ChatGPT-aided Bing makes mistakes, lies to excuse them, and is a serious threat to websites

[highplainsdem]

Last edited Thu Feb 9, 2023, 12:37 PM - Edit history (1)

Oh, and yes, there is the possibility of AI "somehow going wrong in a way that is lights-out for humanity." See the section of the video starting at 6:02.

Now, to the more immediate problems...

As far as the mistakes it makes and how it can offer a glib lie if caught, see the story below, which was on CBS Mornings yesterday morning - especially the part starting about 3:05.

CBS called this "making things up" - not a lie. And I suppose it could be argued that it wasn't an intentional lie, but just another mistake.

But do you really want AI that will not only get things wrong but make things up to excuse any mistake it's caught in?

Re what the thread title says about AI in search engines being a serious threat to websites - which is my take on it, based on decades online:

The sources of the information being provided are not given - whether they're books or websites. If they're websites, links to that website aren't provided by the AI chat version.

So, for instance, if you ask for a recipe for a particular dish, ChatGPT will give you one - but not the website or websites it took the recipe from, let alone websites with interesting variations that will turn up if you do a regular search. You won't learn of new websites you might want to look at more. You won't see the comments from users of that website, giving their opinions of the recipe and tips and variations. You won't be offered a place for your own comments and variations.

And those websites won't be getting the traffic, ad revenue and reader interactions they need to survive.

Instead, every bit of work that was put into the website will just have been ripped off by AI to offer simplified answers with no credit given to anyone.

And as for ad revenue... Does anyone here believe that in a capitalist society, companies will not hope to make their own products and services more visible via these AI-assisted search engines? And do you really think whoever is behind the search engines will be able to resist money being offered? Or will make sponsorship completely visible to the reader/consumer? AI-assisted search engines will be the ultimate in social media influencers.

[Scrivener7]

1. I think ChatGPT is the devil.

[highplainsdem]

3. It's potentially very dangerous. And your reply reminded me of what was said

about the potential threat near the end of this video, which I just edited the OP to mention in the very first sentence.

I caught this news story on CBS yesterday morning, knew I wanted to post about it here, but not when the SOTU and response were the focus. I didn't watch it again before posting this OP, other than to locate where the discussion of mistakes started, and I'd forgotten that warning about AI in general - which is similar to warnings we've seen over past decades, including from science fiction, but striking in this context.

[FakeNoose]

2. It's going to mean the end of slander, libel and defamation enforcement

Any publication that gets sued for libel from now on will say, "Oh it was the chatbot! We didn't write that."

[CousinIT]

4. Related...

https://www.democraticunderground.com/100217631107

[highplainsdem]

6. Just saw that and recced it before I saw your reply here. Thanks for

posting the OP and this reply and link!

[Silent3]

5. "makes mistakes, lies to excuse them, and is a serious threat"

Clearly they used too many Trump tweets in the training data.