Could we ever have AI policy makers?

We live in an age where life is becoming increasingly digital, and with technology taking over jobs and devices now conventional objects in everyday life it is not too hard to imagine a day when artificial intelligence will be given more responsibility. Humans are becoming increasingly dependent on programmes, whether it be to store information or perform complex tasks, and although there is some concern as to what we should let AI do, who is to say we won’t be relying on AI to make impactful decisions in the future? Although in the present moment this may seem purely fictitious, is it possible that one day humans won’t be the only things running our governments? This article will discuss the possibilities and practicality of having artificial intelligence in positions of power.

The Possibilities

Artificial intelligence is increasingly being used to fulfil more than just trivial tasks, and the use of technology in business is an ever-growing field, with the World Economic Forum estimating that by 2025 the time spent on tasks in the workplace will be equal for humans and machines. The social sciences have also started to integrate AI into their practices, and beneficial outcomes are already being seen.

Providing relevant information to voters would support our modern democracies by enabling the population to vote for the candidates who would best suit their political interests; all of this working towards bettering the concept of our ‘representative republics’. Not only this, but AI has also proved promising in aiding governments to make better informed policies by ‘begin(ning) to close data gaps that have long impeded effective policy making’ and guiding the politicians at each of the five policy making stages: identification, formulation, adoption, implementation and evaluation. Examples of AI operation helping to better policy making include how economic development strategists in Quebec are now employing AI to better understand labour, education, and economic variations between regions in order to build more effective economic strategies, and government officials in the UK using AI to evaluate the impact that carbon tax emissions have on business productivity. It is clear that AI has proven useful in the creation of numerous policies and as the technology begins to become more accurate and affordable the capability and market of such programmes only increases.

The future of AI in the social sciences, such as economics and politics, has been viewed as a push towards more statistically informed choices, in turn leading to better outcomes and decisions made, and as new technological advancements occur, we are seeing AI being increasingly utilised. In a briefing done by the European Parliamentary Research Service, a report by Accenture - a consulting company - indicates that AI usage in developing economies could double their annual economic growth rates by 2035. This is predicted to be done by increasing labour productivity, the creation of a virtual workforce (‘intelligent automation’) and the ‘diffusion of innovation’. By incorporating technology such as this on a governmental scale through the route of a virtual policymaker the potential for economic development would be immense. Allowing AI to generate improvements such as these on a nationwide scale could propel the country into a new age of economic and global power and would aid the nation in establishing an international presence. We would be entering an age where data collection and analytics would enhance a country’s position while enabling politics and economics to become more ‘true’ sciences.

The Practicality

Artificial intelligence has the potential to bring about the political and economic changes needed in a prospering and globally present nation, but the ability for this progress has only been demonstrated through AI aiding human policy makers, not an independently run system. The question of whether AI could make policies without human intervention is one of great complexity. There are several factors that come into play when considering this, such as security and hacking, the complications of programming such an AI, the monetary costs, and the public’s fears, all of which have to be dealt with prior to implementing technology in these roles. 

The first problem, security and hacking, is a very reasonable concern to have considering how advanced an AI that is capable of making policies would have to be. Programming an AI that deals with masses of sensitive information would prove to be an extensive task, considering how much security and processing ability the AI would have to have. It would need to have layers upon layers of security protocols considering that ‘increasing dependence on AI for critical functions and services will not only create greater incentives for attackers... and also the potential for each successful attack to have more severe consequences.’. The temptation of having access to a powerful nation’s sensitive files may entice foreign governments to infiltrate their systems to gain a geopolitical advantage, and we could see a wave of cyberwars as more and more internationally established countries make the leap to AI policymaking. This would prove troublesome, especially if people are able to manipulate data to ensure that the AI creates policies that would cause internal economic and political instability. Multitudes of thought would have to go to ensure that all of the information used is protected and that the AI is highly resistant to any infiltration or reprogramming before it would even be considered, and this could easily take years to develop. 

Issues with the costs of hiring programmers to build an AI of this complexity are also to be considered, especially since it is government funding (likely through increased taxes and loans) that would be used to pay for the project. Many people would be opposed to the idea of increased government debt or a rise in taxes at the expense of a programme that may not work or be the best option in terms of safety. Especially in the wake of the pandemic, governments are hoping to revamp the economy and provide support to businesses and individuals who have suffered economically during the lockdowns. The growth in real GDP in the US fell by 31.4%, a very concerning amount considering figures like this had not been seen since the Great Depression, and unemployment reached 14.7% - the highest since WW2. As a result, many economic advisors, and the general public, would prefer to focus on economic recovery rather than ambitious investments. Hopefully as global economies recover and trade and services resume more attention and investments can go into programming. As it becomes more widespread we can increase the number of programmers able to develop advanced AI systems from ten thousand to much more, therefore decreasing the price it would cost to develop such a program. Although a system like this is undoubtedly costly and intricate, steps are being taken to make AI more widely available and make a goal like this much more attainable. 

Public concerns are probably the biggest challenge when trying to implement a change as big as this into a society, particularly a democratic one. A considerable concern with introducing AI into positions of power would be what happens to those already in the positions. Whether they would have to be fired or reassigned is a debate that is consistently going to create opposition if this idea were ever introduced. It would not be taken well by the government officials who create policies as they would understandably fear losing their jobs, and considering that the government is the body that would have to introduce the idea (possibly in the form of a bill) and react to the opinions of the public, it is likely they would avoid introducing the proposition all together. Even if the bill was introduced, the public reaction would likely not be a positive one. In a survey done by the European Commission’s Eurobarometer, 74% of people believed that robots will cause more job disappearance than creation, enforcing a negative view on the impact of robotics and AI on daily life, and it is highly possible that participants were not even considering the high profile jobs that impact millions of people’s lives daily. In a survey, only 23% of people in the UK and 25% of people in the US agreed that they trust AI, and with these countries being second and fourth in government AI readiness respectively it is unlikely that their populations are ready for such a shift, no matter how prepared their infrastructure is.

To conclude, the probability of having an AI policymaker in the near future is very unlikely, but as technology improves and people become more accustomed to artificial intelligence the possibility only rises. The benefits of such a proposition are plenty, but the effort needed to execute the idea may prove to be too much for us to actually go forward with it, and as power-craving politicians and highly skilled hackers are determined to prevent this from becoming our reality it is hard to put much confidence in the idea. This probably will not happen within our lifetimes, but as a species with a history of being unpredictable who knows what the future holds?