11–19 of 19 posts
It's hard not to trust answers which are superhumanly plausible. Imagine being offered a choice of two oracles. The first oracle would give you an answer that is inspiring yet mysterious and oblique. The second gives an answer that is guaranteed to be unoriginal but also highly, highly plausible. It also maintains a record of all the questions and makes them available to unknown persons.
What? Plenty of humans are competent confabulators.
Not really possible when the technology is built around probability.
It is simply *not* deterministic. The results are based on a roll of the dice. They lack consistency.
Ask the same question in a slightly different way or at different time and get a different answer. And not just different wording but also different meaning.
Would you store food in a refrigerator that might *not* keep it from spoiling? Marketing such a product would create a liability issue --- and I expect the same for AI.
The fact that we're investing $ trillions in such a product will not end well.
should we blindly trust what he says
It's hard not to trust answers which are superhumanly plausible. Imagine being offered a choice of two oracles. The first oracle would give you an answer that is inspiring yet mysterious and oblique. The second gives an answer that is guaranteed to be unoriginal but also highly, highly plausible. It also maintains a record of all the questions and makes them available to unknown persons.