26
mergers and acquisitions were about 21.8
billion USD worldwide, and start-ups without
revenue fetched prices that amount to $5-10 million per AI expert.
59
As highly-qualified
experts can now earn very high annual salaries, universities will have great difficulties in
finding competent teachers for this specialty. Some practical implementation work can be
done by relative novices using openly available development tools and learning materials,
but the development of mission-critical applications requires quite advanced skills.
60
One rather immediate result of this situation is that high-level
AI talent and compute
capability will probably be provided as a service. This would perhaps mean that there is
not going to be massive needs for high-level AI competences. Due to the high wage
differentials, many
current students of statistics, mathematics,
mathematical physics,
computer and chip design, and perhaps neurophysiology may, however, reconsider their
career paths and find new identities as experts in AI. Moreover,
in the current informal
learning environment, easy access to state-of-the-art technologies
and research could
also mean that high-level AI competences may emerge from unexpected places, for
example, through open software and open hardware communities.
59
Data from PitchBook, quoted in (Bass 2018).
60
One key bottleneck for neural AI is its energy consumption.
As a result, many chip designers are now
trying to develop semiconductor chips that can be used for specific AI applications, see e.g. (Salvo 2018).