Conclusion 169
misconception around AI is that people think we’re close to it”, a statement that was more or less in agreement with other contributors such
as Joanna Bryson, from the University of Bath and Princeton’s Center
for Information Technology Policy, for whom the very notion of a singularity is logically impossible.
16
While the notion of general AI remains
something of a pipe dream, as Bryson observes we already operate at a
level of super-human activity, in which machines and algorithms operate
much faster than human capabilities, although not yet demonstrating
anything like human consciousness or the transferability of skills to multiple tasks. Software can, for example, play rule-restricted games better
than any person but fail in cognitive tasks that humans take for granted.
As the authors of the Elsevier round table observe, the company itself
makes use of automation via natural language processing to determine
categories for submissions, allowing them to be distributed more quickly
to human peer reviewers, just as Facebook’s algorithms sift millions of
pieces of content, or Ai’s Wordsmith software can produce thousands
of pieces of content much faster than human editors or writers. Such
activities, however, still have to operate within carefully delimited parameters, which is why more complex, non-repetitive cognitive tasks will
still require human agency for many years to come.
Economic pressures, particularly those caused by the fallout from a
collapsing financial model based on advertising that has been disrupted
by big tech in the domain of digital distribution, will mean that publishers will turn more and more to automation to plug gaps that were once
filled by journalists. As with the shift towards automation in factories
this will result in massive economic shifts with important consequences
for those who are failed by the new order. A McKinsey Global Institute
report in 2018 placed skills into five categories: physical and manual;
basic cognitive; higher cognitive; social and emotional; and technological. McKinsey estimates that by 2030, the amount of time required
by human workers for manual and physical tasks such as working on
production lines or driving will fall by 11 per cent from approximately
90 billion hours per year in the USA, and by 16 per cent in western
Europe, from 113 billion hours in 2018. For basic cognitive tasks, such
as handling cash or essential literacy for data inputting, the fall will be
14 per cent fewer hours required in the USA (currently 53 billion hours)
and 17 per cent in western Europe (62 billion hours).
17
Higher cognitive, social and emotional skills, and technological requirements are
likely to increase over the next decade, but how these are categorised
is not always so obvious: it is often assumed that writing, like reading, is a higher cognitive skill, but natural language generation as well
as processing indicates that creating simple stories can be done much
more efficiently by software than by people. For local news providers
struggling to provide the staff to cover essential information regarding healthcare, crime, and government services, algorithmic journalism
misconception around AI is that people think we’re close to it”, a statement that was more or less in agreement with other contributors such
as Joanna Bryson, from the University of Bath and Princeton’s Center
for Information Technology Policy, for whom the very notion of a singularity is logically impossible.
16
While the notion of general AI remains
something of a pipe dream, as Bryson observes we already operate at a
level of super-human activity, in which machines and algorithms operate
much faster than human capabilities, although not yet demonstrating
anything like human consciousness or the transferability of skills to multiple tasks. Software can, for example, play rule-restricted games better
than any person but fail in cognitive tasks that humans take for granted.
As the authors of the Elsevier round table observe, the company itself
makes use of automation via natural language processing to determine
categories for submissions, allowing them to be distributed more quickly
to human peer reviewers, just as Facebook’s algorithms sift millions of
pieces of content, or Ai’s Wordsmith software can produce thousands
of pieces of content much faster than human editors or writers. Such
activities, however, still have to operate within carefully delimited parameters, which is why more complex, non-repetitive cognitive tasks will
still require human agency for many years to come.
Economic pressures, particularly those caused by the fallout from a
collapsing financial model based on advertising that has been disrupted
by big tech in the domain of digital distribution, will mean that publishers will turn more and more to automation to plug gaps that were once
filled by journalists. As with the shift towards automation in factories
this will result in massive economic shifts with important consequences
for those who are failed by the new order. A McKinsey Global Institute
report in 2018 placed skills into five categories: physical and manual;
basic cognitive; higher cognitive; social and emotional; and technological. McKinsey estimates that by 2030, the amount of time required
by human workers for manual and physical tasks such as working on
production lines or driving will fall by 11 per cent from approximately
90 billion hours per year in the USA, and by 16 per cent in western
Europe, from 113 billion hours in 2018. For basic cognitive tasks, such
as handling cash or essential literacy for data inputting, the fall will be
14 per cent fewer hours required in the USA (currently 53 billion hours)
and 17 per cent in western Europe (62 billion hours).
17
Higher cognitive, social and emotional skills, and technological requirements are
likely to increase over the next decade, but how these are categorised
is not always so obvious: it is often assumed that writing, like reading, is a higher cognitive skill, but natural language generation as well
as processing indicates that creating simple stories can be done much
more efficiently by software than by people. For local news providers
struggling to provide the staff to cover essential information regarding healthcare, crime, and government services, algorithmic journalism