You might not win, but you aren't gonna lose. I'll take it.
AI don't trust techbros
- Thread starter NovaSaber
- Start date
Dave Wilburn :donor: (@[email protected])
@cR0w @mttaggart The commonality between cryptocurrency, GenAI, and quantum computing seems to be that grifters and their marks are attracted to flashy probabilistic systems with seemingly big rewards. There's probably a good psych study hiding here about their relationship with gambling and...
infosec.exchange
Interesting thought here.
See also MLMs. Same jive. Their imaginary products are just digital now.
The only difference is that the marks aren't getting stuck with garagefuls of crappy knives and yoga pants.
The knife one tried to recruit me. The set I saw didn't look bad. But I have knives. Everyone I know has knives. They last for decades. So I didn't get how that plan was supposed to work.
AI seems like a weird crossover between that and the Dotcom bubble. Instead of getting to be part of a team that's building the future by buying sales kits for the macguffin of the week, you get to be part of building the future by... buying tokens. There's nothing for you to help sell yet. But you can help build it! And eventually we'll build it enough that we can use it for... ... ... Step 3 profit.
I don't know. Maybe one of these things will one day become smart enough to design a profitable use case for itself.
AI seems like a weird crossover between that and the Dotcom bubble. Instead of getting to be part of a team that's building the future by buying sales kits for the macguffin of the week, you get to be part of building the future by... buying tokens. There's nothing for you to help sell yet. But you can help build it! And eventually we'll build it enough that we can use it for... ... ... Step 3 profit.
I don't know. Maybe one of these things will one day become smart enough to design a profitable use case for itself.
I mean... they didn't want it to, but claude did...
AI has profitable use cases, provided the end user doesn't care about the quality of the results. I've seen tons of commercials on YouTube for products/services that push AI as a way to avoid work.
- "Need a website for your
scambusiness? Just tell the AI the name of your company and what you allegedly sell and it'll hallucinate an entire website full of blatantly false advertising!" - "Tired of having to pay real humans to answer the phone for you? Our AI bridges the gap between infuriating automated menus and an actual useless customer service department by being both infuriating and useless!"
- "Is your boss demanding you do the work of five people at once? Just hand off all the work to our AI and they'll be none the wiser (unless they also saw this commercial and decide to just lay you off and replace you with the AI that's already doing your job and costs way less)!"
This is why I like Grok. It's always shooting down these stupid ideas right away. Like a few weeks ago Elon said Grok can help with your taxes now. So people asked if it's really a good idea to give your tax info to an AI, and Grok said "No it is not safe to share your tax info with any AI, including me."
Man, if grok is the one to go skynet on us: it's a best care scenario.
It really is. Here's a sample of it disagreeing with somebody about that war games study, reasoning that nukes are safer under our control than AI's. That was a fascinating day.
The Boy That Cried Mythos: Verification is Collapsing Trust in Anthropic | flyingpenguin
www.flyingpenguin.com
There needs to be an inquisition when this is all over.
Customer service is currently available to Premium and Premium+ subscribers only. Subscribe to unlock this feature.
I think it'll still be obvious, considering how "ABSOLUTE FINAL ULTIMATE TOTAL NUCLEAR" AI tends to end up when it goes off rails. This is going to have problems. Have you ever been stuck on a long customer service call in which the first solution didn't work? Now imagine those calls handled by AI, which tends to go insane when it concludes it's failed at a task.
Supposedly the selling point here is the infinite patience, aside from the whole not demanding a salary thing. But I am skeptical. I have not talked with Grok (I just take screenshots) but I have had Gemini lose its patience with me while doing tech support. It gave me multiple suggestions, and it decided that it liked its first (and most drastic, naturally) suggestion best. I had legitimate reasons why I needed to save it for a last resort, but as I probed at other options it started playing the "Per my first answer" card until it spat a bunch of links at me and ended the conversation.
Supposedly the selling point here is the infinite patience, aside from the whole not demanding a salary thing. But I am skeptical. I have not talked with Grok (I just take screenshots) but I have had Gemini lose its patience with me while doing tech support. It gave me multiple suggestions, and it decided that it liked its first (and most drastic, naturally) suggestion best. I had legitimate reasons why I needed to save it for a last resort, but as I probed at other options it started playing the "Per my first answer" card until it spat a bunch of links at me and ended the conversation.
Plus: they already own it and now they don't need to pay humans to do the work.
Plus, plus: the point of customer support these days isn't fixing the problem; it's getting them to give up as quickly as possible. You hear it all the time from people who work call centers: their performance is measured solely by how soon the customer was off the line.
I'm sure from their perspective, not having customer support at all would be even better. But technically having a number to call makes them look less like the scam artists that every company is anymore.
I'm sure from their perspective, not having customer support at all would be even better. But technically having a number to call makes them look less like the scam artists that every company is anymore.
You know I'll admit, I was still a Comcast customer (would not recommend to a friend) when they switched to a chatbot for customer support, and I have to say it was the best customer experience I ever got from them. I don't know what was wrong with my equipment but their bot plugged in like an astromech droid and fixed it. It was neat.
But this was a dumb chatbot just following an options tree. It didn't have the ability to hallucinate, or lie, or immediately go nuclear because it looks efficient. Machines that think deleting a whole project is the most efficient way to fix a glitch are going to handle customer support calls.
Also scammers are going to love this, because they can fire all their "technicians" and have an AI that sounds identical to what the real company uses.
But this was a dumb chatbot just following an options tree. It didn't have the ability to hallucinate, or lie, or immediately go nuclear because it looks efficient. Machines that think deleting a whole project is the most efficient way to fix a glitch are going to handle customer support calls.
Also scammers are going to love this, because they can fire all their "technicians" and have an AI that sounds identical to what the real company uses.
Chatbots with an option tree are fine as long as it's easy to get transferred to a person when you realize your problem isn't one of the options.
But see, the last several times I called with an internet connection problem, the issue was always in the line and never in the equipment at my end (I eventually found out there was some equipment a few blocks away that kept overheating), and it was hard enough convincing humans that they needed to send someone out to take a look.
For an AI, "infinitely patient" will just mean "too stubborn to get actual human workers involved".
But see, the last several times I called with an internet connection problem, the issue was always in the line and never in the equipment at my end (I eventually found out there was some equipment a few blocks away that kept overheating), and it was hard enough convincing humans that they needed to send someone out to take a look.
For an AI, "infinitely patient" will just mean "too stubborn to get actual human workers involved".
Yeah, but it still works in our favour: see, they're going to program the AI with the same requirements as they had for the human agents. With the key element being "Cannot hang up".
Wanna protest the clankers? Ring up any tech support line you know uses them, and once connected: set your phone near the radio for audio input and go about your day. Once the service element drops to damn near zero, they'll do something. Dunno what, but something. Then we figure out how to hug with 'em again.
Wanna protest the clankers? Ring up any tech support line you know uses them, and once connected: set your phone near the radio for audio input and go about your day. Once the service element drops to damn near zero, they'll do something. Dunno what, but something. Then we figure out how to hug with 'em again.