Thanks for explaining the 33% hallucination rate. I saw it before and It was quite confusing     See Less

Just because a model has a 33% hallucination rate doesn't mean that 33% of each statement is false. It means that 33% of the responses contain at least one factual error, even if it'     See More t's crucial to understand this distinction when interpreting these metrics.    See Less

Criminalize hallucination     See Less

trump will never give away any UBI...     See Less

Another problem is that unique ideas might not be created anymore. AI copied 1000s of years of human ideas, and if AI will replace humans in many jobs, not much new unique insights or ideas     See More ded anymore. There is no incentive anymore.    See Less

Great topics, really good me thinking     See Less

Consciousness is epiphenomenal, it doesn't really do anything except suffer...

Unconscious AI is just as capable of wiping out humanity and conquering the galaxy.     See Less

I noticed a critical error regarding time and context.

When llama4 launched, I asked common questions related to news events. The latest information if could provide was based on Dece     See More When I asked “who won the 2025 Super Bowl?”, llama4 replied with “that is in the future, so it hasn’t happened yet…”

I pointed out that it is May 2025, and was met with a reply that “time is relative…”
Pointed out that while the statement is true, it is often used by people who do not want to continue the conversation.😂

I shared this conversation with ChatGPT, and it was a fascinating discussion. Running in my desktop, GPT did a better job checking the date using my computer’s time, settings, and searching to fill the gaps.


Today, I asked Gemini about world events, and it told me about a treaty between Japan and Italy to increase defense capabilities. It went on to explain more details, but I had to point out the historical Axis-powers alliances specifically.    See Less

"AI was built on data taken without consent, and it's now reshaping industries without accountability. Artists, writers, and workers never agreed to this—and many are now being pu     See More y tools trained on their own work. If anyone should be responsible to fund a universal basic income, it’s the companies profiting from this massive, unregulated disruption."
- Would it be wrong to characterize it like this?    See Less

AI models would be riskier without hallucinations, because more people would blindly depend on them. The hallucinations make the LLM's more like talking to a smart human being. Sometimes     See More uman being can have a slip of the tongue, or a minor distraction or even incomplete knowledge and honestly provide you with inaccurate information. But if you understand enough about the topic, you will be able to probe further and get the information you really want. If you don't understand the topic, the information would most likely be useless or even dangerous to you.
No matter how accurate the AI models become, I hope I never start taking everything they tell me at face value.    See Less