On Sun, Sep 11, 2016 at 1:47 PM, Lotfi A. Zadeh <zadeh@eecs.berkeley.edu>
[the god of fuzzy logic]:
AI is back in popularity after a long winter. What the cause the change?Here is my thought. Please post yours on the bulletin board:
------------------------------
My reply:
The revival of neural networks and of learning in 1987/1988 was large enough to mobilize a small minority, but the technical culture in general was still too conservative to pay real attention. It acted like the "antibody" forces in the movie Inception. At NSF, I got to see first-hand, and play, in a cultural clash which was quite challenging, with important aspects which don't fit the word limits here.
The recent cultural change was mainly due to the impacts of one of the $2 million grants we funded under a one-year special topic, "COPN", which I fought for and won back in 2008. Search on "COPN" at www.nsf.gov to find that topic. In the old NSF (which maintained the full spirit of Vannevar Bush and Joe Bordogna), I was allowed to fund a proposal by Ng and LeCun, despite very heavy pressures not to. (At www.nsf.gov, search on "COPN Ng LeCun" to see the grant.)
Crudely speaking, many of the great stalwarts of AI had written books with limited research behind them, describing only kindergarten level neural nets, not acknowledging the more advanced work done before, asserting that learning systems in general could never solve real problems. The Ng/LeCun grant simply empowered them to DO THE EXPERIMENT -- to apply slightly more advanced neural networks and other learning systems to challenge problems which conventional AI people had worked on and competed on for decades. When, in one fell swoop, they broke all records quickly for a series of well-publicized open challenges, in image recognition, speech recognition and natural language processing, DARPA and Google both took notice; lots more funding became available, and more people entered the fray. Through very tactful politics, they succeeded in labelling this as a new advance in AI and machine learning.
IN DC, I see very directly what kind of new enthusiasm has resulted in the funding world and in the large companies which tend to drive a lot of the other stuff. The terms "deep learning" really lead the pack.
But "the revenge of the empire" is still a threat. The success of an expert system, Watson, in defeating Jeapordy also had a big psychological impact. Under the new more conservative management of science in the US (which has been spread from DOE and NASA to NSF), there are folks working to run the entire world (the INternet of Things being the base) in a top-down political way, bypassing any enthusiasms in the marketplace or in academia as a whole, let alone little things like the Arab Spring. They are much more determined and ruthless and effective than what we overcame in previous years.
For more details on the recent situation see:www.udc.edu/winter_school/
Video for first 13 slides posted at https://www.youtube.com/watch?
==========================================
===============================================
Before the Ng/LeCun project, and perhaps before the NSF initiative on Learning and Intelligent Systems, the state of AI was similar in many ways to the state of the space program today. Both remind me of the old Russian saying "we pretend to work and you pretend to pay us." 1960's technology were enshrined in power, supported by a few key people (like Senator Shelby and Congressman Lamar Smith) -- and emphatically narrow enough that greater powers were content to leave them alone, with a predictable but small role.
Ironically, I also had a chance in 2003 to do some breakthrough in the space area too, which a mere $2 million grant might have accomplished (with follow-ons of course, as with Ng and LeCun). I deeply regret that I listened to my partner at NASA on that... but there are a lot of other things I would have done differently with 100% hindsight. I do have some foresight, but there are ever so many layers we need to get right here!
Actually, the space example reminds me of the old Zen story about a buy who breaks his leg, who later finds it was for the best, who still later discovers it was not so good after all, but then later still...
When I later tried to find a new way to support that effort, I learned sad but important realities about what is happening in the US right now. For the best, or not? Will that knowledge offer hope of overcoming key problems, or will it just hit the reset button?
Actually, the space example reminds me of the old Zen story about a buy who breaks his leg, who later finds it was for the best, who still later discovers it was not so good after all, but then later still...
When I later tried to find a new way to support that effort, I learned sad but important realities about what is happening in the US right now. For the best, or not? Will that knowledge offer hope of overcoming key problems, or will it just hit the reset button?
==============================
===============================
I am sad it was not appropriate this morning to cite two important new papers in process which go further, one
for NATO and one for neuroscience. I did get an email from Yeshua, who generally agreed with the NATO paper on policy towards terrorism, but would have wanted to go further by adding a paragraph:
On that
note, some final reflections are important. We need to make a clear distinction
between “short” and “long” term solutions together with fundamental vs. quick
fix solutions. When we consider human culture and human potential we must
remember that our highest individual’s spiritual potential lies in the areas of
self- mastery, self-responsibility and self-governance with foundations in the
embodiment of Universal (Spiritual) values [27], towards the application of
constructive intelligence and creativity as antidotes to the limiting
structures created by the brain and nervous system that impose behavioral
values based on reward and fear conditioning, in the form of learning that can lead
to destructive behavior (like terrorism) and unhealthy partial loyalties
towards any group. This human potential development is without a doubt the long
term and fundamental solution for the chaos and destructive forces we actually
face. However, it seems, that this long term solution must be achieved in
scenarios where the application of quick fixes with detrimental side effects
are avoided and instead, we apply some short term solutions that will
strengthen the probabilities of success of the fundamental solution. So far,
this seems to be a well balanced risk taking endeavor in the area of applied quantum
physics and mathematics together with the exploration of Universal Values and
human potential, where for example, a truly sane self-aware person would send a
warning back in time, so that the core version of him or herself has been
changed to a form of avatar or savior who has achieved such a level of altruism
with a well informed consciousness, by having a “real” taste of a ghastly
future and opting for a self-transformation backwards in time, in order to
survive in a scenario of collective freedom and happiness, a better world.
====================
That seemed heavy enough, but a Quaker friend said it was not enough:
Dear Paul et al,
Thank you very much for including me in this fascinating look at cutting-edge science and philosophy. Of course much of it is beyond me, so I’m a fool to rush in where angels fear to tread. But rush I will, as a preface of perhaps discussions to come.
We live in a mostly open society. To the extent that we are open, terrorists always will be able to find vulnerabilities to exploit to their advantage. Otherwise we become even more a closed society that stagnates and feeds terrorism.
To my mind, the best human culture/human potential method to overcome terrorism is to reduce and try to eliminate the inequalities and injustices that give rise to the frustrations that culminate in terrorist acts. Fighting terrorism with military weapons only feeds the frustrations that recruit new terrorists, as much of the 21st Century has shown. As a human race we need to learn from Cain and Abel that we are our brothers’ keepers, and that evil is best overcome by love rather than resisted by force. That’s a hard lesson to learn, but we can try to learn and apply it, one by one, using brilliant insights to live and spread the message, as Ghandi and King have done in our time.
So to me, the final paragraph is not enough.
John
================
Well, there are always important details one must leave out in a brief treatment.
The guy I mentioned in the previous blog had some very specific ideas in similar directions,
beyond the scope of this blog.
Some of this really reminds me of a couple of the novels of Dan Simmons, but today I plan to change tracks yet again.
As a minor matter... I was amused in a way when the TV reported Hillary Clinton having some small issues with dizziness and vertigo. I had the same, even milder, recently... and found it was a side effect of Zantac, which my doctor advised that I take instead of Prilosec/opremazole. So I am back to opremazole as of yesterday. I also pulled a muscle moving heavy bags of mulch, and Luda explained to me that aspercreme is one good part of treatment -- "no, it's not one of those evil pain-killers, it's just antiinflammatory, aspirin for the muscles."
And with hot water and just the right kind of exercise, I hope I will be ready for the next insane adventure...
No comments:
Post a Comment