It has become impossible to tell managers mesmerised by artificial intelligence that the tools are not, in fact, helpful. So employees just play along with the fiction to keep their jobs, writes our tech columnist
AI is any compute model performing complex enough reasoning that the output does not always resemble the input. They are by definition non deterministic and the same ask can provide different outputs without being influenced by other inputs.
Everything is guessing, only death is certain. I’ve worked with AI/ML for over 15 years long before folks got all hot and bothered by it. Your examples you give are great because if you walked up to a person and just said 5 you might get as random of answers and you gave.
Assuming by AI you mean LLMs, because that is what is hot on social media today, they use statistical probability models to be able to rate their input to output consistency. It is measurable but could never be 100%. No compute or person is 100%, but an LLM is orders of magnitude less consistent than traditional compute.
I use AI on a daily basis. I trust it less than people, but I also am able to feed it dumb questions which is nice.
AI is any compute model performing complex enough reasoning that the output does not always resemble the input. They are by definition non deterministic and the same ask can provide different outputs without being influenced by other inputs.
So its a lottery
No
I put in a 5 and get a triangle. I put in the same 5 again and get a transvestite. This is intended behavior.
That system isn’t a magic box. It’s a bag of variables linked together by statistics and what’s basically automated guessing.
Everything is guessing, only death is certain. I’ve worked with AI/ML for over 15 years long before folks got all hot and bothered by it. Your examples you give are great because if you walked up to a person and just said 5 you might get as random of answers and you gave.
Assuming by AI you mean LLMs, because that is what is hot on social media today, they use statistical probability models to be able to rate their input to output consistency. It is measurable but could never be 100%. No compute or person is 100%, but an LLM is orders of magnitude less consistent than traditional compute.
I use AI on a daily basis. I trust it less than people, but I also am able to feed it dumb questions which is nice.