For new readers and those who request to be “好友 good friends” ,  please read my 公告栏 first.

Every year betweensemesters and during the winter recess, the Paulson School of Engineering andApplied Sciences (SEAS) at Harvard through its Institute of Applied Computational Sciences (IACS)  sponsors a week long course/workshop/Seminar entitled COMPUFESTinvolving students faculties, and invited speakers.
Computest 2016 http://computefest.seas.harvard.edu/  carries the title of “BRAIN + MACHINES”. I went to the full day seminaron 1/22/2016 http://computefest.seas.harvard.edu/symposium
Below are my “takeaways”

1.                  Complexity ofour Brian is our distinction as an animal species. We know very little about the functions of our brain.

2.                  Brain has always associated with the technology of the day throughout theages. Currently, it is Computer or Computational Sciences.

3.                  Mental disorder is a problem with the softwares of our Brain .

4.                  Computational neuroscience is Brain+Machines, or  technology-wise: Robotics + Artifical Intelligence

5.                  There are 100 billion neurons and 100 trillion connections inour brain. It consumes about 20watts energy and yet possesses  enormous capability/flexibility not matched by computers.

6.                  Will understanding biology and biotechnology help?  We now have quite sophisticated technology for access to neurons.

7.                  The US government has set up the Intelligence  Advanced Research ProjectAgency (IARPA) for the study of the brain in similar fashion following the successes  of DARPA.

8.                  Self drivingcars is one current example of computers  replacing human brains. However,despite Google Car, the real  problem has not been solved. Google car has very detailed photo map of the city in which it practices driving. It is  still an accomplishments in a narrow domain. But progresses are  being made. The frequencies of necessary human  intervention (or “handoff to humans”, i.e., Google car says to the human,please take over, I cannot handle this) has decreased  from one per 700 miles driven to one per 5000 miles driven.

9.                  Simultaneous Localization and Mapping (SLAM) are  important problems for self-driving cars.
In other words, we need to know What and Where n the world – i.e., object based sensing and understanding

10.               CHARACTERISTICS OF AN INTELLIGENT AGENT, react  to random events, learn , cooperate, decide, and have a goal
E.g. GPS is a simple agent , self driving car is much more complex, soccer playing robots are cooperating  agents

11.               Connections among agents are increasing explosively, 50 billion things will be connected by2020

12.               Group decision making: buying a family car, a hiring  committee, where to go for a group dinner, elections. And how  to aggregate opinions and elicit preferences,  voting rules, and the Arrow paradox.

13.               Ethical principles ( life, promises, confidentiality, loyalty,  public good, ) and moral value, e.g., trust. Learning and  adaptation  in aggregating opinions and in cooperation;  Resolving ethical and preference of agents; Psychology, philosophy, safety and OR constraints, - symbiotic environment for group decision making. These are all problems/issues/characteristics  AI agents must have and resolve.

14.               Is our expectation for AI too high periodically,  i.e., AI goes through cycles of over-promise-and-under-deliver resulting i “AI winter” of disappointment and low research activity?

15.               Some partial answers:

a.  because generalization in AI is still immature,

b.advances are in narrow domain only.

c.  big data help little so far

d.Is biology giving us new path in AI? Probably not

e.  Are we too early in issuing rules and regulation for AI?

f.   In AI, 98% correct is not 100%(which is required)!!

g.AI education for the public who are generally poorly inform about the truth of AI

h.AI will not soon replace human intelligence.

i.    AI even in narrow domain can do harm.

j.    We don’t undertand human intelligence well either. For example, why can we stop from over eating?

As an outsider to this field, and at the age of 82, I was reaching the  point of  diminishing return after 4 hours of concentrated attention. Certainly enough questions were raised by the speakers to  occupy my interest. I decided not to attend the afternoon sessions  which were more concerned with the biology of the brain.