Hi, I'm Tim Tyler - and today I will be discussing the idea that
self-improving systems already exist - in the form of companies and
other organisations - and that future self-improving systems are
likely to arise from the evolution of such existing organisations.
A common perception is that self-improving systems will be developed
directly out of existing computer software projects.
Here are Eric Drexler and Steve Omohundro discussing some of the
efforts that have been made on this front so far.
[Eric Drexler and Steve Omohundro footage]
However the premise that self-improving systems are exclusively a
future development seems suspect to me.
Rather, I consider today's companies to be examples of self-improving
systems. Companies gradually improve themselves over time - therefore,
they qualify as a form of self-improving collective intelligence.
Other organisations can also act to improve themselves - notably
government organsations, charities, religions - and so on.
It is true that companies contain some components which are hard to
incrementally upgrade - namely human brains. However the brains can be
replaced one at a time by the process of automation - a process which
is relatively well-understood.
My expectation is that the self-improving systems that will be
dominant in the future will arise out of today's companies - or other
similar large organisations.
Which organizations are the ones most likely to undergo explosive
growth and development? In my view the three most likely possibilities
are the super-librarian scenario, the super-stock-market-trader
scenario and the super-security scenario. The largest such
organisations today would correspond to Google,
Bridgewater Associates and the NSA.
Though they operate in different domains each of these organisations
has a large proprietary base of sofftware which they do not share with
the rest of the world. This knowledge can be thought of As being part
of the genetic basis of the organisation - and it is also what
improves over time.
Keeping their knowledge secrets hinders these organisations is some
respects - since they may be able to make more money by sharing their
knowledge with their partners. However, secrets do keep their
technology out of the hands of their competitors - these companies do
gradully accumulate a wealth of knowledge that only they posess - and
that can translate into competitive advantages.
If you look at the security services, they have a long track record of
reaching important discoveries first and then keeping them secret -
public-key cryptography, differential cryptanalysis, programmable
computers - and so on.
Companies - or robots?
The nearest all-machine equivalent of a company is probably a robot.
However companies enjoy advantages over robots in several areas.
Companies can own goods, enter into contracts, and have legal rights and
responsibilities. Robots currently have none of these abilities.
By utilising humans, companies can take advantage of the capabilities
of organic brains in areas where these are strong. The humans can then
be gradually made redundant as their skills are surpassed by those of
machines. In this way the human brain is used as a convenient stepping
stone - to be used when needed, and then discarded when exhausted.
It seems likely that an all-machine approach would be handicapped by
an inability to exploit human labour during the early stages of
development - when humans still have things of economic value to
So, it seems reasonable to expect companies - or possibly governments
- which are the most significant self-improving systems of today - to
evolve directly into the main self-improving systems in the future -
with human labour dropping out of the picture gradually - as tasks
within these organisations progressively succumb to automation.