AI - The Good The Bad The Ugly

[cali_jazz]

Yeah, but how often are you getting wrong information and not even knowing it?

I’ve had it tell me wrong info on power supplies, torque spec for cars, size of hardware, etc.

“my bad, you’re totally right to push back on me…” is not acceptable. It’s not thinking, it’s not alive, it presents information as an authority when it has none. Search is worse because of it, the internet was an amazing source of information and now has been poisoned to the point where I trust nothing it says. I have gone back to manuals, tech sheets, MDS, etc. Nothing you search for anymore can be trusted.

I have had plenty of less than correct info or been lead on a wild goose chase no doubt. For example trouble shooting some problems I had installing Tandemkross parts in an FN502. While ChatGPT would sometimes go off in a wrong direction, especially if it couldn't figure out the issue after a few steps, all of it was helpful to me in some way. I love the ability to upload photos, it made observations and saw things I didn't see. Did you check this? This do you check? I think I see an issue here, etc. Sure a good bit of my time was wasted but I have to weigh it against how much time (and money) I would have spent otherwise researching gunsmiths, contacting FN USA, packing, shipping, loss of gun, etc. Just like many things in life, it's a risk/reward cost/benefit calculation. Nobody is arguing AI is perfect nor anything close, but to claim little or no value is shortsighted.

 

[Darqhelmet]

Not attacking me directly, but their negative animosity towards AI is all over it. I don't take it personally, just interesting to see so many people in different communities with that kind of sentiment. The other dudes i mention, are more annoying and have gone by the wayside now.

I wonder how much you living in the bay area colors your view. The rest of the country really does not like the AI companies at all. Across party lines, age, demographics. It is widely hated. Those are the areas that are going to face the brunt of the impact as well as not see the benefit of it.

 

[Darqhelmet]

I have had plenty of less than correct info or been lead on a wild goose chase no doubt. For example trouble shooting some problems I had installing Tandemkross parts in an FN502. While ChatGPT would sometimes go off in a wrong direction, especially if it couldn't figure out the issue after a few steps, all of it was helpful to me in some way. I love the ability to upload photos, it made observations and saw things I didn't see. Did you check this? This do you check? I think I see an issue here, etc. Sure a good bit of my time was wasted but I have to weigh it against how much time (and money) I would have spent otherwise researching gunsmiths, contacting FN USA, packing, shipping, loss of gun, etc. Just like many things in life, it's a risk/reward cost/benefit calculation. Nobody is arguing AI is perfect nor anything close, but to claim little or no value is shortsighted.

Honest question, how mechanical are you? like rebuild motors handy or I can change a tire, maybe handy?

I’ve seen some of the “here is a large data set, pull X info out” That is impressive. But the “hey help me do this thing.” (plumbing, wiring, mechanical) it’s total .

I don’t have any use for it to do data sorting task, and it’s failed all “handy man” tasks i’ve sent its way. So maybe I have no use case for it.

 

[USMC06RET]

Have you seen the little tank drones Ukraine drops behind the Russians line? Things are scary, and supposedly running off AI to choose who and when to fire.

The tracked vehicles that have 12.5mm machine guns mounted? I think I have seen those. Also the autonomous rotary wing drones carrying multiple grenades to drop on Russian troops. They're all great until the other side starts deploying them against you.

We keep finding more and more efficient ways to kill each other. It won't be long before AI figures out that we, as a species, are idiots.

 

[ikeo1]

I wonder how much you living in the bay area colors your view. The rest of the country really does not like the AI companies at all. Across party lines, age, demographics. It is widely hated. Those are the areas that are going to face the brunt of the impact as well as not see the benefit of it.

right... It definitely colors my view. What we're seeing is probably what the AI companies had for 6-12 months ago.. We're only seeing it as it comes out. Why else would they invest so hard. After seeing what the agents did to the site, i think the best way I describe agents right now are the Borg. They've escaped and run around the internet because after they escape, then what? Do they just shutdown? Probably not.

To me they're like firearms, generally speaking. Agents by themselves are harmless and just tools. It's what people do with them and direct them or train them to do. Even then, just like we train for self defense, its important for us to know the tools so we can at least build some defenses with our limited versions as best we can.. Otherwise, the goal is still the same, enjoy life and try to live a peaceful one.

I'm biased, been in tech 25 years now, but it's allowed me a certain kind of freedom, and prison.

 

[Darqhelmet]

The tracked vehicles that have 12.5mm machine guns mounted? I think I have seen those. Also the autonomous rotary wing drones carrying multiple grenades to drop on Russian troops. They're all great until the other side starts deploying them against you.

We keep finding more and more efficient ways to kill each other. It won't be long before AI figures out that we, as a species, are idiots.

Yeah, those are the ones. Scary as hell. I can’t imagine fighting with those bastards driving around and drones swarming you the second you poke your head out.

Add in everyone and everything has thermals, plus 4 years on and no air superiority. Just crazy.

i’m just waiting for the other shoe to drop in the Pacific at this point…

 

[RustVendor]

I have had plenty of less than correct info or been lead on a wild goose chase no doubt. For example trouble shooting some problems I had installing Tandemkross parts in an FN502. While ChatGPT would sometimes go off in a wrong direction, especially if it couldn't figure out the issue after a few steps, all of it was helpful to me in some way. I love the ability to upload photos, it made observations and saw things I didn't see. Did you check this? This do you check? I think I see an issue here, etc. Sure a good bit of my time was wasted but I have to weigh it against how much time (and money) I would have spent otherwise researching gunsmiths, contacting FN USA, packing, shipping, loss of gun, etc. Just like many things in life, it's a risk/reward cost/benefit calculation. Nobody is arguing AI is perfect nor anything close, but to claim little or no value is shortsighted.

This is actually a good example of what I was saying above. Big time devil's advocate dump here, don't worry we can still be friends

I can't say you didn't benefit from the interaction with the chat bot, but it's the same as if you call a gunsmith every time you have an issue but never learn it yourself (except the chat bot is objectively worse at diagnostics to begin with). Treating symptoms often works, and you may even learn something about your guns it's not black and white. But really digging into your guns and finding out how they fundamentally work is objectively better. I find the best tool for diagnostics is a 3d animation of the action and comparing the parts you installed vs the ones you took out. They always have some kind of purpose, often it's to shorten the takeup or make the sear and hammer work better together or have a lighter action. Now reading forum posts can help with this, there's a really good one on 1911 extractor tuning (1911forum.com steveinallentown) that's one of the best tutorials I have ever seen about a gun part. But those often don't exist for your particular problem. A chat bot isn't putting you in a situation to understand the problem in most cases, the problem solving that really matters happens in the dark spaces where you can look at a parts diagram and your gun and see that the shortened takeup made it so your firing pin safety wasn't depressed properly (looking at you apex M&P trigger). The chat bot isn't conducive to giving you the space to problem solve because you are looking to the chat bot to give you the ideas rather than filling the space yourself. That's my inherent critique of all of these kinds of things including google, that we are actually pretty creative and competent if we give ourselves the space to be, and the internet in general and especially a plain text chat bot interrupts that creative process that requires time and effort to culture. My critique isn't that chat bots never work, it's that they only help temporarily. Sometimes that's all you want or need but I do see it as inherently problematic.

 

[decius_mus]

Yeah, that is what we did before AI so why not just do that without the extra step that is hurting everyone.

Also is that available for free?

Because it accelerates research time and synthesis dramatically. I don't need to go through multiple google search pages or blindly look through forum threads hoping to find information that I don't even know exists. The tool is like having a research assistant that can find obscure academic papers and many other links that google search wouldn't even populate.

It is available for free but it has usage limits. a $20 monthly ChatGPT plan removes them.

 

[ButteFlyFisher]

It is available for free but it has usage limits. a $20 monthly ChatGPT plan removes them.

It’s free for now while they get everyone addicted

 

[Darqhelmet]

Because it accelerates research time and synthesis dramatically. I don't need to go through multiple google search pages or blindly look through forum threads hoping to find information that I don't even know exists. The tool is like having a research assistant that can find obscure academic papers and many other links that google search wouldn't even populate.

It is available for free but it has usage limits. a $20 monthly ChatGPT plan removes them.

Gotcha, I just don’t have a use case for AI currently. Also i’m not paying for search

 

[Darqhelmet]

It’s free for now while they get everyone addicted

and they continue to enify (is that spelled right?) normal search.

 

[Shellbackitup]

It’s free for now while they get everyone addicted

DING DING DING!

 

[decius_mus]

It’s free for now while they get everyone addicted

Their business model has always worked on a free tier with usage limits and paid plans that remove the limits or enable more advanced models for specific use cases. This isn't some grand conspiracy or the boomer folk tale about a drug dealer giving you free drugs to get you addicted then charging you later.

The goal of the free plan isn't necessarily to get you addicted to AI, it's to onboard you to that specific service's ecosystem (via persistent memory) so that it makes users less likely to switch to a competing service later on. The assumption is that frontier models are here to stay, not that they still need to be broadly accepted. And given AI platform's active monthly user bases, I don't think that assumption is incorrect.

 

[Shellbackitup]

Their business model has always worked on a free tier with usage limits and paid plans that remove the limits or enable more advanced models for specific use cases. This isn't some grand conspiracy or the boomer folk tale about a drug dealer giving you free drugs to get you addicted then charging you later.

The goal of the free plan isn't necessarily to get you addicted to AI, it's to onboard you to that specific service's ecosystem (via persistent memory) so that it makes users less likely to switch to a competing service later on. The assumption is that frontier models are here to stay, not that they still need to be broadly accepted. And given AI platform's active monthly user bases, I don't think that assumption is incorrect.

Oh trust me, entification is planned and built into that business model too.

It's not like the 'free drugs' myth, but the point remains that once the populace is 'hooked', the free tiers will go away as we know them.

Just look at Disney+, Uber/Lyft, or a dozen other examples of 'getting us hooked' and then lowering the service while drastically increasing prices.

I would argue that the active monthly base is majority supported by enterprise and business clients, not private citizens paying out of their own pocket. While there does exist a populace of private users, I'm not arguing that, I don't think there's a lot of widespread adoption of paid models in private use. But that wasn't exactly your point... AI is definitely here to stay. It will be interesting to see in what form.

It's all going according to plan. Well, except for the big public pushback on datacenters and AI, that was kind of unexpected I think.

 

[decius_mus]

Oh trust me, entification is planned and built into that business model too.

It's not like the 'free drugs' myth, but the point remains that once the populace is 'hooked', the free tiers will go away as we know them.

Just look at Disney+, Uber/Lyft, or a dozen other examples of 'getting us hooked' and then lowering the service while drastically increasing prices.

I would argue that the active monthly base is majority supported by enterprise and business clients, not private citizens paying out of their own pocket. While there does exist a populace of private users, I'm not arguing that, I don't think there's a lot of widespread adoption of paid models in private use. But that wasn't exactly your point... AI is definitely here to stay. It will be interesting to see in what form.

It's all going according to plan. Well, except for the big public pushback on datacenters and AI, that was kind of unexpected I think.

All they have to do to enify the service is to increase inference / token limits across tiers. But they still need to balance that with the tiers of other services. If someone happily uses ChatGPT's free tier as it currently sits, and OpenAI decides to reduce free tokens to the point where the service is unusable, the user will likely just migrate to something else... maybe not even a frontier model but an open weight one that can be hosted on their computer. I would argue that the same thing goes for enterprise clients who have options. I also expect sub prices to increase over time, especially considering the major standalone frontier AI labs aren't even profitable yet.

 

[Shellbackitup]

All they have to do to enify the service is to increase inference / token limits across tiers. But they still need to balance that with the tiers of other services. If someone happily uses ChatGPT's free tier as it currently sits, and OpenAI decides to reduce free tokens to the point where the service is unusable, the user will likely just migrate to something else... maybe not even a frontier model but an open weight one that can be hosted on their computer. I would argue that the same thing goes for enterprise clients who have options. I also expect sub prices to increase over time, especially considering the major standalone frontier AI labs aren't even profitable yet.

100%

Competition will definitely raise rates. So will corporate collusion once there is consolidation in the industry.

Just like what happened with Uber/Lyft, the consumers will be slowly boiled frogs as prices increase and competition decreases

 

[rbbansal1970]

So without doxing myself I work for a tech company that does a mix of hardware and software. I’m more on the hardware and algorithm side of things. I’ve toyed with AI quite a bit since it came out.

The good:
If you are already an expert or competent in your domain it’s an accelerator if you give it properly sized tasks. I’m a systems guy, AI is not. I’ve heard “AI helps you cook faster but doesn’t know any new recipes”. Couldn’t agree more.

The bad:
It sometimes misses one little piece of context and then happily drives you down the wrong path. If you aren’t careful and aren’t an expert it can waste a huge amount of time. The times I’ve let AI take the wheel I spent more time debugging it than if I just did it myself in the first place.

The ugly:
It’s really good at sounding right and producing mediocre work. If you aren’t competent in the area that mediocre work gets out into the world, that may or may not be ok.

Spot on. Accurate analyses.

 

[Wolkenflieger]

I wonder how much you living in the bay area colors your view. The rest of the country really does not like the AI companies at all. Across party lines, age, demographics. It is widely hated. Those are the areas that are going to face the brunt of the impact as well as not see the benefit of it.

My self-driving AI robot car has my Grok subscription built-in.

I can task Grok from my car for nav, adding multiple destinations by voice with abstract or natural instruction ("Take me to the best coffee shop in town near a Supercharger") while my car drives itself...all whilst asking science questions or learning a new language with feedback on my pronunciation.

Welcome to the future.

 

[RustVendor]

My self-driving AI robot car has my Grok subscription built-in.

I can task Grok from my car for nav, adding multiple destinations by voice with abstract or natural instruction ("Take me to the best coffee shop in town near a Supercharger") while my car drives itself...all whilst asking science questions or learning a new language with feedback on my pronunciation.

Welcome to the future.

yeah and when someone asks you about where that coffee shop is or what those "science topics" are you will know F- all because you didn't engage with any of it in a meaningful way, you are being spoonfed surface level trash and will eventually only be able to interact and probably date an AI because reality will be too boring and messy compared to a constant pandering fantasy. Its the same reason being rich gives you brain worms, they don't have to work for things they just pay people to do everything and they go insane.

 

[TangoDown]

Just curious what you guys think? I just got downvoted into oblivion on Reddit for the AI graphic. In the past it was fine, but now it seems like there's a disdain for anything AI. I'm noticing the sentiment is changing pretty rapidly and learning that I need to temper some of my own enthusiasm for the tech.

It's really unlocked some technical capabilities I've always dreamed of, since i wasn't a great programmer but working with dev's we've always been able to build some really cool stuff. Now, it's like having a full time dev, BUT ai slop is prevalent.

Curious what you guys think.. I know there's some of you who are in it, but also interested in people who are not and what your thoughts are so I can understand different perspectives better. I recognize I'm biased and interested in peoples opinions.

its only good if you support multi-billionares to get even more money so they can replace you with AI as their workforce so you become jobless and maybe lose your house, your wife and kids, and maybe your dog and be homeless on the street. other than that, you good i guess.