Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Why is AI so slow to spread?

economist.com

Post reply on HN
Print view View original

51–60 of 189 posts

Re: Why is AI so slow to spread?

#51

Post by lazide »

galaxyLogic wrote: post #16
5350-uiop-1130 wrote: post #5

reminds me of crypto a bit. most people i know are apathetic or dismissive. when i see normies use it - its to make selfies with celebrities. in 5-10 years AI will everywhere. a massive inequality creator. those who know how to use it and those who can afford the best tools. the biggest danger is dependency on AI. i really see people becoming dumber and dumber as they outsource more basic cognitive functions and deci…

> in 5-10 years AI will everywhere. a massive inequality creator. That is possible, even likely. But AI can also decrease inequality. I'm thinking of how rich people and companies spend millions if not hundreds of millions on legal fees which keep them out of prison. But me, I can't afford a lawyer. Heck I can't even afford a doctor. I can't afford Stanford, Yale nor Harvard. But now I can ask legal advice from AI, w…

And when that legal advice is dangerously wrong?

At least lawyers can lose their bar license.

Re: Why is AI so slow to spread?

#52

Post by WarOnPrivacy »

ethan_smith wrote: post #15
WarOnPrivacy wrote: post #8

I don't often use AI in my work because it is not sufficiently useful not sufficiently trustworthy. It is my ongoing experience that AI + My Oversight requires more time than not using AI. Sometimes AI can answer slightly complex things in a helpful way. But for most of the integration troubleshooting I do, AI guidance varies between no help at all and fully wasting my time. Conversely, I support folks who have the c…

The variance in utility you're seeing is largely because AI performs best on problems with clear patterns and abundant training data, while struggling with novel edge cases and specialized domain knowledge that hasn't been well-represented in its training.

This is a reasonable analysis. It explains where AI is useful but I think it doesn't touch on AI's trustworthyness. When data is good, AI may or may not be trusted to complete it's task in an accurate manner. Often it can be trusted.

But sometimes good data is also bad data. HIPAA compliance audit guides are full of questions that are appropriate for a massive medical entity and fully impossible to answer for the much more common small medical practice.

No AI will be trained to know the latter is true. I can say that because every HIPAA audit guide assumes that working patient data is stored on practice-owned hardware - which it isn't. Third parties handle that for small practices.

For small med, HIPAA audit guides are 100 irrelevant questions that require fine details that don't exist.

I predict that AI won't be able overcome the absurdities baked into HIPAA compliance. It can't help where help is needed.

But past all that, there is one particularly painful issue with AI - deployment.

When AI isn't asked for, it is in the way. It is an obstacle to that needs to be removed. That might not be awful if MS, Google, etc didn't continually craft methods to make that as impossible as possible. It smacks of disdain for end users.

If this one last paragraph wasn't endlessly true, AI evangelists wouldn't have so many premade enemies to face - and there would be less friction all around.

Re: Why is AI so slow to spread?

#53

Post by kaoD »

rckt wrote: post #9

Slow?? AI is literally being shoved into everything. It took only several years to see AI being advertised as a magic pill everywhere. It’s not meeting the expectations, probably because of this aggressive advertising. But I would in no way say that it’s spreading slow. It is fast.

They mean slow to actually be used by people. Doesn't matter that it's shoved everywhere if in the end there are 0 users clicking the magic wand icon.

Re: Why is AI so slow to spread?

#54

Post by sbt »

I have been using it for coding for some time, but I don't think I'm getting much value out of it. It's useful for some boilerplate generation, but for more complex stuff I find that it's more tedious to explain to the AI what I'm trying to do. The issue, I think, is lack of big picture context in a large codebase. It's not useless, but I wouldn't trade it for say access to StackOverflow.

My non-technical friends are essentially using ChatGPT as a search engine. They like the interface, but in the end it's used to find information. I personally just still use a search engine, and I almost always go to straight to Wikipedia, where I think the real value is. Wikipedia has added much more value to the world than AI, but you don't see it reflected in stock market valuations.

My conclusion is that the technology is currently very overhyped, but I'm also excited for where the general AI space may go in the medium term. For chat bots (including voice) in particular, I think it could already offer some very clear improvements.

Re: Why is AI so slow to spread?

#56

Post by gmag »

Opocio wrote: post #50
somenameforme wrote: post #20

Earlier quoted context omitted.

Another issue, one that you alluded to, is imagine AI actually was reliable. And a company does lay off e.g. 30% of their employees to replace them with AI systems. How long before they get a letter from AI Inc 'Hi, we're increasing prices 500x in order to enhance our offerings and and improve customer satisfaction. Enjoy.' The entire MO of big tech is trying to create a monopoly by the software equivalent of dumping…

It's quite easy to switch LLM api, so you can just transition to a competitor. Competition between AI providers is quite fierce, I don't see them setting up a cartel anytime soon. And open source models are not that far beyond commercial ones.

It's easy to switch the LLM API, but in practice this requires having a strong eval suite so that the expected behavior of whatever is built on top changes within acceptable limits. It's really the implications of the LLM switch that matter.

Re: Why is AI so slow to spread?

#57

Post by rini17 »

stillpointlab wrote: post #6

> such as datasets that are not properly integrated into the cloud I believe this is a core issue that needs to be addressed. I believe companies will need tools to make their data "AI ready" beyond things like RAG. I believe there needs to be a bridge between companies data-lakes and the LLM (or GenAI) systems. Instead of cutting people out of the loop (which a lot of systems seem to be attempting) I believe we need…

Give rank-and-file employees access to all the data? LOL. Middle managers will never allow that and will shift blame to intrasingent employees. Of course Economist is pandering to that. LLMs are fundamentally very bad at compartmentalized access.

Re: Why is AI so slow to spread?

#58

Post by rwmj »

I spent a bit of time reviewing and cleaning up the mess of someone who had taken text that I'd carefully written, put it through an AI to make it more "impactful", and revising it to remove the confusions and things I hadn't said. The text the AI wrote was certainly impactful, but it was also exaggerated and wrong in several places.

So did AI add value here? It seems to me that it wasted a bunch of my time.

Re: Why is AI so slow to spread?

#59

Post by hansvm »

wordofx wrote: post #14
WarOnPrivacy wrote: post #8

I don't often use AI in my work because it is not sufficiently useful not sufficiently trustworthy. It is my ongoing experience that AI + My Oversight requires more time than not using AI. Sometimes AI can answer slightly complex things in a helpful way. But for most of the integration troubleshooting I do, AI guidance varies between no help at all and fully wasting my time. Conversely, I support folks who have the c…

I still haven’t found anyone who AI wouldn’t be helpful or that isn’t trustworthy enough. People make the /claim/ it’s not useful or they are better without it. When you sit down with them it often turns out they just don’t know how to use AI effectively.

I'll pick a few concrete tasks: Building a substantially faster protobuf parser, building a differentiable database, and building a protobuf pre-compression library. So far, AI's abilities have been:

1. Piss-poor at the brainstorming and planning phase. For the compression thing I got one halfway decent idea, and it's one I already planned on using.

2. Even worse at generating a usable project structure or high-level API/skeleton. The code is unusable because it's not just subtly wrong; it doesn't match any cohesive mental model, meaning the first step is building that model and then figuring out how to ram-rod that solution into your model.

3. Really not great at generating APIs/skeletons matching your mental model. The context is too large, and performance drops.

4. Terrible at filling in the details for any particular method. It'll have subtle mistakes like handling carryover data at the end of a loop, but handling it always instead of just when it hasn't already been handled. Everything type checks, and if it doesn't then I can't rely on the AI to give a correct result instead of the easiest way to silence the compiler.

5. Very bad at incorporating invariants (lifetimes, allocation patterns, etc) into its code when I ask it to make even minor tweaks, even when explicitly promoted to consider such-and-such edge case.

6. Blatantly wrong when suggesting code improvements, usually breaking things, and in a way you can't easily paper over the issue to create something working "from" the AI code.

Etc. It just wasn't well suited to any of those tasks. On my end, the real work is deeply understanding the problem, deriving the only possible conclusions, banging that into code, and then doing a pass or three cleaning up the semicolon orgasm from the page. AI is sometimes helpful in that last phase, but I'm certain it's not useful for the rest yet.

My current view is that the difference in viewpoints stems from a combination of the tasks being completed (certain boilerplate automation crap I've definitely leaned into AI to handle, maybe that's all some devs work on?) and current skill progression (I've interviewed enough people to know that the work I'm describing as trivial doesn't come naturally to everyone yet, so it's tempting to say that it's you holding your compiler wrong rather than me holding the AI wrong).

Am I wrong? Should AI be able to help with those things? Is it more than a ~5% boost?

Re: Why is AI so slow to spread?

#60

Post by crinkly »

rwmj wrote: post #58

I spent a bit of time reviewing and cleaning up the mess of someone who had taken text that I'd carefully written, put it through an AI to make it more "impactful", and revising it to remove the confusions and things I hadn't said. The text the AI wrote was certainly impactful, but it was also exaggerated and wrong in several places. So did AI add value here? It seems to me that it wasted a bunch of my time.

This is the paradox. It's often harder correcting someone else's work than it is doing it in the first place. So you end up having spent more effort.
Post reply on HN