One of immediate goals of TruthGPT Musk is planning to create is to offset ‘left-wing lies’ in chatbots. He believes large-language models were being trained to be politically correct = untruthful.

Is it technically possible for AI to lie?

Yes. Especially in models that rely on human interaction or that are subject to a central managing entity.

🔎 If it is trained with inaccurate or biased data
🔎 If it generates information that seems plausible but has no real basis (AI hallucinations)
🔎 If it assumes the wrong data is true due to a higher number of incorrect reinforcements (an AI equivalent to a 51% attack)

What is known for sure is that manual fact-checking is everything when working with AI (at least for now). There are many tools created to use AI to fact-check information: AI Fact Checker Tool, Edge browser, Fact Check Tools, Full Fact, FactInsect, Fact Protocol.

Would you rather use...?

🔥 Elon's TruthGPT
🥰 OpenAI's ChatGPT
👨‍💻 Neither

@Coach