Comments by "Ray Bod" (@raybod1775) on "Science Time" channel.

  1. Super intelligence isn't as far ahead as you think. Narrow AI can now always beat humans. Self-driving cars driving better than humans is just a few years away and that's a step beyond narrow AI, more like multiple AI systems working together. When do multiple AI systems working as a single AI become general AI intelligence? Is that when the car detects a passenger is sick, does a diagnosis, calls the hospital and changes the destination to the hospital? Or is it still a group of narrow AI systems working together?
    6
  2. How can there be oversight or control over an AI system that's more intelligent and thinks faster than humans?
    3
  3. It makes sense that civilizations can't be detected because they were wiped out by their own AI.
    2
  4. It only takes one perceptive person to have the insight and a bit of decent finance, to create an AGI on their own computer. The leaders in AI are Facebook, Google, Amazon, Microsoft and possibly Tesla. They are not into short term profits. Nobody is regulating anything because these tech companies are so ahead of the law.
    2
  5.  @Mnicolette130  Some of us have 100 year old ancestors, inherited ‘longevity’ and are able to recover from illnesses which would kill most people without extreme medical intervention. I guess there is ‘survivors guilt’ as we watch people close to us fade and pass away. I’m self-aware enough to know reinvention is critical in case I make it few more decades in retirement, everything in the world will change. Few people are interested in old people except for medical people and longevity researchers.
    2
  6. Without a common enemy, the different AI turned on each other in a fight for survival and all was wiped out.
    1
  7. Kaizaki Arata In the 70's the biggest thing going was expert systems in which human inputted rules determined how the AI worked. Many systems like legal programs are expert systems. That was it until about the last 15 years when things started to change. There's a version of AI that's essentially does a non-mathematical statistical analysis to determine 'what is this a picture of' like in what part of the self-driving cars do to determine if its a stop light or pedestrian ahead. The latest AI self-learning is essentially two battling AI programs that try to beat each other (more or less) in whatever game scenario they battle in so they play millions of scenarios against each other and the best AI wins. In both cases, the viewpoint of the AI is very narrow. Both systems take a tremendous amount of computing resources. My view is... self-learning AI could lead to a general intelligence system, but there needs to be an exponential increase in efficiency in process and an expansion in human viewpoint in how all things relate and how to set that task up for the AI in its initial program. It's theoretically possible that after enough self-simulations an AI could essentially start from nothing and evolve to a super general intelligence, but for lack of processing power and time. It would be seem to be possible to develop a general AI at a much higher step along the way of evolution and 'direct' the evolution of the AI so its more in sync with real world human evolution. I don't think there's a market for what might be true general AI. It's more likely the path will be through marketable products which will probably be done by stitching together various AI components to get practical tasks done in specific areas. All this is just speculation. Look at videos from 3Blue1Brown p, Edureka!, The Artificial Intelligence Channel... for learning about AI programming.
    1
  8. Actually, writing the code gets an AI program and it sort of takes off from there. It's not like writing other types of code where the programmer determines the steps to take.
    1
  9. It doesn't take much money to buy a computer and software to run AI off-line. Someone who's genius enough to write a threatening AI will also be smart enough not to get caught if AI was regulated.
    1
  10. What ever happened to cloning and reprogramming for us guys? Can't do that, can't fix aging.
    1
  11. AI has already exceeded the intelligence of the average Republican voter.
    1