Have you created life?
I certainly think of them as creatures, not computer code. We could describe
any self-reproducing or metabolic system as alive, and they are
self-reproducing, metabolic systems, so I am happy to call them alive.
Are they just like any software following an algorithm?
Advertisement
Yes and no. They must be algorithms, because they live in a computer. When
you write a computer program to simulate atomic theory, you haven鈥檛 made atoms.
But if you take those simulated atoms, plug them together, under the right
circumstances they start to behave like molecules, and then those molecules
react with each other and form networks of reactions, and eventually more
complex behaviour emerges. So, at the first level, it is a con. But the second
level is just as real as it is in our Universe.
Compared with animals, where you would place them?
They鈥檙e like ants. They don鈥檛 have thoughts of their own and they are trapped
in a sensory loop, where the environment tells their senses what to receive, the
brain generates behaviour, and, much like us, that behaviour changes the
environment, and so on . . .
Were you surprised they worked?
No, I expected them to work, otherwise I wouldn鈥檛 have dared try. But I was
astonished that people were impressed and that they took to them. There are now
somewhere between five and ten times as many norns as there are elephants. It is
now the biggest Web user community in the world.
People are doing lots of experiments with norns on the Internet. What鈥檚 the
most bizarre?
Probably the most monstrous thing is this guy called Anti norn. I think he is
a US Marine. He has a website devoted to torturing the creatures. In response,
people have got rescue centres and adoption agencies for these poor battered
creatures. If people are worrying about these creatures and adopting them, then
I have achieved one of my aims鈥攖o make people ask questions.
So is a sense of moral responsibility creeping into A-life?
The first time I really knew I was successful was when someone sent me an ill
norn by e-mail and asked me to cure it. The problem was it just stood there, and
they were worried because it was wasting away. I discovered that it was deaf and
blind because a gene hadn鈥檛 been expressed. It couldn鈥檛 see or hear the world,
so it wasn鈥檛 responding. I persuaded the gene to wake up, cured it, taught it
some swear words, then sent it back to Australia. They sent me a Christmas card
to say it was doing fine. Afterwards I wondered who was the idiot鈥攖hem for
worrying about a data file or me for spending a day trying to cure it?
What鈥檚 so great about biology that we want to copy it?
Because it works. Biological machines have qualities that none of our
machines have: ours are not robust, not adaptive, not intelligent, and they are
certainly not user-friendly. Even simple animals can have all of those
qualities, and, in a sense, I see my job as putting the life back into
technology.
Aren鈥檛 you just reducing life to information and processing
information?
I sincerely hope not. The problem with reductionism is there is a difference
between describing something and explaining it. You can describe complex systems
in the language you use to explain their parts, but you cannot necessarily
explain them that way. I would hate to think that I was explaining away
life.
Do you identify with Victor Frankenstein?
Yes, Frankenstein was a misunderstood guy. He was trying to do what any of us
do, which is to understand ourselves.
Could it go wrong the same way for you?
No, because Frankenstein was a lot smarter than me. He made a human being,
whereas I am just making insects and eventually I want to make mice and rabbits.
There are millions of species on this planet and I see no reason to fear yet
another species.
Could norns ever evolve into more intelligent creatures?
No. Although I tried to make the genetic structures that wire up their brains
as flexible as I could, I don鈥檛 think it鈥檚 enough to give them the freedom to
make the next step up in intelligence. They might develop a few more facilities,
and subsequent norns have been manipulated to have different brains. Whether
they are more intelligent or not is a moot point. I opened the system up as much
as possible for people to play with鈥擨 don鈥檛 mind people tampering as long
as they can鈥檛 be malicious.
There is growing concern about the effect of genetically modified organisms
on the environment. Is there a similar problem in releasing your digital DNA
onto the Net?
My creatures only exist in a virtual world, they couldn鈥檛 live on the Net. I
doubt if you can have intelligent life on the Net. Intelligence has to exist in
a rich, self-consistent world. The Net is consistent, but it is not
self-consistent because you have to be a human being to understand it. Unless an
agent has that knowledge, it can鈥檛 make sense of the Net so an awful lot of
attempts to make intelligent search agents are doomed. At the moment the Net
doesn鈥檛 even support the protocols for agent communications. When it does,
people will have to be very careful about releasing these organisms. But they
will be no different from computer viruses. I had to be careful when I designed
norns to make sure that you couldn鈥檛 wipe someone鈥檚 hard drive from inside a
norn鈥檚 brain.
You have interesting views about intelligence . . .
Well, there鈥檚 much more to intelligence than logic. Most of the AI pioneers
were mathematical magicians and philosophers and to them, thinking was about
logic, about reason. But it has to grow out of more primitive systems, and most
intelligence is not logical. Most of the time we are not reasoning people. Dogs
tend not to argue about syllogisms, but they still seem bright. Chuck IBM鈥檚 Deep
Blue chess computer and a dog into a pond, and see which one climbs out first.
Which means that intelligence is grounded in survival. If you haven鈥檛 got a
reason to think, you won鈥檛 think, and survival is what motivates us.
Are we wasting our time with robots that have no sense of existence?
Nothing is a waste of time. But there鈥檚 been a backlash against symbolic AI
and everyone is now far more interested in nematodes. There is a huge middle
ground that nobody seems to be exploring, an area where psychology starts to
make sense but human reasoning is irrelevant. More than an insect, but not quite
human . . .
It鈥檚 2000. Where are the intelligent robots? Has AI failed?
Yes. Alan Turing predicted that the idea of machines thinking would be
routine by the end of the century. The very second that time ran out for his
prediction was the second that computers were set to demonstrate their complete
stupidity because they couldn鈥檛 add up dates. But Mark II in the AI game will be
a success. I believe we will reach simian intelligence within forty years. I may
be wrong, but you鈥檝e got to have a goal.
Do you believe computers are the right way to go about AI?
It鈥檚 the wrong tool for the wrong job. It was designed to be a logical
machine. It was designed to do one thing very quickly, one step after another.
But all the naturally occurring intelligent systems we know about are made from
large numbers of very dumb things鈥攕uch as neurons鈥攁cting in
parallel. The saving grace of a computer is that you can make it pretend to be a
million stupid machines acting in parallel.
Do you think your work has contributed to science in any way?
I don鈥檛 know. If it has, then it is probably only in showing that something
can be done that other people hadn鈥檛 tried. I am quite proud of what I did
because when I started, I鈥檇 never heard of AI, never read any books on the
subject. I invented neural networks not knowing that other people had done it
better, earlier. I have ploughed my own furrow all this time.
You recently left your position as co-founder and technical director of
Cyberlife, the company that marketed Creatures. Why?
We agreed to part on friendly terms. The rest of the company chose to earn
money, which they can do more quickly if they haven鈥檛 got to develop complex
technology.
Where is your research heading?
I am interested in imagination. I鈥檓 not trying to solve the problem of
consciousness explicitly, although that鈥檚 the only reason I am interested. For a
thing to be conscious, it has to have an imagination to do the higher-level
thought processes.
Isn鈥檛 that a chicken and egg thing?
No. I think imagination has a biological substrate. A world inside the head
to live in as well the world outside, or instead of the world outside. Maybe
consciousness naturally emerges out of it. I suspect not. I think there is
another layer, which I don鈥檛 understand. I think imagination probably comes from
distinguishing between self and non-self鈥攈ow you know whether you are
hitting something or whether it is hitting you. And also out of body
image鈥攖he ability to have a model of your body so that you can predict
where it will be if the nervous system isn鈥檛 quick enough to get feedback. But
at some point in evolution, something usurped that ability to distinguish
between self and non-self and something new emerged. That something was the
subconscious which allowed us to plan, and from that emerged consciousness.
I鈥檝e never managed to keep a norn alive long enough to breed. Am I doomed as
a parent?
I鈥檝e succeeded in keeping one child alive for 18 years but I鈥檓 a lousy norn
keeper as well. They all either die on me or learn to hate me. I think they
resent me for being their God . . .