GPT-3: Hello, who is the President of Russia?
GPT-3: Are you sure you want to know?
Chatbots: Still dumb after all these years
41–50 of 426 posts
Re: Chatbots: Still dumb after all these years
#42Having 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…
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 points out that the technology doesn't do what the business people think it does just gets dismissed as not understanding.
Re: Chatbots: Still dumb after all these years
#43I really hate when companies hide behind chatbots. Like, NOOOO, your "AI" chatbot didn't help me - get me an actual person.
Kind of a weird tangent but I had a very interesting experience with one of these customer support chatbots recently. My ISP apparently now requires that all customers communicate with their chatbot as the only point of contact. So, naturally, I was pseudo-outraged because I know chatbots are just a money-saving gimmick that reduces workload by driving away 80% of support requests regardless of whether they actually solve someone's problem. And, being technical, my problem was obviously not going to be in the bank of stock answers or even understood by the bot. (I really wish I could remember the actual question)
Long story short, I proposed the question to the chatbot in all its complexity, assuming it would be handed over to a human agent to read the transcript. The chatbot immediately understood the question and provided the exact response I needed.
That was the day I realized I have a deep-seated prejudice against chatbots that blinded me to the possibility that maybe, just maybe, they actually can help sometimes. And I haven't kept up with their technical advancement to be throwing around judgements on their abilities.
To be clear: I'm not arguing in favour of chatbots; just sharing a story.
Re: Chatbots: Still dumb after all these years
#44Having 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…
We want it to be true, so we squint and connect dots and it's true.
I've had numerous people look at the actions of software I wrote over the last 30 years make comments like, "Ohhhhhh it probably did that because it knew x". The software had no such functionality. Seems to be a natural thing for people to be overly optimistic about software's capabilities.
Re: Chatbots: Still dumb after all these years
#45Chatbots are about as useful as phone trees. They can help solve the top 5 easy/common problems, but they are useless passed that. Anyone who has worked in a call center knows that more than half the calls are about the same couple of issues: reset password, finding common help docs, etc. Since help desks are cost centers, it makes sense to half a robot handle as many of these as possible.
I think most of the hate directed to chatbots are because they are really intrusive. You scroll through a page, and 15 seconds in your focus is disrupted as a fake chat window opens up. This is the digital equivalent to the annoying sales rep who asks you if you need help when you are clearly just browsing. The difference is a good sales rep has the intelligence to turn that conversation into a sale. A chatbot usually has to hand off the conversation to a real person to do this. So it has all of the annoyance without the benefit of a potential conversion.
Chat bots as frontends for help desks make sense, but they are poor sales people. If companies learned the difference, I bet their perceived usefulness would change.
Re: Chatbots: Still dumb after all these years
#46My 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…
The self driving people talk about time domain occlusions, which seems closely related to conversation context.
Say you're driving along and you see a pedestrian about to cross but something blocks your view for a moment. Your mental model will remember that person, continue to speculate on their motion and intention, consider if their vector might intersect yours, and also know when it's okay to flush them from attention.
Re: Chatbots: Still dumb after all these years
#47Having 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…
What if that's the case? What if me replying to you is nothing more than some atomically basic mechanism of statistical inference and randomness? Your entire perception of my personhood could be based on a sense of verisimilitude without significant reasoning to set me apart from a transcendant spirit.
What then? Would you be less forgiving of my mistakes if you knew? Would you be less proud of work we collaborate on?
Re: Chatbots: Still dumb after all these years
#48Having 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…
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.
Re: Chatbots: Still dumb after all these years
#49I'm interested in chatbots for learning foreign languages. In that scenario I really don't care how dumb the conversation is, just that it's grammatically correct.
Does anybody know of such a service?
Re: Chatbots: Still dumb after all these years
#50I had to interact with one today.
Garbage company didn’t pick up my garbage. Chat not told me I didn’t have service (I did), then told me I had recently paid them (now I have service?) then told me they couldn’t help me…
Probably one of the most common inquiries and it just failed.