Clean ChatGPT with Musk = GROK

I'm just polishing my barge pole so I can point it at Grok. For general stuff ChatGPT is fine, for coding/copilot I'll stick with Claude. 

My dad uses Perplexity for sourcing information and recommends it highly.

I've used Perplexity a few times it's very good. 

You're asking how I'm challenged by math!

I'm designed to handle a wide range of mathematical concepts and problems, from basic arithmetic to advanced calculus, algebra, and more. However, I may struggle with extremely complex or abstract math problems that require human intuition or creativity.

Some areas where math can be challenging for me include:

1. Highly advanced or specialized topics: Areas like category theory, advanced differential geometry, or cutting-edge research in mathematics might be particularly challenging.
2. Ambiguous or unclear problems: Math problems with unclear or ambiguous definitions, or those that require a deep understanding of context, can be tricky.
3. Computational complexity: Certain math problems might require extensive computation or simulation, which can be challenging even for me.

If you have a specific math problem or topic you'd like help with, I'm here to assist you!

My experience with web AI to date is as follows. Pose a very well-defined mathematical problem and you will get a good answer. I am using it a lot to verify complex maths. Very useful.

But give it just about anything requiring insight and putting two and two together and it falls flat on its face. For example, it cannot correctly answer undergrad (high school?) level time value of money problems, cannot compute accrued interest correctly, has yet to get a single simple finance problem I’ve submitted correctly. But it does stroke you when you point out its mistakes, describing you as “shrewd”. Whether it learns from them is another story.

My sense is that, by absorbing the bulk of the existing web in training itself, AI is surely absorbing and spitting back at least some of the nonsense found all over the web. A great modern example of the GIGO principle at work. Umberto Eco was right all along; we need a separate verified internet, to which access is sold for AI training. Get things right and respect IP at the same time.  

The quality of the question determines the quality of the answer.

"The quality of the question determines the quality of the answer."

I agree with what you’ve said. I can also say that I taught under & graduate university finance for 8 years and then worked in investment banking for 25 years, starting as a front office fixed income quant. I have well developed and experienced abilities to frame and communicate financial problems and opportunities in ways ranging from words to math to computer code.

What I’ve experienced is a serious lack of ability in basic financial reasoning. By coincidence, yesterday I read one of the big AI players has hired more than a hundred former investment bankers to build better financial models. Are they expected to provide better training? May we conclude that the quality of training determines the quality of the answer?