AI? LMAO...We're Not There Yet

[wagtunes]

This has nothing to do with music but it illustrates how stupid AI is.

I had a transaction declined through PayPal. That's what the email said. But it didn't tell me which transaction. So I called Paypal to find out.

The AI assistant asked me what I needed and I said "Question about declined transaction."

It proceeded to tell me what would cause a declined transaction. I then told it "I need to find out which transaction was declined."

It then proceeded to tell me the same thing again. So I said "I need to speak to an agent." It then asked me why. I told it again, "I need to find out which transaction was declined."

It then proceeded to tell me AGAIN what would decline a transaction. And the loop began again. Finally, realizing I wasn't going to get anywhere, I lied to it and said "I need to find out how to open a new account" to which the AI assistant finally transferred me to an agent because it could no longer help me.

AI is stupid, even when you CLEARLY tell it what you need. If it hasn't been programmed to give you what you want, it gets stuck and you get stuck in an endless loop.

My problem should have been a basic thing that the AI should have easily handled. It couldn't because it's stupid.

The bigger point is this. If you're worried about AI taking over the music industry and possibly even destroy the world, it couldn't even send me to an agent when I told it I needed to find out why a transaction was declined.

We are not there yet folks.

And from what I've seen today, we may NEVER get there.

[TheMaestro]

[ghettosynth]

The purpose of AI agents in support is to reduce costs. It really isn't about it being dumb, per se, it's about controlling costs and one way to do that is to not commit much compute to the agent. The ideal resolution for your concern is that you give up and just let it go so that they don't have to provide any actual support. I'm surprised that there aren't more recordings of people having fun with these agents. "Are you AI, because you sound very smart."

[wagtunes]

The purpose of AI agents in support is to reduce costs. It really isn't about it being dumb, per se, it's about controlling costs and one way to do that is to not commit much compute to the agent. The ideal resolution for your concern is that you give up and just let it go so that they don't have to provide any actual support. I'm surprised that there aren't more recordings of people having fun with these agents. "Are you AI, because you sound very smart."

Well if that's true then shame on them.

[vurt]

i have a camera with ai recognition for animals.
so far, ive had a pigeon described as a human, a robin was a cat and a magpie was a "dog" not even a specific species, just "dog".

ai my arse.

[BertKoor]

@wagtunes: While annoying, pointless, and total waste of time, what you have described has nothing to do with AI. It exists in this form well over a decade. Robotic client "service" at its worst. It likely also refers you to the website. Plain stupid. If you could find it on the website then you would not need to reach for the phone.

[wagtunes]

@wagtunes: While annoying, pointless, and total waste of time, what you have described has nothing to do with AI. It exists in this form well over a decade. Robotic client "service" at its worst. It likely also refers you to the website. Plain stupid. If you could find it on the website then you would not need to reach for the phone.

Well it's called an AI assistant.

[el-bo (formerly ebow)]

i have a camera with ai recognition for animals.
so far, ive had a pigeon described as a human, a robin was a cat and a magpie was a "dog" not even a specific species, just "dog".

ai my arse.

Stool pigeon?

[ghettosynth]

@wagtunes: While annoying, pointless, and total waste of time, what you have described has nothing to do with AI. It exists in this form well over a decade. Robotic client "service" at its worst. It likely also refers you to the website. Plain stupid. If you could find it on the website then you would not need to reach for the phone.

Well it's called an AI assistant.

Don't believe everything that agents, computer or otherwise, tell you. They are often limited in what they can do, on purpose, to reduce risk to the organization. There are fun stories about when this breaks. In any case, expect more of this going forward. You are encountering cost cutting, not the cutting edge of AI.

[wagtunes]

@wagtunes: While annoying, pointless, and total waste of time, what you have described has nothing to do with AI. It exists in this form well over a decade. Robotic client "service" at its worst. It likely also refers you to the website. Plain stupid. If you could find it on the website then you would not need to reach for the phone.

Well it's called an AI assistant.

Don't believe everything that agents, computer or otherwise, tell you. They are often limited in what they can do, on purpose, to reduce risk to the organization. There are fun stories about when this breaks. In any case, expect more of this going forward. You are encountering cost cutting, not the cutting edge of AI.

Then again, shame on these people. Especially Paypal which is arguably one of the biggest payment processing services in the world.

[ghettosynth]

@wagtunes: While annoying, pointless, and total waste of time, what you have described has nothing to do with AI. It exists in this form well over a decade. Robotic client "service" at its worst. It likely also refers you to the website. Plain stupid. If you could find it on the website then you would not need to reach for the phone.

Well it's called an AI assistant.

Don't believe everything that agents, computer or otherwise, tell you. They are often limited in what they can do, on purpose, to reduce risk to the organization. There are fun stories about when this breaks. In any case, expect more of this going forward. You are encountering cost cutting, not the cutting edge of AI.

Then again, shame on these people. Especially Paypal which is arguably one of the biggest payment processing services in the world.

Dude, you sound like this is news to you. It has always been this way. Outsourcing to humans reading scripts with no authority to solve your actual problem without escalation was a pre-AI response. I'm not defending their behavior here, but firms are going to compute profits, that's what they do.

[wagtunes]

@wagtunes: While annoying, pointless, and total waste of time, what you have described has nothing to do with AI. It exists in this form well over a decade. Robotic client "service" at its worst. It likely also refers you to the website. Plain stupid. If you could find it on the website then you would not need to reach for the phone.

My problem with the whole thing is the waste of my time. At least prior to these automated answering bots. I got a real person who at least handled my call right away even if they couldn't help me. Now I have to argue with a machine for God knows how long just to get to talk to somebody. That's my gripe. Not the fact that it can't help me. Just don't waste my time.
Well it's called an AI assistant.

Don't believe everything that agents, computer or otherwise, tell you. They are often limited in what they can do, on purpose, to reduce risk to the organization. There are fun stories about when this breaks. In any case, expect more of this going forward. You are encountering cost cutting, not the cutting edge of AI.

Then again, shame on these people. Especially Paypal which is arguably one of the biggest payment processing services in the world.

Dude, you sound like this is news to you. It has always been this way. Outsourcing to humans reading scripts with no authority to solve your actual problem without escalation was a pre-AI response. I'm not defending their behavior here, but firms are going to compute profits, that's what they do.

[zerocrossing]

@wagtunes: While annoying, pointless, and total waste of time, what you have described has nothing to do with AI. It exists in this form well over a decade. Robotic client "service" at its worst. It likely also refers you to the website. Plain stupid. If you could find it on the website then you would not need to reach for the phone.

My problem with the whole thing is the waste of my time. At least prior to these automated answering bots. I got a real person who at least handled my call right away even if they couldn't help me. Now I have to argue with a machine for God knows how long just to get to talk to somebody. That's my gripe. Not the fact that it can't help me. Just don't waste my time.
Well it's called an AI assistant.

Don't believe everything that agents, computer or otherwise, tell you. They are often limited in what they can do, on purpose, to reduce risk to the organization. There are fun stories about when this breaks. In any case, expect more of this going forward. You are encountering cost cutting, not the cutting edge of AI.

Then again, shame on these people. Especially Paypal which is arguably one of the biggest payment processing services in the world.

Dude, you sound like this is news to you. It has always been this way. Outsourcing to humans reading scripts with no authority to solve your actual problem without escalation was a pre-AI response. I'm not defending their behavior here, but firms are going to compute profits, that's what they do.

All of that is true, but AI agents are a step above a human copy-pasting a script into a support email. The company my wife works for has replaced their entire Level 1 support team with AI agents, and the customer satisfaction surveys have shown that people are actually a lot more satisfied. Of course, these agents were specifically trained on past successful customer service interactions, the software implementation documentation, features, and it’s all baby sat by humans to make sure that it’s not hallucinating. They expect to be able to stop the babysitting at some point, but level 2 support still needs to be done by a human.

I’ve also experienced an AI agent used by the company who did our HVAC and water heater work last year. It was mostly good. When it didn’t know what to say, it apologized and had a human get back to me.

The most recent bad experience was when Positive Grid’s customer service bot told me to increase the oversampling to help reduce aliasing… and there is no way to do that. When I pointed this out to the bot, it pretended to be a human and apologized. I think the issue is that they haven’t properly trained the agent on their software and do no monitor it properly.

[keys_au1]

beware hubris! AI is on an exponential curve not a steady one!

[ghettosynth]

All of that is true, but AI agents are a step above a human copy-pasting a script into a support email. The company my wife works for has replaced their entire Level 1 support team with AI agents, and the customer satisfaction surveys have shown that people are actually a lot more satisfied. Of course, these agents were specifically trained on past successful customer service interactions, the software implementation documentation, features, and it’s all baby sat by humans to make sure that it’s not hallucinating. They expect to be able to stop the babysitting at some point, but level 2 support still needs to be done by a human.

I’ve also experienced an AI agent used by the company who did our HVAC and water heater work last year. It was mostly good. When it didn’t know what to say, it apologized and had a human get back to me.

Yep, but, wags is complaining about what's not working, and that is a familiar pattern. There is a structure to problems where this works at low cost and where it is likely to be problematic.

The most recent bad experience was when Positive Grid’s customer service bot told me to increase the oversampling to help reduce aliasing… and there is no way to do that. When I pointed this out to the bot, it pretended to be a human and apologized. I think the issue is that they haven’t properly trained the agent on their software and do no monitor it properly.

I'm not surprised. That sounds like general advice, which is not intrinsically wrong. The apology part though is also not surprising. Here's a pro top, ignore apologies from AI, they don't mean anything more than "please user."

I don't disagree BTW, that some support bots are useful. However, my primary point that they are generally not empowered to incur liability is key, whether they are useful or not.