Research

The Genesis Group

To better understand our intelligence in general, and our keystone story understanding competence in particular, my students and I formed The Genesis Group. You can find detailed descriptions of what we do at the Genesis page and learn more about why we do it in Moonshot, below.

Moon Shot

Copernicus taught us about the solar system. Darwin did the same for evolution. Then, Watson and Crick determined the structure of DNA. Collectively they answered fundamental questions about who we are. Now, we can realistically dream of another scientific achievement of equal importance: constructing a top-to-bottom, computational account of our own intelligence.

My students and I want to do it now because we are curious, because the problems are hard, because the problems are exciting, and because we need a better understanding of ourselves and each other and the forces that make us what we are.

We need to do it now because the scientific answers will revolutionize the engineering of intelligent systems. Applications with humanlike intelligence will emerge and empower in education, health care, policy development, business, energy, the environment, cybersecurity, and all the other high-impact areas with unsolvable problems that we must solve.

We can to do it now because we are asking better questions; because computer power is now effectively unlimited; because of encouraging progress in the contributing fields; because of maturing techniques for studying the neural substrate; and because there is immense student interest.

Our better questions include: How are we different from other species? And what are the competences we share with other species such that the difference matters.

Our answer is that we do, in fact, have a differentiating, keystone competence: we build complex, highly nested symbolic descriptions of situations and events. Together with the competences we share with other species, the keystone competence enables story telling, story understanding, and story composition, and all that enables much, perhaps most, perhaps all of education.

We may be wasting our time, of course, but the potential reward is that 1,000 years from now, everyone could say that we first understood our own intelligence.

Will they take over

Links to WCVB TV Chronical show devoted to concerns about the future of AI, first aired 2 March 2016.

  • Segment 1
  • Segment 2
  • Segment 3
  • Segment 4
  • The Genesis Group is part of MIT's Computer Science and Artificial Intelligence Laboratory.

    Members of the Genesis Group particpate in MIT's Center for Brains, Minds, and Machines, sponsored by the National Science Foundation.

    The center is dedicated to developing a deep understanding of intelligence and the ability to engineer it, ephasizing how it works, how it develops in childhood, how it is implemented in neurobiology, and how it is amplified through social interaction.


    Copyright © 2016 Patrick Henry Winston
    This site was updated on 24 October 2017
    Designer: Chiai Takahashi

    Teaching

     

    In the fall, I teach 6.034, Introduction to Artificial Intelligence, a large class.

    Grades are calculated in accord with several student-oriented principles. For example, because we want to encourage each student to compete with himself/herself and not with each other, we no longer compute a class average. Because anyone can have a bad day, each segment of the material is tested both in a quiz and on the final, and we only count the higher grade.

    We who teach the subject are pleased to note that grades are correlated with attendance at lectures, quiz reviews, recitations, and tutorials.

    Anticipating a shift of skill building to the on-line world, we have enriched the subject with what we call Right-now Talks, aimed at exposing exciting, up-to-the-minute research occuring at MIT.

    In the spring, I teach 6.xxx (also known as 6.803 and 6.833), The Human Intelligence Enterprise, a class in which I focus on contributions that visionary thinkers have made toward developing a computational explanation of intelligence.

     

    During the Independent Activities period, I talk on the subject of How To Speak. The talk helps people do a better job in lectures, theses defenses, and job talks.

    10-250, The Center of the Universe. The next offering will be at 11:00 am, on Friday,
    2 February 2018
    , in Room 10-250, The Center of the Universe.

    If you can't wait, an aging version, lacking the latest material, is available from the Harvard Bok Center.

    Awards

    All of us involved in developing and teaching 6.034 and 6.xxx over the years are immensely pleased that our work on these subjects has been appreciated by students. Our collective efforts led to the 2011 Eta Kappa Nu Teaching Award for excellence in instruction, a MacVicar Faculty Fellowship in 2011, the Baker Award for undergraduate teaching in 2010, and the Graduate Student Council Teaching Award in 2006. Unknown, but heroic students went to a lot of trouble to make it happen.

    We were also pleased to note that, according to a Bloomberg article, 6.034 is among Five of the Best Computer Science Classes in the U.S. Of course, what they meant to say was that 6.034 has had graduate-student teaching assistants that are consistently outstanding. Jessica Noss, for example, was the 2016 winner of the EECS Carlton E. Tucker Award for outstanding teaching as a graduate-student teaching assistant.

    Books


    During my book-writing phase, I wrote numerous editions of various textbooks and collections, 17 in all.

    Free, online editions of On To Java, On to Smalltalk On to C and On to C++ are available via the website.

    Video

    Various people have recommended my lectures on Neural Nets and Deep Neural Nets. Maybe that is why 6.034 is among the 10 most visited subjects of the 2340 MIT subjects with materials offered free to all via MIT's OpenCourseWare.

    In the fall, I hope to record for OCW a new lecture on deep-net enabled AlphaGo. I'll also hope to record a second lecture on Support Vector Machines featuring the latest ideas of Vladimir Vapnik and Rauf Izmailov.

    Pensées

    Artificial Intelligence and Chocolate

    In the old days, if you wanted to attract viewers to your page, you mentioned chocolate. Now, you mention Artificial Intelligence.

    For those of us who have been around for a while, we are overcome with déjà vu.

    The first wave of excitement followed from James Slagle's symbolic integration program in 1961. It did better work than most MIT freshman, so some feared the end of human usefulness was near.

    The second wave of excitement came with the introduction of rule-based expert systems in the 1980s. One diagnosed infections of the blood better than primary-care physicians. Many such systems did and still do useful work in areas from resource allocation to medicine, but they are not an existential threat.

    Now we have a third wave; we have Machine Learning and the subdiscipline of Deep Neural-Net Learning. Both are best viewed as Computational Statistics, a kind of processing enabled by newly available massive computing on massive data sets.

    The Deep Neural Nets classify objects in pictures, invent captions, and translate from one language to another. AlphaGo beat Lee Sedol, a top-ranked human Go player, and the most recent version, announced in a paper in Nature's 19 October 2017 issue, became the world's best Go player by learning from games it played against itself.

    All these give the appearance of humanlike understanding but without much actual humanlike understanding. None has any idea what it is doing.

    Nevertheless, the world has responded with astonishing gusto. Fantastic headlines appear frequently, such as this one from Forbes in June, 2017: “5 Ways Facebook Uses Artificial Intelligence To Counter Terrorism.”

    Important people weigh in, as when Valdimir Putin told Russian school children, in September, 2017: “Artificial intelligence is the future, not only for Russia, but for all humankind. It comes with colossal opportunities, but also threats that are difficult to predict. Whoever becomes the leader in this sphere will become the ruler of the world.”

    Hmmm. Déjà vu. I think it useful to see how such headlines and comments sound with statistics substituted for Artificial Intelligence.

    Of course all new technology can have unanticipated downsides, and there is no harm in some people thinking about how to mitigate those unanticipated downsides, but for as for me, I think it early to put Artificial Intelligence on par with, say, climate change, engineered pandemics, and nuclear conflagration.

    I like Andrew Ng's response to the most extreme form of overexcitement. Ng, who has considerable talent for composing pithy characterizations, said in a conference in March, 2015, that fearing a rise of killer robots is like worrying about overpopulation on Mars.

    7 October 2017

    Most recent pensées
    The General Patton diet
    Favorite pensées
    The Arcosanti file:
      Great works
      The MIT 100
      The Infinite Connection Advisor
      The End of Live Lectures?

    . Subscribe to pensées.

    The Faculty News Letter

    I am on the board of the MIT Faculty News Letter, the voice of the MIT faculty. In that capacity I have contributed substantially to several editorials, including two on MITx, MIT's on-line initiative:

    MITx discusses pros and cons of going on line.

    What's Next with MITx comments on possible futures and suggests questions that should be asked.

    I also contributed substantially to an editorial on the Abelson Report:

    Not Blameless but Not to Blame comments on lessons to be learned from Professor Hal Abelson's committee's Report to the President: MIT and the Prosecution of Aaron Swartz

    I also contributed several Teach Talk articles:

    Skills, Big Ideas, and Getting Grades Out of the Way explains how we give students two shots at showing they know the material.

    Looking at the Numbers looks at the correlation between lecture attendance and grades.

    Dialog on Right-Now Talks describes an experiment aimed at exposing students to what's going on in research right now.

    And finally, a Phillipic against the idea of a faculty senate.

    Where.i.am

    My office is room 32-251 in MIT's Stata Center, which was designed by Frank Gehry

    Who.i.am.not

    I do not use social media. For a time, there was a fraudulent Facebook page with my picture and other material that attempted to deceive. Now I have a defensive Facebook page saying that I do not use Facebook or other social media.

    Why do imposters do what they do? Some may be psychotic. Many are criminals attempting to extract personal information from “friends” via a form of so-called spear phishing.

    Who.i.am
    You can learn more about who I am and what I'm up to from a short biography, my Curriculum Vitae, and the rest of this home page.

    Several friends and I started Ascent Technology, Inc., a company that develops products that solve complex resource-planning, resource-scheduling, resource-allocation, and situation-assessment problems.



    I served three six-year terms as a member of The Naval Research Advisory Committee*, which advises the Chief of Naval Operations and the Chief of Naval Research on technical matters.

    During my second term, while I was chair, NRAC focused on manning and on the concept of an all-electric Navy. One high-impact NRAC study, Reduced Ship Manning, led to the Smart Ship Program. Another, CVX Flexibility/Integrated Electric Power emphasized the need to move toward electric drive on our carriers.

    * Alas, I have learned that NRAC, along with many other advisory committees, did not survive the sequester, leaving the Navy, strangely, with no analog of the Air Force Science Advisory Board or the Army Science Board, or the Defense Science Board.

    Will.i.am

    Will likes to see stuff at MIT whenever he is in town. This time I took him to see robots in the Computer Science and Artificial Intelligence Laboratory, wearable computing in the Media Lab, and miscellaneous cool stuff in the Precision Engineering Research Group. It wasn't hard to find people to help out.

    Continued...

    What MIT should do

    I believe technology will take university education through a period of instability—what Andy Grove would call a 10X period—as new educational technology is introduced for the first time since the invention of movable type. This period of instability coincides with a window of global scientific opportunity and engineering challenge.

    Accordingly, I believe that technical universities that want to be important in 2050 should chart a new course now, which I lay out, somewhat telegraphically in a sample mission statement and in a fanciful interview, recorded in 2050.

    The MIT FSILG Task Force

    I have strong views about MIT's system of fraternities, sororities, and independent living groups. During 2002–2004 I co-chaired, with Stephen Immerman, the MIT Task Force on FSILGs: Status and Future Development. The work of the task force led to a detailed report recommending a six-step plan and the creation of Project Aurora.

    For the Record

    The MIT150 Celebration

    If you missed the symposia that were part of MIT150 Celebration, too bad. They were great. Fortunately, you didn't really miss them, as they are all on line.

    In the Computation and the Transformation of Practically Everything Symposium, I spoke in the History panel.

    In the Brains, Minds, and Machines Symposium, I spoke in the Golden Age Keynote Panel. and the Language and Thought Panel.

    The State of the Institute

    From 2006 to 2012, I expressed concern from time to time about a drift of MIT away from mission, community, and collegiality. Now, however, we have new leadership, with a different voice, so I have moved the philippics to page two, preserved for the record, and focus page one on the future.

    hit counter