AI - Artificial Intelligence or Artificial Ignorance?

[Carolinian]

Having read multiple articles about lawyers using AI to write briefs and being embarrassed by AI making up fake cases and citing them and a law clerk for a judge having the same problem with writing judicial opinions where AI made up cases and statutes, I have wondered about the accuracy of these AI machines. There was also a case of an AI "friend" advising a child to commit suicide and instructing him how to do it, which he did.

In doing internet searches, I usually ignore the unwanted "AI analysis" presented first. This morning I was searching for the website of a state senator running for congress in a nearby district and was curious to see what AI said about him. First, they did not even know he was running for congress, plus they called him a state representative, which he was 6 or 7 years ago. AI was that far behind the curve. That is Artificial Ignorance. Later a did a search on a 1900 Danish 10 kronor, a coin I recently acquired. AI said it was a 90% silver coin, when it is actually a 90% gold coin. AI mentioned a coin designer and I looked that up and it was wrong, too. More Artificial Ignorance.

The real search results followed. On the state senator, the first one listed was the one for his Congressional campaign website, which I was looking for. On the Danish coin, the first one that came up was a Numista article on Christian IX 10 kronors, which included my gold 1900 coin, again exactly what I was looking for. I wish the search engines would not bother sticking that useless AI crap ahead of the real search results.

 

[TolmiePeak]

I just wonder how AI is able to tell facts from fabrications. There is so much garbage out there on the internet. I don't know how AI tools are able to figure out what to regurgitate.

 

[JohnnyO]

I had an argument with Grok AI about Ozzy being dead or not. It took some doing but I finally convinced her after sending her a gootube link of the funeral. She also tried to tell me that high season is low season in Barbados. These are just two examples of the ridiculous info. Grok AI has been helpful with several subjects but I always verify the information with other sources and common sense. It has a long way to go for sure.

 

[billymach4]

It's well understood that AI is not accurate. In my experience it works best on known documented examples and existing proven solutions.

 

[Brett]

.

 

[jp10558]

Having read multiple articles about lawyers using AI to write briefs and being embarrassed by AI making up fake cases and citing them and a law clerk for a judge having the same problem with writing judicial opinions where AI made up cases and statutes, I have wondered about the accuracy of these AI machines. There was also a case of an AI "friend" advising a child to commit suicide and instructing him how to do it, which he did.

In doing internet searches, I usually ignore the unwanted "AI analysis" presented first. This morning I was searching for the website of a state senator running for congress in a nearby district and was curious to see what AI said about him. First, they did not even know he was running for congress, plus they called him a state representative, which he was 6 or 7 years ago. AI was that far behind the curve. That is Artificial Ignorance. Later a did a search on a 1900 Danish 10 kronor, a coin I recently acquired. AI said it was a 90% silver coin, when it is actually a 90% gold coin. AI mentioned a coin designer and I looked that up and it was wrong, too. More Artificial Ignorance.

The real search results followed. On the state senator, the first one listed was the one for his Congressional campaign website, which I was looking for. On the Danish coin, the first one that came up was a Numista article on Christian IX 10 kronors, which included my gold 1900 coin, again exactly what I was looking for. I wish the search engines would not bother sticking that useless AI crap ahead of the real search results.

Generally I feel that AI can be a good search replacement, if you pay for Kagi Ultimate anyway. Free, well it's always worse than paid, and in this case worse in many ways - the free search is generally more product placement to begin with. And garbage in is garbage out as we all know. But free AI is also quite a bit behind paid options, though I'm sure it's debatable. Worse though is the "free, limited AI summaries" that you get on a search engine before the search results. These seem to be the lowest of the low in quality IMO.

What I say is good AI with Internet access that links sources can be a decent search replacement because it tends to filter out irrelevant search results. You can then click through to the sources where you usually at least find something relevant to what you were looking for compared to some search results where they seem to be about something completely different.

All that said, AI is not reliable, and can only go based on what it was trained on or found on the net. So you're compounding issues there. And we all know you need to filter the info on the net to try and get rid of the complete made up crap before AI even gets its hands on it.

I think one thing to pay attention to in many of the people boosting AI is - are they making money from it? Well, that's a big red flag. But even if they're not invested or something - are they using free or paid? Because you just can't compare the different classes of AI in my opinion, they're so different.

 

[Luanne]

I have found quite a few inaccuracies with AI.