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

Chatbots: Still dumb after all these years

mindmatters.ai

Post reply on HN
Print view View original

111–120 of 426 posts

Re: Chatbots: Still dumb after all these years

#111

Post by mrweasel »

arthur_sav wrote: post #5

I really hate when companies hide behind chatbots. Like, NOOOO, your "AI" chatbot didn't help me - get me an actual person.

Bots do not serve ANY purpose in most interaction with people. They are capable of a limited set of tasks and should be used carefully. Mainly they piss people off, and if a bot could handle the interaction, so could a website.

Previously I worked for a company that took pride in not being like the big players, doing this the right way, but apparently that has fallen completely apart. I know it's not the same thing exactly, but it made my a little angry to see some of their web pages having a text saying: "Blip Blop, I'm a tiny bot and I've translated this page. I don't always get it right, but I'm learning". Just leave the reasonable English version or do a proper translation, this automated crap, and don't try to excuse bad translation and messy language with "I'm learning". If you KNOW the bot makes enough mistakes that you have to let people know, then maybe it's not ready yet.

Re: Chatbots: Still dumb after all these years

#112

Post by the_only_law »

gregmac wrote: post #87
waynesonfire wrote: post #27

Earlier quoted context omitted.

> Still doesn’t piss me off as much as automated phone support systems. Yes, the AI voice bot is marginally better because I can request "customer service" without waiting to discover the right numeric code. That's about the extent of that.

I accidentally discovered a cheat code a few years ago, interacting with one of those IVR voice systems, as I was getting frustrated with it and eventually exclaimed "fuck!" -- its response was brilliant: "Alright.. ok.. it looks like you're having troubles. Please hold while I transfer you to a human operator."

I’ve had that work a few times, but other times, I can verbally abuse the bot all I want to no avail.

Re: Chatbots: Still dumb after all these years

#113

Post by p1esk »

pwinnski wrote: post #13

Having worked in ML at two different companies now, I think that people interpreting model output as intelligence or understanding says much more about the people than about the model output. We want it to be true, so we squint and connect dots and it's true. But it isn't. It's math and tricks, and if human intelligence is truly nothing more than math and tricks, then what we have today is a tiny, tiny, tiny fraction…

As far as chatbots, we went from "tricks" to "math" in just ~10 years. Yes, still as dumb, but the underlying "technology" is very different. GPT-3 is a lot closer than ELIZA to how our brains do it.

Re: Chatbots: Still dumb after all these years

#115

Post by mrweasel »

zerkten wrote: post #108
pydry wrote: post #29

Earlier quoted context omitted.

They could do a better job of separating the two. Most companies would save their time and mine if they had a callback system which took as many details as possible up front & didnt have to ask my name and account details because Im already logged in. Even better if it could give me some warning about where I am in the queue and when the call might be coming (e.g. via push notification). I'd 1000x prefer that over an…

> They could do a better job of separating the two. When support is purely viewed as a cost, then this will never happen. 99% plus of your call volume may be for the obvious things. If you offer two options all of those people are going to be more confused than they already are and you will have some of them engaged by costly humans. > Unfortunately it seems call centres are driven by very traditional metrics which w…

> When support is purely viewed as a cost, then this will never happen. 99% plus of your call volume may be for the obvious things.

Maybe, but not my experience. I worked at a telco, and as developers we had to sit in on support calls a few times, to help identify areas that could be improved with minor effort. The majority of the calls I listened into on a given day had to be assigned to an engineer. The remaining, they just wanted a better deal or help reading their bill.

Re: Chatbots: Still dumb after all these years

#116

Post by Animats »

And that's a GPT-3 chatbot. Commercial customer service chatbots are worse. They're usually just phone trees with delusions of grandeur. That's underperforming.

Is there any known customer service chatbot that really has a good model of what's going on? That ought to be possible for banking, travel, and product ordering, at least. Something where, each time it has to refer someone to a human, collects data to get better.

Re: Chatbots: Still dumb after all these years

#117

Post by jhbadger »

hotpotamus wrote: post #23
klyrs wrote: post #12

My favourite flaw of chatbots is exposed by ELIZA. Not chatting with ELIZA, (though, it does suffer this flaw) but using responses inspired by that program. "Please elaborate on that" or "tell me more about [noun]" etc. Bots appear to have zero lines of short term memory, and utterly fail to pick up a reference to the thing that they just said. My favorite being bot> [something plausibly human-sounding] me> What do y…

I'm not trying to be mean or break any HN rules, but did you read the article? It basically covers what you asked and was quite a revelation to me. Others in here point out that GPT-3 is not a chatbot which is good info, but I also wonder if there is anything out there that can even "remember" the context of a conversation it's supposed to be having. Particularly interesting is the question from the article, "who is…

>who is the current president?

I wonder what happened to IBM Watson's technology that was good at answering trivia -- it actually won a game of Jeopardy! a decade ago (although of course it answered things with a question as per the rules). I know that they weren't that successful at applying it to biomedical research as they had hoped, but it would seem it would be better at chatbots than GPT-3 and other deep-learning models.

Re: Chatbots: Still dumb after all these years

#118

Post by chubot »

p1esk wrote: post #113
pwinnski wrote: post #13

Having worked in ML at two different companies now, I think that people interpreting model output as intelligence or understanding says much more about the people than about the model output. We want it to be true, so we squint and connect dots and it's true. But it isn't. It's math and tricks, and if human intelligence is truly nothing more than math and tricks, then what we have today is a tiny, tiny, tiny fraction…

As far as chatbots, we went from "tricks" to "math" in just ~10 years. Yes, still as dumb, but the underlying "technology" is very different. GPT-3 is a lot closer than ELIZA to how our brains do it.

It was more than 10 years ... People reacted the same way to SHRDLU 40+ years ago, and 20 years ago when I encountered it:

https://en.wikipedia.org/wiki/SHRDLU

There's definitely an illusion where we anthropomorphize very simple software, ascribe intention to it, etc.

Re: Chatbots: Still dumb after all these years

#119

Post by rchaud »

Animats wrote: post #116

And that's a GPT-3 chatbot. Commercial customer service chatbots are worse. They're usually just phone trees with delusions of grandeur. That's underperforming. Is there any known customer service chatbot that really has a good model of what's going on? That ought to be possible for banking, travel, and product ordering, at least. Something where, each time it has to refer someone to a human, collects data to get bet…

I'm not sure what I dislike more: the cust. service chatbots or the sales chatbots.

Re: Chatbots: Still dumb after all these years

#120

Post by numpad0 »

melony wrote: post #48
version_five wrote: post #42

Earlier quoted context omitted.

I think a problem is the tighter cycle between academic discoveries and business people trying to monetize them. Large language models were developed, objectively a great achievement, and immediately someone nontechnical wants to apply their own interpretation and imagine that we can build a chatbot that you won't have to pay, and before you know it, people are selling and even deploying them. Anyone who questions or…

That's a good thing. The market is very good at empirically validating research. It helps filter out the chaff and ensure research is useful. Better than academia wasting years on concepts like the "semantic web" where nobody can even agree on what it means. Academia is under attack from many political directions right now, being able to show useful output will help it thrive in the long run.

The problem with the market is it wants consistent, short-term successive, quantifiable production, and that seems to instantly kill all progress in a field of research. They just go laid back watching investments burn and prints worse and worse nonsense.
Post reply on HN