Mr. James Bryant:

I might break up the two main questions on how automation and artificial intelligence, AI, will affect jobs and algorithmic bias. I will look first at how AI will affect jobs. I use AI loosely because it has come to mean many different things, both to the technology sector and to the public. Many view automation in general like the invention of the railroad or the written language in that there was much fear around those inventions at the time. The agricultural sector was afraid that passing trains would cause cattle to stop breeding, something we, as we look back, think is quite silly. We will look back in many years' time and think many of our fears about automation were quite silly. However, it is not correct to compare it to such modest inventions. Rather, we should view it more as a cultural changing point. The move is more similar to that from the agricultural economy to the industrial economy. As Mr. Marshall stated earlier, there are great lessons to be taken from history. If we look back at that transformation, we see developments such as 12 years of formal education were necessary to bring someone from a farming background up to what was then considered the base-level minimum required education to be able to jump between jobs. Similarly, with such a high level of digitisation, one must look at a similar transformation in education, or even what we consider education, and the topics of lifelong learning in order to have people who are able to dynamically move between jobs, sectors and roles. The idea of digital literacy is not something we can confine to a single department or a single executive in a company. It is the responsibility of all citizens to be digitally literate, which is why it is so important that we are having this discussion today.

If we look at the highest-paid jobs, they usually require tertiary level degrees such as a masters of education or a PhD. The required level has been increasing over the past decades, as one would expect. Similarly, the entry-level requirements of education are also increasing. This is also set to increase over the coming decades. In the case of those who are most at risk of having their jobs automated, as mentioned by previous speakers, those working in agriculture, transport and factories, there is a real question as to how we as a society react to and pre-empt the massive automation that is predicted to be coming down the line within the next few decades. The primary solutions are to not only take the infrastructural approach of throwing ICT computers, laptops and facilities at people and saying, "Go and learn", but to take a more comprehensive look and to state that here are the skills and teachers one needs, the community centres necessary for it and the address of foundational inequalities in society that are causing this. As previous speakers mentioned, the community approach can often work quite well because one learns with one's peers and one is not expected to go on one's own and venture into uncharted territory when it comes to digital skills.

Second, I will touch on the topic of algorithmic bias that has been raised. There are many examples of this. It could be as simple as being on social media, seeing many articles and news sources that have similar opinions to oneself, and thinking one's opinions are mainstream opinions and they are reinforced. There are issues with this that large companies, such as Google and Microsoft, have tackled. If we look at another system, such as policing technologies, there are examples of policing technologies, the use of which is reasonably widespread in the United States, which direct police forces on where best to use their limited resources but this can be a self-fulfilling prophecy in that the most traditionally under-serviced regions experience slightly higher crime. This means that these regions are more likely to be policed and more people are more likely to be charged with crimes in these regions. Similarly, when one goes back and looks at the data set, it looks as though this model was correct and that one should police these regions even more. One can see how that would negatively affect different communities. Such algorithms are not always comprehensive and they will not necessarily look at aspects such as financial crime. They might be focused on violent crime. That is merely one example. When one looks at the idea of digital literacy, as the Senator mentioned, it is also important that policymakers, judges and administrators have these skills and understand not only the possibilities and the use cases but the limitations of such technology. They should be able to say that maybe this is not the place for statistical analysis and that it is important to have a human in the loop making final decisions in these cases.

Finally, this brings me back to the idea that one should not only educate those who are easy to educate. It is easy to say, if one looks at ICT graduates, that most of them are men with an interest in the sector from middle-class and more affluent backgrounds. While it is fantastic that we have such a high rate of ICT graduates, it should also be a real point of worry that 50% of citizens do not have basic digital literacy skills. It is a large divide. Working on ways to attract that 50%, not necessarily to specialist-level literacy but at least basic digital literacy, is of vital importance.