Here's a statement of the obvious: The opinions expressed here are those of the participants, not those of the Mutual Fund Observer. We cannot vouch for the accuracy or appropriateness of any of it, though we do encourage civility and good humor.
Reason I posted was I something I came across on another site where somebody wondered if retail investors using AI might be helping stoke the overheated market. Dunno. Lots of other factors. Managed futures etfs are hot now. The ones I’m familiar with seem to chase momentum across many different assets (stocks, currencies, bonds, commodities, etc.) Supposedly they protect to some extent in down markets by running in reverse! AI would seem tailor made for these types of strategies. I’d imagine it’s used a lot by those managers.
Overall Accuracy: Studies indicate that AI search engines provide incorrect answers over 60% of the time.
Best Performing Model: The most accurate AI model, Perplexity AI, still has a 37% error rate.
Worst Performing Model: Elon Musk's Grok 3 chatbot has been reported to be wrong 94% of the time.
Specific Findings
AI Overviews: A survey found that 75% of users experienced significant issues with AI-generated summaries, with 42.1% reporting inaccuracies.
Trust Levels: Only 8.5% of users always trust AI-generated content, while 21% never trust it.
Implications
Bias and Misinformation: AI systems can perpetuate biases and misinformation, leading to potentially harmful consequences in critical areas like healthcare and legal judgments.
Need for Verification: Users are encouraged to verify AI-generated information due to the high likelihood of inaccuracies.
Thank you very much for that, @Derf. Looks like we wont have to worry about 100% accuracy in my lifetime.
How much of that error stems from poor questioning input, however? There is a direct impact in the outcome based on the quality of the input questioning. Apparently, the simple expedient of asking AI to verify its own answers can catch a lot of the errors.
How much of that error stems from poor questioning input, however? There is a direct impact in the outcome based on the quality of the input questioning. Apparently, the simple expedient of asking AI to verify its own answers can catch a lot of the errors.
"Poor questioning input" was the problem even before AI. lol
That's pretty interesting. Looking ahead, let's propose that the bugs get worked out of AI, and a basically reliable information source becomes widely available to almost anyone. What sort of impact would that have on financial markets?
OJ, I always thought that once the average person had access to the WWW, they would become more formidable intellectually.
If you think too much about this it can drive you nuts. OK, so above I asked about the accuracy of AI answers, and the response, from some type of AI, was "not so good".
The thing is, how do we know what the accuracy is of the answer that I received? What if that answer has only a 50/50 chance of being right?
Comments
General Accuracy Rates
Overall Accuracy: Studies indicate that AI search engines provide incorrect answers over 60% of the time.
Best Performing Model: The most accurate AI model, Perplexity AI, still has a 37% error rate.
Worst Performing Model: Elon Musk's Grok 3 chatbot has been reported to be wrong 94% of the time.
Specific Findings
AI Overviews: A survey found that 75% of users experienced significant issues with AI-generated summaries, with 42.1% reporting inaccuracies.
Trust Levels: Only 8.5% of users always trust AI-generated content, while 21% never trust it.
Implications
Bias and Misinformation: AI systems can perpetuate biases and misinformation, leading to potentially harmful consequences in critical areas like healthcare and legal judgments.
Need for Verification: Users are encouraged to verify AI-generated information due to the high likelihood of inaccuracies.
Thank you very much for that, @Derf. Looks like we wont have to worry about 100% accuracy in my lifetime.
Hence the phrase, "Ask a stupid question...".
We have all seen how that turned out.
The thing is, how do we know what the accuracy is of the answer that I received? What if that answer has only a 50/50 chance of being right?