4 ms·
A dialog might look reasonable, smart even, but it often contains serious flaws when the questions get harder. This is quite dangerous if people believe the out
by nopinsight 6y ago
A dialog might look reasonable, smart even, but it often contains serious flaws when the questions get harder. This is quite dangerous if people believe the output. Being expressive and being right are quite different.
Note: We might have something that looks very much like AGI quite a while before we do AGI.
-----
Examples:
Me: What do you think of Donald Knuth's literate programming?
"Paul Graham: In general, the concept of literate programming is bad. They're hard to read, and take too much time to write. They are usually unnecessary, because the natural separation between content and implementation is usually fine. Usually, it's better to not look at the code at all and to just read the comment in order to understand."
-----
I asked the exact same question to "Paul Graham" in multiple sessions and always got different answers, many of which are quite problematic:
Me: If you need to pick between a short, but cryptic program and longer, but clearer code, which one would you pick? and why?
"Paul Graham: I would pick the shorter program if it was the same length as the longer program. If they were different lengths, I would pick the shorter one if it contained a bug. I would pick the longer program if it contained a bug. I would pick the shorter program if the longer one contained a bug that had already been fixed in the shorter one."
Another "Paul Graham" with the exact same question:
"Paul Graham: A succinct program is usually the clearer one. Clear code explains its reason for existing, succinct code just exists. For example the following program is not very clear and is hard to follow. for i <- 1 to 1000, j <- 1 to 1000 if i + j < 0 fail with Exit"
-----
I applaud the effort, but this should not be used for serious purposes, like teaching children (or adults).
- blueblisters 6y agoIt' the uncanny valley of NLP. I hope researchers pay more attention to how these answers are being generated. I would prefer interpretable AI over intelligible but random AI.