Quote:
Originally Posted by soundman98
the comments from experienced robotic/programmers are interesting:
https://hackaday.com/2020/12/30/bost...leon-dynamite/
this seems advanced on the surface, but when one realizes that it's only following a pre-programmed routine because active processing is too taxing, it gets a lot less impressive |
If by active processing you mean the ability to improvise or dynamically respond to the environment then they have done that. The major movements would have been preprogrammed, but the dynamic movements require continuous balance, adjustments and refinements. The nice thing about robots is that they are not constrained by their own body, so depending on distance and bandwidth, as it pertains to latency, there can be central processing at a large supercomputer, and the executions can be remote. The limits in response and smoothness will be balancing central/distancing processing versus local/body processing, which is very similar to the human body in how it moves with regard to the central nervous system--specifically, the interactions between the brain and spinal cord.
We don't need to watch the video and start shaking in our boots, but the rate of advances suggest we are not far away from.
1940-1960 - Computers are huge and basic.
1964 - Prototype modern computer
1976 - Apple I
1984 - Macintosh
1997 - Deep Blue supercomputer beats a grand master in chess.
2001 - iPod debut.
2007 - iPhone debut.
2009 - a chess engine running on a HTC mobile phone attains grandmaster level.
2011 - a super computer named Watson decisively beats Ken Jennings and Brad Rutter in Jeopardy.
2017 - AlphaGo beats the top Go champion
2019 - Google's Sycamore, a quantum computer, completes a task in 200 seconds at 53 qubits, which they claim would take a supercomputer 10,000 years.
2020 - China's quantum computer achieved a processing power of 76 qubits and was 10 billion times faster than Sycamore.
We could do the same type of thing for battery technology development, camera and other sensor development, neural network development, behavior and personality development, speech learning, object recognition, self-learning, etc. Computers and machines have exceeded parts of our abilities a long time ago, and they have exceeded the cognitive limitations of some animals, so I think it is a matter of a few decades--maybe less--until we see weak AI.