The probability of the tiger actually being able to touch you are zero. The atoms will simply get close but then just hover next to you. The tiger just wants to cuddle :3
To the ppl who think chatgpt and other LLMs will create terminator style robots, you can make the argument that LLMs are just a text prediction engine and incapable of that.
However, you can't use that argument to say LLMs are harmless. They are capable of other types of harm, mostly when users blindly trust what LLM's output.
It's less ridiculous when you realize the human brain is also a prediction engine. It can just operate in a wide variety of ways instead of just being limited to only talking, or only folding proteins, or only playing chess or whatever.