Turing’s Test 101
top page hits are clearly news stories written by humans about the links
between companies such as AP with Automated Insights, but most –
with titles such as “Strong travel demand lifts Boeing in first quarter” or
“PepsiCo beats earnings forecasts” – are generated by software. The VT
of Vermilion Sands does exist – it’s just that there is more serious money
in journalism than poetry.
Robo-Writers and the Algorithm of News
As we saw in a previous chapter, the Turing Test was devised by Alan
Turing as a means of determining whether humans would be able to
differentiate between machine and human responses. While John Searle
among others has criticised whether this would demonstrate true artificial intelligence – or, more accurately artificial consciousness, what is often referred to in this book as general artificial intelligence – for practical
considerations, this is much less important than the intentional stance of
readers and audiences. What matters most in practical terms is whether
those readers can distinguish (or, indeed, even care) as to whether the
author of a story is a human or algorithm. Writing in 2013, Spyridou
et al argued that technological innovation was viewed at that time in
the newsroom as a means of improving professional practice – that indeed such professional culture articulated as skills, ideas, and practices
worked to weaken the potential impact of such innovation.
4
In the space
of five years, following even more rapid closures across multiple newsrooms than had even been witnessed in the first decade of the twenty-first
century, it is probably not unfair to suggest that this professional culture has been severely attenuated and that publishers – if not journalists
and editors – are more willing to experiment with software-generated
content, so-called robo-writers or algorithmic journalism. Certainly, a
study by Young and Hermida in 2015 found that the Los Angeles Times
was engaging in what the authors called “computational journalism”,
the use of the newspaper’s Homicide Report and Data Desk to generate
material for news stories, if not always the final copy.
5
The activities at the LA Times are part of a much wider phenomenon,
what Örnebring and Conill refer to as the “outsourcing of newswork”.
6
This includes a wide range of activities, some of which – such as a reliance on external press agencies or using PR agencies – have their roots in
twentieth-century practices, but steep challenges to newsroom budgets
also push for increased automation of all aspects of the news production process wherever possible.
7
The conditions that would enable the
next step away from merely using technology networks to better manage
freelancers, for example, towards moving towards full-scale softwaregenerated content represent not solely a technical issue. As Flew et al
have observed regarding algorithmic journalism proper (what he, like
Young and Hermida, refers to as “computational journalism”), fully
top page hits are clearly news stories written by humans about the links
between companies such as AP with Automated Insights, but most –
with titles such as “Strong travel demand lifts Boeing in first quarter” or
“PepsiCo beats earnings forecasts” – are generated by software. The VT
of Vermilion Sands does exist – it’s just that there is more serious money
in journalism than poetry.
Robo-Writers and the Algorithm of News
As we saw in a previous chapter, the Turing Test was devised by Alan
Turing as a means of determining whether humans would be able to
differentiate between machine and human responses. While John Searle
among others has criticised whether this would demonstrate true artificial intelligence – or, more accurately artificial consciousness, what is often referred to in this book as general artificial intelligence – for practical
considerations, this is much less important than the intentional stance of
readers and audiences. What matters most in practical terms is whether
those readers can distinguish (or, indeed, even care) as to whether the
author of a story is a human or algorithm. Writing in 2013, Spyridou
et al argued that technological innovation was viewed at that time in
the newsroom as a means of improving professional practice – that indeed such professional culture articulated as skills, ideas, and practices
worked to weaken the potential impact of such innovation.
4
In the space
of five years, following even more rapid closures across multiple newsrooms than had even been witnessed in the first decade of the twenty-first
century, it is probably not unfair to suggest that this professional culture has been severely attenuated and that publishers – if not journalists
and editors – are more willing to experiment with software-generated
content, so-called robo-writers or algorithmic journalism. Certainly, a
study by Young and Hermida in 2015 found that the Los Angeles Times
was engaging in what the authors called “computational journalism”,
the use of the newspaper’s Homicide Report and Data Desk to generate
material for news stories, if not always the final copy.
5
The activities at the LA Times are part of a much wider phenomenon,
what Örnebring and Conill refer to as the “outsourcing of newswork”.
6
This includes a wide range of activities, some of which – such as a reliance on external press agencies or using PR agencies – have their roots in
twentieth-century practices, but steep challenges to newsroom budgets
also push for increased automation of all aspects of the news production process wherever possible.
7
The conditions that would enable the
next step away from merely using technology networks to better manage
freelancers, for example, towards moving towards full-scale softwaregenerated content represent not solely a technical issue. As Flew et al
have observed regarding algorithmic journalism proper (what he, like
Young and Hermida, refers to as “computational journalism”), fully