Chapter 10, Conclusions and Reflections

1. I do not feel that robots or AI are a threat in any way, because the necessary and sufficient conditions for robots to be a threat do not exist.  Which is to say, the robots have to want to take over the world, and have a need to take over.  Currently, robots and AI have no wants or needs.
2. We would need project managers, packaging designers, advertising and marketing, sales people, and support staff.
3. Psychologists study normal and abnormal mental states and cognitive processes, exactly what we are trying to simulate for artificial personality.  We want the robot to not trigger bad responses in people.  I once had a robot with flashing red eyes that caused small children to have panic attacks.  Psychologists would help avoid such errors.
4. GPS receivers, radios, Wi-Fi, Bluetooth, accelerometers, gyroscopes, and these days, apps.
5. Because they are universal approximation systems that work in probabilities and averages, not in discrete numbers and logic.  ANNs can take a different amount of time because a particular bit of data may take different paths at different times, going through a different number of neurons and thus not taking the same amount of time to process.
6. You can use a neural network based system to model a bad human operator for a driving simulation to help teach other drivers (and self-driving cars) how to avoid bad drivers.  The desired state is an unpredictable driver, so just train the neural network to 60% or so.

7. So we have a group of 100 stocks picked by our AI program.  Of that set, an indeterminate number are winners and losers.  There is a 43% chance the stock is a winner, and a 57% chance it is a loser. We have no way of judging the stocks being winners or losers except by investing our money, which is what we are trying to avoid – investing in bad stocks.  A 43% chance of winning is not good.  The second AI has an 80% chance of telling you that the first AI chose a bad stock. So 80 times out of 100, you will know that the stock was not a winner.  So you are left with an 80% chance of correctly identifying one of the 57 bad stocks, which eliminates 45 stocks.  That leaves you with 55 stocks, of which 43 are winners (on average), which raises your odds to 78%.

Using Bayes Theorem, I recomputed the combined probabilities as 75.1% - so I’ll take either answer:

Bayes theorem p(x|c) = (px* pc) / (px*pc)+(1-px)(1-pc)