Dr. Kissinger, you're an
elder statesman. Why did you think AI was an important enough
subject for you?
Kissinger: When I was an undergraduate, I wrote my
undergraduate thesis of 300 pages - which was banned after that
ever to be permitted - called "The Meaning of History."
The
subject of the meaning of history and where we go has occupied
my life.
The technological miracle doesn't fascinate me so much;
what fascinates me is that we are moving into a new period of
human consciousness which we don't yet fully understand.
When we
say a new period of human consciousness, we mean that the
perception of the world will be different, at least as different
as between the age of enlightenment and the medieval period,
when the Western world moved from a religious perception of the
world to a perception of the world on the basis of reason,
slowly.
This will be faster.
There's one important difference.
In the Enlightenment, there
was a conceptual world based on faith. And so Galileo and the
late pioneers of the Enlightenment had a prevailing philosophy
against which they had to test their thinking.
You can trace the
evolution of that thinking. We live in a world which, in effect,
has no philosophy; there is no dominant philosophical view.
So
the technologists can run wild.
They can develop world-changing
things, but there's nobody there to say,
'We've got to integrate this into
something.'
When you met Eric [Schmidt] and he invited you
to speak at Google, you said that you considered it a threat to
civilization.
Why did you feel that way?
Kissinger: I did not want one organization to have a monopoly on
supplying information.
I thought it was extremely dangerous for
one company to be able to supply information and be able to
adjust what it supplied to its study of what the public wanted
or found plausible.
So the truth became relative. That was all I
knew at the time.
And the reason he invited me to meet his
algorithmic group was to have me understand that this was not
arbitrary, but the choice of what was presented had some thought
and analysis behind it.
It didn't obviate my fear of one private
organization having that power. But that's how I got into it.
Schmidt: The visit to Google got him thinking.
And when we
started talking about this, Dr. Kissinger said that he is very
worried that the impact that this collection of technologies
will have on humans and their existence, and that the
technologists are operating without the benefit of understanding
their impact or history.
And that, I think, is absolutely
correct.
Given that many people feel the way that you
do or did about technology companies - that they are not really to
be trusted, that many of the manipulations that they have used
to improve their business have not been necessarily great for
society - what role do you see technology leaders playing in this
new system?
Kissinger: I think the technology companies have led the way
into a new period of human consciousness, like the Enlightenment
generations did when they moved from religion to reason, and the
technologists are showing us how to relate reason to artificial
intelligence.
It's a different kind of knowledge in some
respects, because with reason - the world in which I grew up - each
evidence supports the other.
With artificial intelligence, the
astounding thing is, you come up with a conclusion which is
correct.
But you don't know why. That's a totally new challenge.
And so in some ways, what they have invented is dangerous. But
it advances our culture.
Would we be better off if it had never
been invented? I don't know that.
But now that it exists, we
have to understand it. And it cannot be eliminated.
Too much of
our life is already consumed by it.
What do you think is the
primary geopolitical implication of the growth of artificial
intelligence?
Kissinger: I don't think we have examined this thoughtfully yet.
If you imagine a war between China and the United States, you
have artificial-intelligence weapons.
Like every artificial
intelligence, they are more effective at what you plan. But they
might be also effective in what they think their objective is.
And so if you say, 'Target A is what I want,' they might decide
that something else meets these criteria even better. So you're
in a world of slight uncertainty.
Secondly, since nobody has
really tested these things on a broad-scale operation, you can't
tell exactly what will happen when AI fighter planes on both
sides interact.
So you are then in a world of potentially total
destructiveness and substantial uncertainty as to what you're
doing.
World War I was almost like that in the sense that everybody had
planned very complicated scenarios of mobilization, and they
were so finely geared that once this thing got going, they
couldn't stop it, because they would put themselves at a bad
disadvantage.
So your concern is that the
AIs are too effective? And we don't exactly know why they're
doing what they're doing?
Kissinger: I have studied what I'm talking about most of my
life; this I've only studied for four years.
The
Deep Thought
computer was taught to play chess by playing against itself for
four hours. And it played a game of chess no human being had
ever seen before.
Our best computers only beat it occasionally.
If this happens in other fields, as it must and it is, that is
something, and our world is not at all prepared for it.
The book argues that
because AI processes are so fast and satisfying, there's some
concern about whether humans will lose the capacity for thought,
conceptualizing and reflection.
How?
Schmidt: So, again, using Dr. Kissinger as our example, let's
think about how much time he had to do his work 50 years ago, in
terms of conceptual time, the ability to think, to communicate
and so forth.
In 50 years, what is the big narrative? The
compression of time.
We've gone from the ability to read books
to being described books, to neither having the time to read
them nor conceive of them nor to discuss them, because there's
another thing coming.
So this acceleration of time and
information, I think, really exceeds humans capacities. It's
overwhelming, and people complain about this; they're addicted,
they can't think, they can't have dinner by themselves.
I don't
think humans were built for this. It sets off cortisone levels,
and things like that. So in the extreme, the overload of
information is likely to exceed our ability to process
everything going on.
What I have said - and is in the book - is that you're going to need
an assistant.
So in your case, you're a reporter, you've got a
zillion things going on, you're going to need an assistant in
the form of a computer that says,
'These are the important
things going on. These are the things to think about, search the
records, that would make you even more effective.'
A physicist
is the same, a chemist is the same, a writer is the same, a
musician is the same.
So the problem is now you've become very
dependent upon this AI system.
And in the book, we say, well,
Who controls what the AI system does?
What about its prejudices?
What regulates what happens?
And especially with young people,
this is a great concern.
One of the things you write
about in the book is how AI has a kind of good and bad side.
What do you mean?
Kissinger: Well, I inherently meant what I said at Google.
Up to
now humanity assumed that its technological progress was
beneficial or manageable. We are saying that it can be hugely
beneficial.
It may be manageable, but there are aspects to the
managing part of it that we haven't studied at all or
sufficiently. I remain worried. I'm opposed to saying we
therefore have to eliminate it. It's there now.
One of the major
points is that we think there should be created some philosophy
to guide to the research.
Who would you suggest would
make that philosophy? What's the next step?
Kissinger: We need a number of little groups that ask questions.
When I was a graduate student, nuclear weapons were new.
And at
that time, a number of concerned professors at Harvard, MIT and
Caltech met most Saturday afternoons to ask,
What is the answer?
How do we deal with it?
And they came up with the arms-control
idea.
Schmidt: We need a similar process.
It won't be one
place, it will be a set of such initiatives. One of my hopes is
to help organize those post-book, if we get a good reception to
the book.
I think that the first thing is that this stuff is too powerful
to be done by tech alone. It's also unlikely that it will just
get regulated correctly.
So you have to build a philosophy.
I
can't say it as well as Dr. Kissinger, but you need a
philosophical framework, a set of understandings of where the
limits of this technology should go.
In my experience in
science, the only way that happens is when you get the
scientists and the policy people together in some form.
This is
true in biology, is true in recombinant DNA and so forth.
These groups need to be
international in scale? Under the aegis of the U.N., or whom?
Schmidt: The way these things typically work is there are
relatively small, relatively elite groups that have been
thinking about this, and they need to get stitched together.
So
for example, there is an Oxford AI and Ethics Strategy Group,
which is quite good. There are little pockets around the world.
There's also a number that I'm aware of in China.
But they're
not stitched together; it's the beginning.
So if you believe
what we believe - which is that in a decade, this stuff will be
enormously powerful - we'd better start now to think about the
implications.
I'll give you my favorite example, which is in military
doctrine.
Everything's getting faster. The thing we don't want
is weapons that are automatically launched, based on their own
analysis of the situation.
Kissinger: Because the attacker may be faster than the human
brain can analyze, so it's a vicious circle.
You have an
incentive to make it automatic, but you don't want to make it so
automatic that it can act on a judgment you might not make.
Schmidt: So there is not discussion today on this point between
the different major countries. And yet, it's the obvious
problem.
We have lots of discussions about things which are
human speed.
But what about when everything happens too fast for
humans?
We need to agree to some limits, mutual limits, on how
fast these systems run, because otherwise we could get into a
very unstable situation.
You can understand how
people might find that hard to swallow coming from you.
Because
the whole success of Google was based on how much information
could be delivered, how quickly.
A lot of people would say,
Well, this is actually a problem that you helped bring in.
Schmidt: I did, I am guilty.
Along with many other people, we
have built platforms that are very, very fast. And sometimes
they're faster than what humans can understand.
That's a
problem...
Have we ever gotten ahead
of technology? Haven't we always responded after it arrives?
It's true that we don't understand what's going on.
But people
initially didn't understand why the light came on when they
turned the switch. In the same way, a lot of people are not
concerned about AI.
Schmidt: I am very concerned about the misuse of all of these
technologies.
I did not expect
the Internet to be used by
governments to interfere in elections. It just never occurred to
me. I was wrong.
I did not expect that the Internet would be
used to power the antivax movement in such a terrible way. I was
wrong. I missed that.
We're not going to miss the next one.
We're going to call it ahead of time.
Kissinger: If you had known, what would you have done?
Schmidt: I don't know. I could have done something different.
Had I known it 10 years ago, I could have built different
products. I could have lobbied in a different way. I could have
given speeches in a different way. I could have given people the
alarm before it happened.
I don't agree with the line of your argument that it's
fatalistic.
We do roughly know what technology is going to
deliver. We can typically predict technology pretty accurately
within a 10-year horizon, certainly a five-year horizon.
So we
tried in our book to write down what is going to happen. And we
want people to deal with it. I have my own pet answers to how we
would solve these problems.
We have a minor reference in the
book how you would solve misinformation, which is going to get
much worse.
And the way you solve that is by essentially knowing
where the information came from cryptographically and then
ranking so the best information is at the top.
Kissinger: I don't know whether anyone could have foreseen how
politics are changing as a result of it.
It may be the nature of
the human destiny and human tragedy that they have been given
the gift to invent things. But the punishment may be that they
have to find the solutions themselves.
I had no incentive to get
into any technological discussions. In my 90s, I started to work
with Eric. He set up little seminars of four or five people
every three or four weeks, which he joined.
We were discussing
these issues, and we were raising many of the questions you
raised here to see what we could do.
At that time, it was just
argumentative; then, at the end of the period, we invited Dan Huttenlocher, because he's technically so competent, to see how
we would write it down.
Then the three of us met for a year,
every Sunday afternoon.
So this not just popping off. It's a
serious set of concerns.
Schmidt: So what we hope we have done is we've laid out the
problems for the groups to figure out how to solve them.
And
there's a number of them: the impact on children, the impact on
war, the impact on science, the impact on politics, the impact
on humanity.
But we want to say right now that those that
initiatives need to start now.
Finally, I want to ask you
each a question that sort of relates to each other.
Dr.
Kissinger, when, in 50 years, somebody Googles your name, what
would you like the first fact about you to be?
Kissinger: That I made some contribution to the conception of
peace. I'd like to be remembered for some things I actually did
also.
But if you ask me to sum it up in one sentence, I think if
you look at what I've written, it all works back together toward
that same theme.
And Mr. Schmidt, what would
you like people to think of as your contribution to the
conception of peace?
Well, the odds of Google being in existence in 50 years, given
the history of American corporations, is not so high.
I grew up
in the tech industry, which is a simplified version of humanity.
We've gotten rid of all the pesky hard problems, right?
I hope
I've bridged technology and humanity in a way that is more
profound than any other person in my generation.
