Here are the notes for my talk at IAAI-10, 15July2010.
At the moment, this files is also available at: www.rohan.sdsu.edu/faculty/vinge/misc/iaai10
The Technological Singularity
It seems plausible that with technology we can, in the fairly near future,
create (or become) creatures who surpass humans in every intellectual and
creative dimension.
In 1993, I said I'd be surprised if the Singularity happened before
2005 or after 2030.
Nowadays, my friends ask: "So Vernor, are we on schedule?"
And my answer is that I still think the Technological Singularity is the most
likely non-catastrophic outcome for the relatively near future, and with the
time schedule shown above.
But there are interesting follow-on questions.
Hey, since it's already 2010:
- Do we know more about what the world would be like after the Singularity?
- I think the answer to this question is "no". The Singularity
would be qualitatively different from tech advances of the past
(the Industrial Revolution, the development of agriculture), in that
the intelligence of the top players afterwards would be beyond us.
- Do we know more about what the run-up to the Singularity might look like?
- This should be a more tractable question; maybe "yes".
Paths to the Singularity
each with its special flavor and danger and promise (though developments are
concurrent and interacting):
- Artificial Intelligence (AI)
- Intelligence Amplification (IA)
- Computer Networks plus Humanity
- Digital Gaia: Fine-grained distributed systems
Note that altogether this list involves hundreds of thousands of serious
researchers, many of them not even knowing they are part of the enterprise.
Artificial Intelligence
Certainly 2010 is a good year for assessment of one of the classic
arguments for strong AI:
We are into petaflops computing and still seem to be on track with the hardware
side of Hans Moravec's picture after Robot: Mere Machine to Transcendent Mind (Oxford University Press): http://www.frc.ri.cmu.edu/~hpm/book98/fig.ch3/p060.html
- In 2010, perhaps the starkest evidence of progress in the classical AI
area is the way that computer-run models and bayesian methods have taken
over the sciences.
- Going forward from 2010: Whether the application is autonomous
automobiles or natural language translation, the next few years will be
profoundly interesting. In the end, I think we (I mean, you!) will capture
all that AI optimists ever hoped for -- and if you don't, what refractory
problems remain will be revelations in themselves.
Intelligence Amplification
How intimate can the user interface become?
Paraphrasing David Brin's take on the close coupling of computers with the
human intellect: ~"The computer portions might make up a kind of neo-neocortex,
providing scalable processing power, while the organic part provides what we
natural humans have always been good at: desire, setting goals and aims.~"
- Many people who are enthusiastic about the Singularity, regard
Intelligence Amplification as a safe and comforting path. After
all, if you are becoming smarter and smarter, then the Singularity
might be as transparent and navigable as past technological revolutions.
- If extreme IA happened abruptly, I don't think it would be
especially safe; we humans are naturally bloody-minded. On the other
hand, IA could proceed in a gradual way.
- Biological life has its virtues (see further on in this talk), but in this
context, biological life doesn't have legs.
Computer Networks plus Humanity
- There were also a number of papers at AAAI-10/IAAI-10 in this area.
In 2010, Humans plus the cloud is already evidence of radical intellectual
change in Human culture. The Human side is limited to a few billion people,
no matter how effective the networks. The machine and network side has a
lot more
headroom.
- Along with IA, this path to the Singularity gives hope that
something like humanity can provide guidance to further progress.
Digital Gaia: Fine-grained distributed systems
- Extreme ubiquity of networked sensors/effectors/logic,
both free-standing and embedded in ourselves and all our artifacts.
- An aside: If you popularize the notion of the Technological
Singularity, you will run into journalists (and friends :-) who say:
"But isn't this Singularity stuff isomorphic to claims of religious
apocalypse? You know -- a Rapture of the Nerds?" (Ray Kurzweil had a
good-natured response to this sort of question with the picture on page
368 of his book, The Singularity Is Near.)
- As of 2010, Digital Gaia seems a clearcut, credible path to the
Singularity. But if there is any religious isomorphism to claims about
Digital Gaia, it would be with animism, not apocalypse.
Thus, there are religious analogies in all directions, but that's mainly
because we are an era where technology interacts with the deepest issues of
mind and philosophy.
- Science-fiction writer Karl Schroeder has written a number of novels about
the fine-grain, distributed scenario, in particular:
- Ventus -- the conflict between cloudy "big iron" and
ubiquity is pursued into the far future, where reality has become its
own database.
- The Sunless Countries -- wherein the dust motes are so
powerful that when asked to produce an artifact (say an aircraft),
they evolve it on the spot.
The difference between a trend and an avalanche
From year to year, watching these developments gives me the feeling that I'm
seeing not a trend, but an avalanche. (For an amusing -- and chilling -- riff
on tech inevitability, see Charles Stross's short story, "Antibodies".)
- Over the last twenty years, those who are enthusiastic about the
Singularity have concentrated on human-centered possibilities:
- Intelligence Amplification
- Mind uploads
- Virtual realities (though the most extreme version of VR
is analogous to still another style of religion).
- While these are interesting, they are only a part of the broad
range of possibilities coming our way.
- In fact, one of the strongest arguments for the likelihood of the
Singularity is that there is such a broad range of projects that might lead
to it.
There's another consequence of this broad advance. What if most or all of
the above approaches are successful?
Looking for historical (or paleontological) precedents
I used to think that the best analogous prior event to the Singularity was the
rise of the Humankind within the animal kingdom. That event certainly had the
quality of unknowability (for the animal precursors of Humans).
I still think the rise of Humans is the most recent analog of the Singularity,
but I've come to think that there is another event that may be more like what
is happening now. That is the Cambrian Explosion
-- the "sudden" appearance about half a billion years ago of a great variety of
complex, new life forms.
The events of the Cambrian Explosion stretched across millions of years; the
upcoming era could see as much variety and change in decades. A major
question is whether the potential varieties of mind will be realized, or be
blocked by the "first-movers". Of course, another question is what happens
to us ordinary Humans, and to superhumans that still have Human-like thought
processes!
Mitigating factors
- We and our nearterm enhancements are a form of first-mover. Despite
biological limitations, we are in a position to keep up with early events,
perhaps even influence them.
- What is the unit of natural selection in this new era?
Getting an answer to this question (or creating an answer!) may be the most
important issue with regard to longterm optimism about the new era.
- In fact, contrary to Bill Joy's assessment, the future may need
us (or something very like us):
Life considered as subroutine-threaded code
A basic principle for most conservation and green groups is the interdependence
of living things. I think this principle is valid at all levels, if one is
careful with definitions; it is starkly evident at the cellular level.
Just as higher biological life can't survive without much of the rest of the
domain of life, some of the deficiencies of machines compared to biological
life may never be cured!
Even if this aeons-long tradition of operational dependence is broken by the
machines, post-Singularity life should welcome the continued presence of
creatures like good old-fashioned, natural Humans. After all, existential
disasters can happen to anybeing, and Humans are the all-purpose backup.