9.15 What We Can Learn from IBM’s Watson Computer
191
might assume the most common disease to cause the symptoms, Watson would also
consider rare diseases, which may otherwise be overlooked.
In order to work well, it is important to feed Watson in a non-selective way with
information reflecting different perspectives and potentially contradictory pieces of
evidence. Then, Watson is trained by experts how to evaluate information sources,
and in the end, Watson may outperform humans. The power, quantity and variety of
information sources the cognitive computer can scan and the number of hypotheses it
can generate and evaluate, are vastly greater than any single human could conceivably
cope with. Whereas humans tend to seek and keep information that confirms their
existing beliefs, Watson seems less prone to this bias, as it can consider much more
pieces of information.
9.16 Collective Intelligence Versus Superintelligence
Let us finally address the question whether one could eventually create something
akin to an omniscient superintelligence, if we used all the data of the Internet? In fact,
the Google Brain project probably seeks to create such a superintelligence (and other
projects, too). However, I expect a distributed, networked system of multiple brains
interacting with Artificially Intelligences to be superior to a superbrain. Remember
that the “wisdom of crowds” often exceeds that of the best experts.
25 This implies
that collective intelligence can outperform superintelligence. In a complex world
the consideration of multiple different, independent perspectives and goals is key to
success. Therefore, to cope with the challenges of the future, we need a pluralistic
and participatory approach, as I will discuss it later.
It’s probably not by accident that our society is based on distributed intelligence,
characterized by many moderately sized brains which think differently, with relatively small information flows among them. If this wouldn’t be useful, biological
evolution might have created different solutions already.
26 Smaller brains learn more
quickly. Thus, collective intelligence can probably better adjust to the variability and
combinatorial complexity of our world.
27
25 While Google can easily implement many different algorithms, the very nature of a large corporation with its self-regarding goals, uniform standards, hiring practices and communications means
that the teams developing these approaches will be inherently more prone to observe and follow
each other’s successes. Similar ways of thinking produces less diverse opinions.
26 High bandwidth communication could, for example, make it impossible to stop destructive ideas
from spreading. One could certainly not be interested in fostering the “madness of crowd” (or “social
flash crashes”), which can lead to violence or war, as known from the psychology of masses.
27 As we know, intellectual discourse can be a very effective way of producing new insights and
knowledge.
191
might assume the most common disease to cause the symptoms, Watson would also
consider rare diseases, which may otherwise be overlooked.
In order to work well, it is important to feed Watson in a non-selective way with
information reflecting different perspectives and potentially contradictory pieces of
evidence. Then, Watson is trained by experts how to evaluate information sources,
and in the end, Watson may outperform humans. The power, quantity and variety of
information sources the cognitive computer can scan and the number of hypotheses it
can generate and evaluate, are vastly greater than any single human could conceivably
cope with. Whereas humans tend to seek and keep information that confirms their
existing beliefs, Watson seems less prone to this bias, as it can consider much more
pieces of information.
9.16 Collective Intelligence Versus Superintelligence
Let us finally address the question whether one could eventually create something
akin to an omniscient superintelligence, if we used all the data of the Internet? In fact,
the Google Brain project probably seeks to create such a superintelligence (and other
projects, too). However, I expect a distributed, networked system of multiple brains
interacting with Artificially Intelligences to be superior to a superbrain. Remember
that the “wisdom of crowds” often exceeds that of the best experts.
25 This implies
that collective intelligence can outperform superintelligence. In a complex world
the consideration of multiple different, independent perspectives and goals is key to
success. Therefore, to cope with the challenges of the future, we need a pluralistic
and participatory approach, as I will discuss it later.
It’s probably not by accident that our society is based on distributed intelligence,
characterized by many moderately sized brains which think differently, with relatively small information flows among them. If this wouldn’t be useful, biological
evolution might have created different solutions already.
26 Smaller brains learn more
quickly. Thus, collective intelligence can probably better adjust to the variability and
combinatorial complexity of our world.
27
25 While Google can easily implement many different algorithms, the very nature of a large corporation with its self-regarding goals, uniform standards, hiring practices and communications means
that the teams developing these approaches will be inherently more prone to observe and follow
each other’s successes. Similar ways of thinking produces less diverse opinions.
26 High bandwidth communication could, for example, make it impossible to stop destructive ideas
from spreading. One could certainly not be interested in fostering the “madness of crowd” (or “social
flash crashes”), which can lead to violence or war, as known from the psychology of masses.
27 As we know, intellectual discourse can be a very effective way of producing new insights and
knowledge.