4 ms·

I know the “AI label” drives me away. It means the product is unreliable and a black box. The “AI” label also indicates the solution is way over complicated an

by lumb63 2y ago

I know the “AI label” drives me away. It means the product is unreliable and a black box. The “AI” label also indicates the solution is way over complicated and simpler ways to improve the product have been ignored. For instance, Confluence now has an “AI” chatbot. Search is still substantially worse than grep.

javier123454321 2y ago
Yeah, for example, nerdwallet did an AI search/ask chat GPT about finances feature. I believe that not a single customer asked for their budgeting software to have an AI chatbot.
- tail_exchange 2y ago
  But how else would they stamp an AI label to their product? Bolt a chatbot onto it, say you are AI-powered, and profit!
  - jerf 2y ago
    It's even worse than that. It's "bolt a chatbot on to your product, force it into people's workflow against their will, claim huge usage numbers on your quarterly result calls, stock price goes up". Not profits, just stock price. The evidence that anybody who isn't "selling shovels" has made any significant money yet with AI remains thin. I can't remember if it was Google or Facebook, but in the last earnings season one of them claimed directly that AI has improved their advertising revenue, but the improvement was not terribly out of line from what these companies have been doing for a while (not like suddenly they popped out a revenue triple, it was not obviously out of line with their normal reporting), and I wouldn't be surprised they were doing exactly what I said in my first paragraph. I did not find their claim terribly compelling, even if it was completely true and not just a sop to the stock market. Nobody else is even claiming that AI has increased profits. I'm sure there's some startups with some profits, maybe even very exciting amounts of profit for a startup, but nothing that would move the needle for established companies. In fact we seem to already be in the consolidation phase for the industry, the startups are starting to go under. If we use that as a timer for where we are in the hype cycle, this is underperforming compared to other bubbles; at this point in the dot com crash, while the sector was heavily overinvested it was also quite obvious to a calm observer not panicking about their portfolio that there was definitely a there there, it just wasn't ready to sustain that much investment. There is so far a lot less there there with AI. This hype is going to prove disastrous for the entire technology, I fear. If it was treated as a more normal technology, it would improve, it would be experimented with, companies would learn how to use it, failures would be shaken out and successes doubled down on, and in 3-10 years it would be a healthy industry making reliable, good money with a bright future for growth. But the way the stock market went absolutely ape shit, now the bar is, in roughly 6-12 months if AI has not quintupled the profit margin of all of the already-largest companies on Earth, it is a failure, and so are those companies. We've got a lot of people now with all the incentives in the world to blatently lie about their progress because as soon as they are truthful they become personally bankrupt as their stock options tank. Not because of any particular aspect of AI, except that it wasn't a miracle "drop it into any process instantly see quintupled profits", and I'm not inclined to be too annoyed at any technology for failing to be that. And it could take AI down with it. Again. I'd expect the resulting "AI winter" to be relatively short this time because the technology really is becoming promising, but the medium-term prospects for the tech are probably a lot dimmer than they theoretically should be because of this disproportionate frenzy.
  - tcgv 2y ago
    I argue the main issue is that many companies have invested significant resources in poorly assessed, designed, and planned AI implementations, rather than focusing on simpler, achievable, and impactful solutions [1] [1] https://thomasvilhena.com/2024/06/easy-wins-for-generative-ai https://thomasvilhena.com/2024/06/easy-wins-for-generative-a...
ToucanLoucan 2y ago
It will never stop being remarkable to me the sheer volume of enterprises of all scales, from the smallest startups to the most massive titans of our industry, all trip over and fall on their faces when it comes to search. A good search feature can turn a great product into an amazing product, and yet seemingly in every product, it is at best, an afterthought. Even Google which forged it's position in the market today off the back of just doing one thing, search, really, really well, has entirely lost it's way and utterly destroyed it's search product on the altar of profit. And now it too is following suit deploying AI as a half-assed solution for that situation.
lolinder 2y ago
What's worse is companies replacing their search with an AI chat bot. I noticed that on Amazon the other day—where there used to be a feature where you could search all the Q&As and all the reviews for a product, there's now a chatbot that seems to just be ChatGPT with the product description given as context (if you poke at it you can get it to talk about itself as ChatGPT, so I don't think it's an Amazon model). Completely worthless. There's still a search for reviews, but I can't find one for Q&A or for both together.
- wasteduniverse 2y ago
  I used that chat bot to try and find the dimensions of a product and it couldn't even fetch them from the information table that was right above it in the page. What's the point of working your developers to the bone (as Amazon is famous for doing) if the result is going to be this bad?
JoshuaRogers 2y ago
If anyone from JetBrains is reading this thread, please take the above to heart. I have yet to have any devs in my circle make the comment that "the JetBrains AI assistant really saved me time" but I've heard more than once about how distracting (and generally incorrect) its guesses are. I'd really rather have time invested in making the memory footprint smaller again.
- rurp 2y ago
  JetBrains products are such a dichotomy. The core software is outstanding but for years now I've found almost every update to be net negative. They rarely add new features that I find useful and they spend a lot of time adding useless stuff that adds clutter and hurts performance, or they change existing features for no apparent reason.
  - ffsm8 2y ago
    They probably put product people in charge. Which is an incredibly misguided idea specifically in the context of a developer IDE... I mean I do understand the rationale and value for the job in other contexts. Not so much in this one though.
alkonaut 2y ago
Because decisionmakers literally think that potential customers who are comparing products side by side look at whether ProductA has a tick in the "AI" column and ProductB doesn't, so A must be better. I think it's as simple as that. They cater to the feature-matrix thing, they get a win for their yearly review because they "implemented AI". A good search makes users love your product. But that doesn't show up in a feature matrix, nor is it a great review bullet point to have "improved the search somewhat".
ffsm8 2y ago
I for one love well implemented AI features. Take immich search for example. No cloud connection, and let's you search for pictures in your library by describing it i.e. a kid with a red jacket on a bicycle... And you get the picture you made, but didn't know exactly when etc. Very rare example though. As a matter of fact, I couldn't name another one right now.

by lumb63 2y ago

javier123454321 2y ago

Yeah, for example, nerdwallet did an AI search/ask chat GPT about finances feature. I believe that not a single customer asked for their budgeting software to have an AI chatbot.

tail_exchange 2y ago
But how else would they stamp an AI label to their product? Bolt a chatbot onto it, say you are AI-powered, and profit!
- jerf 2y ago
  It's even worse than that. It's "bolt a chatbot on to your product, force it into people's workflow against their will, claim huge usage numbers on your quarterly result calls, stock price goes up". Not profits, just stock price. The evidence that anybody who isn't "selling shovels" has made any significant money yet with AI remains thin. I can't remember if it was Google or Facebook, but in the last earnings season one of them claimed directly that AI has improved their advertising revenue, but the improvement was not terribly out of line from what these companies have been doing for a while (not like suddenly they popped out a revenue triple, it was not obviously out of line with their normal reporting), and I wouldn't be surprised they were doing exactly what I said in my first paragraph. I did not find their claim terribly compelling, even if it was completely true and not just a sop to the stock market. Nobody else is even claiming that AI has increased profits. I'm sure there's some startups with some profits, maybe even very exciting amounts of profit for a startup, but nothing that would move the needle for established companies. In fact we seem to already be in the consolidation phase for the industry, the startups are starting to go under. If we use that as a timer for where we are in the hype cycle, this is underperforming compared to other bubbles; at this point in the dot com crash, while the sector was heavily overinvested it was also quite obvious to a calm observer not panicking about their portfolio that there was definitely a there there, it just wasn't ready to sustain that much investment. There is so far a lot less there there with AI. This hype is going to prove disastrous for the entire technology, I fear. If it was treated as a more normal technology, it would improve, it would be experimented with, companies would learn how to use it, failures would be shaken out and successes doubled down on, and in 3-10 years it would be a healthy industry making reliable, good money with a bright future for growth. But the way the stock market went absolutely ape shit, now the bar is, in roughly 6-12 months if AI has not quintupled the profit margin of all of the already-largest companies on Earth, it is a failure, and so are those companies. We've got a lot of people now with all the incentives in the world to blatently lie about their progress because as soon as they are truthful they become personally bankrupt as their stock options tank. Not because of any particular aspect of AI, except that it wasn't a miracle "drop it into any process instantly see quintupled profits", and I'm not inclined to be too annoyed at any technology for failing to be that. And it could take AI down with it. Again. I'd expect the resulting "AI winter" to be relatively short this time because the technology really is becoming promising, but the medium-term prospects for the tech are probably a lot dimmer than they theoretically should be because of this disproportionate frenzy.
- tcgv 2y ago
  I argue the main issue is that many companies have invested significant resources in poorly assessed, designed, and planned AI implementations, rather than focusing on simpler, achievable, and impactful solutions [1] [1] https://thomasvilhena.com/2024/06/easy-wins-for-generative-ai https://thomasvilhena.com/2024/06/easy-wins-for-generative-a...

ToucanLoucan 2y ago

It will never stop being remarkable to me the sheer volume of enterprises of all scales, from the smallest startups to the most massive titans of our industry, all trip over and fall on their faces when it comes to search. A good search feature can turn a great product into an amazing product, and yet seemingly in every product, it is at best, an afterthought. Even Google which forged it's position in the market today off the back of just doing one thing, search, really, really well, has entirely lost it's way and utterly destroyed it's search product on the altar of profit. And now it too is following suit deploying AI as a half-assed solution for that situation.

lolinder 2y ago

What's worse is companies replacing their search with an AI chat bot. I noticed that on Amazon the other day—where there used to be a feature where you could search all the Q&As and all the reviews for a product, there's now a chatbot that seems to just be ChatGPT with the product description given as context (if you poke at it you can get it to talk about itself as ChatGPT, so I don't think it's an Amazon model). Completely worthless. There's still a search for reviews, but I can't find one for Q&A or for both together.

wasteduniverse 2y ago
I used that chat bot to try and find the dimensions of a product and it couldn't even fetch them from the information table that was right above it in the page. What's the point of working your developers to the bone (as Amazon is famous for doing) if the result is going to be this bad?

JoshuaRogers 2y ago

If anyone from JetBrains is reading this thread, please take the above to heart. I have yet to have any devs in my circle make the comment that "the JetBrains AI assistant really saved me time" but I've heard more than once about how distracting (and generally incorrect) its guesses are. I'd really rather have time invested in making the memory footprint smaller again.

rurp 2y ago
JetBrains products are such a dichotomy. The core software is outstanding but for years now I've found almost every update to be net negative. They rarely add new features that I find useful and they spend a lot of time adding useless stuff that adds clutter and hurts performance, or they change existing features for no apparent reason.
- ffsm8 2y ago
  They probably put product people in charge. Which is an incredibly misguided idea specifically in the context of a developer IDE... I mean I do understand the rationale and value for the job in other contexts. Not so much in this one though.

alkonaut 2y ago

Because decisionmakers literally think that potential customers who are comparing products side by side look at whether ProductA has a tick in the "AI" column and ProductB doesn't, so A must be better. I think it's as simple as that. They cater to the feature-matrix thing, they get a win for their yearly review because they "implemented AI". A good search makes users love your product. But that doesn't show up in a feature matrix, nor is it a great review bullet point to have "improved the search somewhat".

ffsm8 2y ago

I for one love well implemented AI features. Take immich search for example. No cloud connection, and let's you search for pictures in your library by describing it i.e. a kid with a red jacket on a bicycle... And you get the picture you made, but didn't know exactly when etc. Very rare example though. As a matter of fact, I couldn't name another one right now.