3 ms·

>Factual correctness is not very important, you don't have to trust it. If these chat bots are going to replace or compete with the traditional search interfac

>Factual correctness is not very important, you don't have to trust it. If these chat bots are going to replace or compete with the traditional search interface, then correctness is absolutely key. When a traditional search engine points you to some spammy site, you have far more context to decide for yourself how much credibility you ascribe to it. You may know the site from earlier visits. You can see whether it's completely ad infested and what the likely business model is. You can see what else is on the site. You can check who the publisher is, etc. But when a chat bot delivers results in a conversational style, there is absolutely no indication as to whether a particular claim is fact or fiction. If the bot provides links to sources that's great, but then you're back to square one reading through 10 paragraphs of blog spam trying to filter out the one bit that may answer your question. Also, the way in which these bots fail compared to traditional content farms and spammy blogs is very different. These sites tend to copy stuff from elsewhere. They may repeat falsehoods published by others. But they are unlikely to just make up numbers and present them as fact.

mrtksn 4y ago
I don't think its going to be drop-in replacement. With ChatGPT, you can ask follow up questions. So if something doesn't sound right, you can pressure it to explain its position. It's like talking with people, sometimes people are wrong and we happen to have a lot of linguistic tools to deal with it. Like, follow up questions or asking for specifics.
- fauigerzigerk 4y ago
  >I don't think its going to be drop-in replacement. I agree, but part of the reason for that is that it doesn't get the facts right (yet?). A chat bot that can be trusted would be far more useful and could largely replace the traditional search interface. I tried ChatGPT (and some other transformer models). I know you can ask follow up questions and make it reconsider its responses or change its mind. But why does it change its mind? It's not because it finds a flaw in its own logic or discovers new information. I think it's simply because the pattern of inquisitorial questioning is statistically associated with the pattern of mind changing. And then it just follows your lead in a new direction making claims that are just as unreliable as the original ones. I'm keeping an open mind though. I don't think I have understood everything there is to understand about this.

mrtksn 4y ago

I don't think its going to be drop-in replacement. With ChatGPT, you can ask follow up questions. So if something doesn't sound right, you can pressure it to explain its position. It's like talking with people, sometimes people are wrong and we happen to have a lot of linguistic tools to deal with it. Like, follow up questions or asking for specifics.

fauigerzigerk 4y ago
>I don't think its going to be drop-in replacement. I agree, but part of the reason for that is that it doesn't get the facts right (yet?). A chat bot that can be trusted would be far more useful and could largely replace the traditional search interface. I tried ChatGPT (and some other transformer models). I know you can ask follow up questions and make it reconsider its responses or change its mind. But why does it change its mind? It's not because it finds a flaw in its own logic or discovers new information. I think it's simply because the pattern of inquisitorial questioning is statistically associated with the pattern of mind changing. And then it just follows your lead in a new direction making claims that are just as unreliable as the original ones. I'm keeping an open mind though. I don't think I have understood everything there is to understand about this.