Arrgh, Henry Kissinger talking about expert systems, nothing could trigger me faster. Thank you Bernard, here I go:
Firstly: Chess and go are examples of how well cognitive systems perform is closed
environments, where all the rules are known and the goal perfectly clear.
Expert systems are just algorithms designed to optimise the win conditions and
nothing more, it is indeed humans who set this win conditions, and they (as
with all engineering projects) need to choose the speed at which they develop.
Self driving cars have been in development since 1995. Companies that have
pushed too hard too fast (Tesla, Uber) have seen what often happens to such
efforts, and more cautious companies (Google) are developing solutions that are
much more robust.
Now to the three areas of concern raised by Henry:
1 - Unintended results, Yes! science love
unintended results as it can point the way to new paths of understanding, but
this is how we as humans can challenge our assumptions. We set the win
conditions and are then amazed when the computer finds an original new path too
them. This is the very nature of science. It works better than just bombing Laos.
3 - Expert systems cannot explain the
rationale, I am very sceptical about this, as again it is the programmer who
set (and can audit) the win conditions. The solutions may seem strange, but
they will always be defined by clear criteria, which can be interrogated. If you want to look for weird win conditions, Henry was the master of that in his foreign policy.
2 - AI may change human thought patterns
and goals. The more I thought about this the more I realised that HK had a
point, albeit one that is regularly made. Facebook and Google are platforms
using cognitive systems to sell clicks of advertisements (or lie about clicks
and then sell them). This has lead to everyone getting to live in their little
echo chamber and struggling to engage with alternate points of view.
So the real concern is how to we broaden
the human mind to consider multiple perspectives and care about the results of
our actions. I struggle to see how a national vision on A.I. will help us,
although a bit more philosophy in our schools and
universities might be a good step. Seems a shame Kissinger didn't raise that.
