Page 71 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

2. Third Gen AI 59

To manifest the short fall of second Gen AGI we consider the DAV, especially
“what to do at a red light?” besides “stop.” We adopt the aforementioned KR in
terms of (Attribute A, Object O, Value V) Relational databases (RDB) for 1000
DAVs experience FMF, then the Boolean logic among these sets of FMF’s, for
example, (1) Attribute A: Lyaponov FMF to Control dynamically the high speed
vehicle on a fixed turning radius, DAVuses the accelerating pedal generating the out-
ward centrifugal force that must be balanced by the steering wheel turning inward,
and vice versa; (2) Object O: Car weight and tire friction in Langevin equation gen-
erates FMF that might contribute to glide over safely; (3) Value V: space-time GPS
FMF. A combined RDB is (attribute [collision avoidance, speed up], Object [corner
radars, center top Lidar, window video sensors], value [traffic situation in space-
time]) FMF to provide machine with the judgment decision, for example, when there
are no pedestrians including uniformed police, siren of ambulance, or police car.
When there is no incoming traffic during the midnight at mid-West Arizona,
Phoenix. Stopping can cause an impatient passenger (human passenger) to take
over DAV. Third Gen AI, or i-AI, will be “Experience Based Expert System
(EBES)” which is so-to-speak, older and wiser. The state of art of AI deep learning
is that the planar siliconebased computer hardware has furthermore miniaturized
with 3D multiple layers connectivity GPU into the backplane of tabletop, and like-
wise ANN software has developed a matching hardware backward error propagation
Tensor Flow Python Language, like wearing a matching gloves with hands. Together
the result has been that it impressively passed the Alan Turing test, because the other
end could beat human genius Mr. Lee Se-dol 4:1 in Go game in March 2017. This
state of art has changed first Gen AI to become learnable rule-based system second
Gen AI (e.g., developed knowledge representation and judgment reasoning [KR] by
RDB to medical diagnoses, psychology, and natural language).
To point out what’s the future, the critical shortfall in second Gen AI in case of
DAV which has been pointed by Science Magazine (December 15, 2017) weekly
issue. It said “not so fast” that DAV is not going to happen We could not believe
TSA committee announced that inspite of so many stake holders in DAV, e.g.,
NASA (Martian Lander); DARPA (Grand Challenge AV); Uber, Google (taxi
without driver); Auto insurance companies level 4 automation with man-in-the-
loop, 13 years, level 5 automation driverless for few decades c.2040w2050. The
leaders in planning and navigation of autonomous vehicles are Waymo (Google)
and Uber invested $1B per annum. The rest of the industry is desperately trying
to catch up. This is absolutely a disruptive technology. Thrun (Stanford) lead the
early DAV work at Google. Sebastian Thrun (PI, Darpa, now Waymo), “Toward Ro-
botic Cars.” (Comm. ACM 53 (April 2010) 99e106). More modern approaches for
AV sensor fusion, planning, and fast time decision and controls are closely held trade
secrets. The general public TSA is always interested in the traffic safety, for
example, the saving of possible 36,000 lives of fatal car accidents in the United
States per year. That’s why it may be presumptuous to ask why one of the best in-
dustries in the United States cannot move AI ANN rapidly forward? We point out