AI and chatbots

[Maz]

This is nothing about hamsters Just wondered what everyone thinks of all the new technologies. There is a big focus on AI at the moment and it's even in Windows 11 now. Plus just about every company has a "chat" feature if you want to talk to someone. I've had to try and contact about three companies recently and the chatbots have their limitations - it's hard to get to speak to an adviser. One was fairly straightforward if you typed "I'd like to speak to an adviser". Another gave you a list of topics to choose from, none of which were relevant. Typing "other" led to the chatbot asking you to type in a short sentence what you wanted to talk about, and then said - I don't understand that - and the loop started again!

So I'm not sure AI is that clever - it is programmed by humans after all. The ChatGPT is quite interesting though and a bit of an encyclopaedia. You can get all kinds of answers on that.

 

[Daisy]

I actually work partly with AI chatbots, more the ChatGPT type ones than the customer service ones

They can be really helpful for some things:
- Simple factual queries on which there is a widely established right answer.
- Brainstorming ideas on which there is no right or wrong answer, like recipes for dinner, hamster names etc
- Producing outlines or first drafts of essays etc.
- Sometimes creative writing, but they tend to be unoriginal and have a distinctive style which some people will recognise.

But you do have to be careful.

There's a big focus on trying to make chatbots more honest (i.e produce more factual responses), but one thing they are really bad at currently is admitting when they don't know something. If you ask them a question they don't know the answer to, usually they will fill in the gaps by making up something plausible.

Because they are trained based on data available on the web, as mentioned above, if it's a simple factual query where the right answer is all over the web, e.g what's the capital of France, they're always going to get that right. But if it's a topic where there's a lot of bad information on the web or not much information, they're prone to make mistakes. Hamster care is one of those topics on which chatbots would do badly, because there's a lot of bad information on the web, and a lot of the bad information actually comes from "good" sources like vet websites, so the chatbots have no way of knowing it's bad. I had a bit of an argument with Google Bard the other day regarding whether hamsters should be fed seed mixes or pellets. It was absolutely convinced that they should be fed pellets because that's what the "authoritative" sources say.

Another thing to be aware of is that one area where the chatbots absolutely excel is in eloquence. No matter how nonsensical what they are saying is, they will always 100% of the time, sound authoritative and convincing. You know how sometimes in online conversations between humans, you can get a sense for who might be a more reliable source because some people will speak with less conviction than others (i.e I think this might be right but I'm not totally sure, wait for others to respond). AI chatbots will always respond with maximum conviction except in certain specific areas like medical or legal advice where they are trained to provide disclaimers. So you have to look beyond the polished exterior and think carefully about what they're saying.

You're right that AI isn't really that clever. It's a bit like a person who has a lot of general knowledge, but limited intelligence to know how to apply it, and limited common sense. The people who created these chatbots are genuises but the chatbots themselves are pretty average. They're fun to play around with and use for certain applications but I would personally never trust it for factual advice or to produce writing or images for me.

 

[Maz]

I had a bit of an argument with Google Bard the other day regarding whether hamsters should be fed seed mixes or pellets.

I would like to have seen that

Yes there was an issue with some AI trying to be too "woke" and producing inappropriate photos of Nazis and real historical people, but with different skin colour. Which would really confuse a school pupil. I've only mainly used them for contacting companies - some are better than others, but one was really annoying this week - it took me an hour to get to speak to someone at the bank eventually!

 

[Maz]

I didn't even know about Google Bard - off to check it.

 

[Maz]

Apparently Bard has changed his name to Gemini. I couldn't try it because it thinks I'm under 18 for some reason!