(August 31, 2020 at 7:31 pm)Sal Wrote:(August 31, 2020 at 6:56 pm)Rhizomorph13 Wrote: Good news though, computers never get drunk or tired.
The amount of stuff that can go wrong with machine intelligence is staggering. Any piece of hardware (and software) that increases in sophistication takes comparatively that more safeguards as well as fail-safes along with increasingly more testing and development the more sophisticated it gets. We're nowhere near an A.I. good enough for a fully automated driving vehicle. There are stretches of roads they got down, but places where the real tests are, like unmarked dirt roads or what have you, will make or break all autonomous vehicles.
I reckon we'll be closer to an A.G.I. before then.
In a lot of ways computers are already better than people. If we had more data I would bet that actuarial charts would bear out that the risk is low. The problem is that people have a miss-guided idea that machines are somehow perfect and fail to understand that ANY process has a failure rate. I'm an engineering technician in a research lab at Intel. I know of so many things that can cause havoc in the second level interconnect alone. Add software issues and those problems increase, but the good thing is that redundancies and the shear computational speed can overcome many of those problems.