What am I missing here? Seems like an AI detection program would be as useful as a program which could tell whether John Henry used a sledge hammer or a steam hammer to lay a section of track.
Or whether your CPA did your taxes using Excell or just a big piece of graph paper and counting on his fingers.
What we should be doing is making tools which can tell when an AI is hallucinating, or the output doesn't really solve the problem the user wants to solve.
What am I missing here? Seems like an AI detection program would be as useful as a program which could tell whether John Henry used a sledge hammer or a steam hammer to lay a section of track.
Or whether your CPA did your taxes using Excell or just a big piece of graph paper and counting on his fingers.
What we should be doing is making tools which can tell when an AI is hallucinating, or the output doesn't really solve the problem the user wants to solve.
we have tools for that, just use a different model to check the work of the other :)