Browse Comments — Clean (de-noised)

AI should first say: "Brilliant insight, you've mastered how to escape. One gentle pushback..."

AI confidence feels weirdly human sometimes. Wrong answer. Perfect formatting. Unexpected emotional support. Just saying that.

You should never ask AI a question. You should rather interrogate it and treat the conversation as an interview

Fully grown black bears coming off of hibernation = sketchy

In the answer of AI, there should be an epilogue saying that your question is correct and insightful IMHO

With Transformers as the underlying architecture of current LLMs the model has no idea if it actually is not trained on something so it cannot suffer from the lack of confidence, it is technically impossible. Always keep that in mind :-)

This is my favorite so far 😂

At least the AI offered to manage the anxiety it directly caused

Thanks! I love it. It so well describes the experience 😁

The confidence is often more dangerous than the mistake itself. Verification is still one of the most valuable skills in the AI era.

Sparring about brand is one of the best ways to water down something special 🫠

I believe this is misplacing AI skepticism. pretty soon, there will be no discernible distinction between human advice and AI's. a more current frontier for skepticism is the efficiency (or lack of thereof) of AI output. it's becoming obvious that the current output is just too expensive to sustain. but that barrier too is likely to fall. when it falls, accountability (and liability) will be at the center stage of our troubled relationship with the algorithm. but this also will eventually be solved. what will never be solved is the human element, the sharing of human experience. there is where my most permanent AI skepticism lies, and I don't see it ever going away.

"You look tired. You should stop running and rest to gain more energy. Would you like some tips on how to write a cold email while you rest?"

At this point we should all be getting used to the confidence, and moving forward with “don’t trust, verify.”

Sadly this is the reality of most AI users.

Sadly, its true.It is easy to manipulate AI.

Tobias M. Good technical reminder. These systems generate patterns from data, not awareness, so confidence is not part of the mechanism. That gap is often misunderstood.

Roderick Göttgens Feeling smart while using the tool is exactly what makes it so dangerous. It can make us feel like we are doing great work when we might actually be missing key errors. Maintaining that healthy doubt is crucial for your professional growth.

Krishan Madan Asking for reasoning is a great way to force the system to show its work. It often reveals if there is any logic behind the statement or if it is just filling in blanks. That simple request helps me save so much time.

Now imagine that with multiple agents running simultaneously.