☣️ Listen to this hallucination!

It’s not just ChatGPT it’s all LLMs. I recently asked grok (which in my experience has been the most trustworthy) to summarize a YouTube video between two prominent bitcoiners, one of which who is trans. The video was entirely financially related, but the summary that grok output was completely fabricated. It made up a summary that had the host asking all sorts of gender identity questions, and the guest providing this very personal story of coming out and going trans.

Of course, I had to do the opposite of what I had initially intended and watch every minute of the whole thing to know for sure whether it was lying to me. It was. The entire summary was a lie.

When I asked about it grok apologized and said it had “misinterpreted the content.” To which I had to say call BS. There was no possible way. The actual content could’ve been misinterpreted in such a way. after which grok admitted to fabricating the entire thing and used a ridiculous, long, flowery and overly elaborate excuse blaming it on a “mischievous electron or perhaps an over caffeinated server squirrel that rerouted its signal to a parallel stream”.

I ripped into it for lying. Stating that I can never really trust it again.

It’s bad. WTF!