> On 15-Mar-2019, at 2:28 AM, Brian Holmes <bhcontinentaldr...@gmail.com> wrote:
>
> There is much to critique in the operations of Boeing and of the FAA. But
> it's not about AI taking full control.
https://www.architecturalrecord.com/articles/13464-structural-design-and-thinking-in-approximations
<https://www.architecturalrecord.com/articles/13464-structural-design-and-thinking-in-approximations>
This short essay by Robert Silman is about another field totally - structural
engineering, but the point it makes about our relationship with computers and
thinking in approximations is significant. Humans can get a overall ‘feel’ for
a system that is far more efficient than a computer in understanding the
holistic character of the system - and to do this requires thinking in
approximations.
The challenge with the computer is that:
Its capabilities are based in computing power rather than contextual
understanding, and the learning and decision making in its intelligence comes
from harnessing this computing power to discern sensible patterns within a host
of randomly collected factors. The system works well when it is inserted into
a context that is within the predictable range of prior learning, but put the
system into a complex non-linear context (like wind flow, climate, collective
social choice) and every now and then it will hit a situation that lies outside
this predictable range. It then falls apart as its analysis is based on
finding correlations rather than building empathy or understanding, and it has
no way of assessing whether the error it finds is minor or major.
It is expected that the human will intervene in such situations. But because
these situations are so rare and random, the human gets habituated to the
routine reality revealed by the computer. And because the computer can reveal
tremendous visual detail, the human thinks that he/she is getting a far better
feel for reality. The human stops thinking in approximations, loses the ‘feel’
for the overall system, and is therefore also ill equipped to deal with crises
or errors thrown up by the machine.
# distributed via <nettime>: no commercial use without permission
# <nettime> is a moderated mailing list for net criticism,
# collaborative text filtering and cultural politics of the nets
# more info: http://mx.kein.org/mailman/listinfo/nettime-l
# archive: http://www.nettime.org contact: nett...@kein.org
# @nettime_bot tweets mail w/ sender unless #ANON is in Subject:
