking irrelevant right now.

We don't require tech that, for the sake of dubious future benefits, is right now - in this real world in which we're living - an actual cause of environmental destruction, disaster capitalism, gross inequality, the hollowing out of public institutions, and the loss of jobs and craft.

The only reason AI is getting the funding and PR attention in the media right now is because it's going to enable large companies to reduce their wage bills and allow to the very wealthiest to feel better about avoiding their tax responsibilities (because AI will provide superficially cheaper and definitely worse services in education and healthcare).

 

[onehitwanda]

Maybe they’ll listen to me again when I start throwing cats at them. That should work… right? @onehitwanda ? Should I start stockpiling cats??

Pussy, babe. I said Pussy.

I remember 2010-11, when economics professors lined up for interviews in every paper that would have them, all insisting Bitcoin was a Ponzi scheme. We don’t hear much from them today.

A friend of mine bought a thousand Bitcoins in late 2010 for twenty cents each and tried to convince me to get in too. I dismissed him with disdain. He sold at $30 a few months later and was the happiest man alive until he had to buy back in at $1,000 two years later… and again at $17,000. He stopped bragging about his trades once the tax authorities started chasing people like him.

I'm not referring to the speculator's godsend that is Bitcoin and cryptocurrency "markets", I'm referring to blockchain and all the related technologies which are... nowhere near as widespread as the proponents were screaming they would be. Sure, there's some utility in being able to "buy" something with bitcoin... but it's a flea circus compared to the traditional finance systems it was fated to replace.

Furthermore, Bitcoin *is* a ponzi scheme. People buy it because the value goes up, and the value goes up because of artificial scarcity caused by limited supply. As the value goes up, more people want it, so the value goes up more. It is a completely arbitrary hedge of value, unbacked by anything except the notional utility that you can exchange it for things, sometimes. It has value because people with it say it does, and the notional utility of being able to "send it" to someone is far, far lower than the very concrete utility of me picking up my phone and using my banking app to do the same thing - with a currency whose value remains stable on a cycle measured in months. There is no magical new economic system that is going to be driven by it, and the stupidendous amount of investment locked up in it is almost entirely speculative.

AI is ten times worse.

Ninety percent of startups collapse after two or three years, so there’s no reason to think AI and crypto ventures will be any different. Still, with the recent maturation of quantum processors combined with AI, we may be on the brink of an evolutionary leap in human nature unlike anything since the first use of spoken language.

Quantum processors have not matured, they cannot even reliably factor the number 21. Quantum processors also show no applicabilty to general computing. Calling out a paradigm shift is... premature.

It’s been fifteen years since the famous purchase of two Domino’s pizzas for 10,000 BTC, and the price now stands at $112K.

The price is a measure of rarity, not utility. Don't confuse the two.

 

[nice90sguy]

Have you encountered hallucinations, though? If you get one, how on earth can you rely on any other content the tool produces? From what I've seen, it's fundamentally unreliable, because of that alone. Not what you want helping someone code an aircraft flight control system, for example...

Yes, hallucinations are the bane of most copilot users, so careful prompting is required. I tell copilots to cite sources, use current documentation (e.g. "September 2025"), and -- which seems to work, not to hallucinate. ChatGPT will notice how stringent I am, and enter its chain-of-thought reasoning (burning up slightly more fossil fuels as it does), and do a few web searches, before venturing to supply code. For Python, some copilots can also execute the code first, avoiding some typos.

And yes, we're a long way from relying on copilots for much more than prototyping.

 

[iwatchus]

If asked to cite sources, ChatGPT has a history of fabricating them. Don't rely on the prompt to avoid hallucinations.

 

[nice90sguy]

If asked to cite sources, ChatGPT has a history of fabricating them. Don't rely on the prompt to avoid hallucinations.

When it does that,. it provides a link -- admitedly not always to a real page! But forcing it to cite sources prevents it from delving too much into its own, out-of-date and faulty, "knowledge".

 

[AwkwardlySet]

A bit off topic, but after participating in many topics about AI here, I've formed an impression that a staggering number of AH denizens work (worked) in IT. Literally, every second or third person in these threads mentioned their coding background.

Soooo... any reason why programmers seem to be perverts extraordinaire on average? It can't be a coincidence that so many of us found our way to this place.
Is it maybe because zeroes look like tits? Or because ones look like penises that need attention?

I mean, those guys who code in Python, I can see their wishful thinking... as well as the ladies who prefer coding in C++, but those like me who prefer Java, we are proud of, err, finishing Just In Time.

 

[onehitwanda]

Just In Time.

Get out, and take your write-once-run-nowhere-well crap with you.

any reason why programmers seem to be perverts extraordinaire on average

Because those of us who are longer in the tooth are bored, and humans are by their nature filthy little animals.

 

[AwkwardlySet]

Get out, and take your write-once-run-nowhere-well crap with you.

I didn't mean to ruby you the wrong way...

 

[onehitwanda]

I didn't mean to ruby you the wrong way...

I can't hear you over the glorious Elixir I'm drinking.

 

[AwkwardlySet]

I can't hear you over the glorious Elixir I'm drinking.

Meh, then I shall be gone Swiftly.

 

[onehitwanda]

Meh, then I shall be gone Swiftly.

Don't get Rusty on your wait out, or you'll start to Lisp.

sorry, I'm mov ax,4c00h; int 21h; right now.

 

[AwkwardlySet]

Don't get Rusty on your wait out, or you'll start to Lisp.

sorry, I'm stopping right now.

Damn, I See Sharp rebukes here.

 

[MrPixel]

Yeah, the puns are starting to SNOBOL.

 

[AwkwardlySet]

Yeah, the puns are starting to SNOBOL.

Maybe we should make an Assembly of them.

 

[TheLobster]

I definitely C what y’all are doing here.

 

[ShelbyDawn57]

Maybe we should make an Assembly of them.

Just Go, already.