Last weekend I attended the "Humanity+" conference in Melbourne
(
http://hplusconf.com.au), held at RMIT. It consisted of an
eclectic mix of presentations by invited speakers, without
contributed papers or a published proceedings, though videos of the
talks will become available. The conference was under the
auspices of the Humanity+ organisation (
http://humanityplus.org),
whose aim is to promote thinking about the "next steps" of
humanity. The main areas of focus appear to be biomedical and
bioengineering developments for longer and healthier life, leading
on to enhancements of the body, and artificial intelligence and
enhancements of the mind. The chair of Humanity+, Natasha
Vita-More, was one of the presenters. I went because I thought
the gerontologist Aubrey de Grey would be worth hearing, and because
the artist Stelarc was giving a presentation.
This conference was more optimistic than pessimistic. Climate
change and population pressures were there in the background, and
sustainability was a theme, but on the whole the intent was to look
beyond these problems to longer-term possible futures for
humanity. The organiser was Adam Ford, who has just become a
board member of Humanity+, and who has had a considerable
involvement in this general area.
Maybe 80 people attended, predominantly but by no means exclusively
male, and a mixture of young and old, with relatively few people in
the middle age range. I got the impression that almost everyone there
had a background in science, engineering or computing.
Aubrey de Grey was well worth hearing. His view on ageing is
that normal metabolic processes produce "damage" of various kinds,
such as junk inside cells that the body cannot dissolve. We
can tolerate a certain amount of such damage, but eventually it
starts to harm us. De Grey listed all the classes of damage
that are known (and indicated that no fundamentally new classes of
damage had come to light in the last 30 years), and indicated
plausible approaches to dealing with all of them. He mentioned
two specific projects at his laboratory dealing with junk inside the
cell, targeted at macular degeneration, which is a leading cause of
blindness, and at atherosclerosis, which is inflammation of the
walls of the arteries, heading to heart disease and strokes.
All of this comes under the heading of "regenerative medicine",
therapies to rejuvenate (that is, to make young again) systems in
the body by clearing out damage and taking the bodily systems back
some way towards the healthy young adult state. Once such
therapies are in place for all the major types of damage (which is
quite a few years away), de Grey thinks that we will be able to have
another 30 years of healthy middle age. These days 60 is the
new 50; with these therapies 80 or 90 would be the new 50. But
that is only the start. As techniques improve, clearing out a
greater proportion of damage, repeated rejuvenation would allow
enormous prolongation of healthy, active life, to ultimately maybe
1,000 years. This doesn't imply a cure for cancer, but it does
imply a method of avoiding cancer by manipulating telomeres (the
caps at the end of DNA strands).
All of this provoked a lot of discussion, and de Grey devoted his
second presentation to discussing objections to his program.
The diseases of ageing are not just a first world problem: de Grey
said that already two-thirds of the deaths in the world are due to
them. Of course if we do have the potential to live to 1,000
years there will have to be massive changes in society, but de Grey
pointed out that by the time such long life becomes feasible there
will have been massive changes in society anyway.
Incidentally de Grey is not a food faddist or anything of that
sort. He was asked about diet, and said that as long as one is
reasonably sensible about diet and exercise (and doesn't smoke),
things like the "paleo diet" and the like don't achieve much.
And he enjoyed a beer at the pub afterwards.
The other presentation that contained a road map for future
developments was that of Tim Josling on artificial
intelligence. He outlined the so-called hype cycle that tends
to apply to new technologies. Once a new technology becomes
known, at first there is a great deal of hype, resulting in wildly
inflated expectations. When the technology doesn't live up to
these, there is a "trough of disillusionment", and then after than
attitudes to the technology finally settle to a realistic view of
what it can achieve.
Artificial intelligence (AI) went through this cycle: after quite a
long initial period of hype the "AI winter" descended in the 1980s,
when funding dried up and AI was generally regarded to have
failed. In fact it developed quietly in various specialised
areas. Josling listed several techniques developed years or decades ago
that were impractical at the time but are now coming in to their own
as increased computer power has made them feasible.
Incidentally Josling is more optimistic about the continuation of
Moore's Law (that the number of transistors on a chip doubles every
two years) than Herb Sutter (whom I mentioned in a previous post),
but it doesn't matter for Josling's argument whether increased
computing power arrives via Moore's Law in one box or via
networks, as Sutter expects.
Josling expects that more and more low-level white-collar jobs will
be cheaper to do by machines, on a relatively short time frame, and
he ended by posing the question: "Leisured aristocracy or unemployed
underclass?"
This sort of prophecy was made in my youth, and hasn't really come
to pass. However, the "acceptable" minimum rate of unemployment
has risen from 2% to 5% in my lifetime, and since the official
figures are constructed to be as low as possible, the true unemployment figure is
at least 10%. I also think that the availability of cheap
Third World workers has delayed the development of automation, but
that is beginning to come to an end. Eventually the machines
will be cheaper than even a Third World worker.
In the background of Josling's presentation is a concept known as
"The Singularity", and there was a panel discussion around this at
the conference. The Singularity is when machines become
smarter than we are; this may be a long way off, but it is hard to
argue convincingly that it can never happen. The Singularity is a sort
of "event horizon", as we cannot predict what would happen after that.
As far as raw processing power is concerned, by one estimate a
current desktop machine with a good graphics card has maybe 1/2000
of the raw power of a human brain. Networks of 2000 such
machines already exist. Though one of the panellists,
Colin Hales, indicated
that recent discoveries have indicated that the brain may have far
more power than the above estimate implies.
The work up until now has been in specialised domains, for example
making driverless trucks for mining sites. There was mention
of a possible approach to general artificial intelligence being
pioneered by Marcus Hutter at the Australian National
University. Josling indicated that the promising advances in
artificial intelligence involve various forms of machine learning
(and I got the impression that this applies to Hutter's work); this
led into a discussion of risks. If a machine has learnt from
experience rather than being explicitly programmed (and this already
happens in some areas) then we don't know in detail how it does what it does. If it does something unexpected and kills or injures
someone, it is not at all clear who should be held
accountable. One of the attendees, who works as a safety
engineer (I didn't catch his name) said that once a technology such
as that for driverless trucks is mature, it is more reliable than
having human drivers; it is the early period of introduction of such
technologies that is really dangerous. In this context, the
Google Car has driven itself autonomously around Los Angeles.
One of the panellists, James Newton-Thomas, who works with
autonomous mining equipment, indicated that the current approach is
to segregate the equipment behind physical barriers, as well as fitting independent safety systems.
A discussion that was only touched on at the conference was how to
make sure that a super-intelligent machine would be friendly towards
us, and there was some discussion about the relationship among
consciousness, intelligence and morality. There was also some
discussion about the uses to which governments and large
corporations would put super-intelligent machines. The
prospect of large-scale technological unemployment and the
thought-police-like powers already available via automated
surveillance and data mining are much more immediate concerns.
(To be continued...)
