“Technology’s gone robot happy. Any job has to have a robot, or the engineer in charge feels cheated.” (Isaac Asimov, “Risk”)
In “Risk”, a short story by Isaac Asimov, a man is irate at the head of a robotics corporation because he was made to go do a job under extremely dangerous circumstances while a robot was available to do it. This is while, as he felt, robots had been designed precisely so that they could perform any job whose incidental circumstances were too dangerous for a human being. The anger lasts until he understands that the robot could have been given a direction such as “solve it”, because no one knew what the problem was until a human went in and a robot would not have had the lateral thinking skills and the flexibility of approach to identify what the problem was and how to solve it. That required a human.
In that text, I think, is contained the crux of why no artificial intelligence is going to wrench away control from humans. Humans have the capacity to solve problems and figure out plans using indirect and creative approaches, through lenses that are new, unusual, never before seen or never before thought of. Artificial intelligence is never going to have that ability, precisely because it is artificial. Humans are also always going to be needed to devise the training programs that any new artificial intelligence would require to be developed through, and to test them. So, robots and computers wrenching away control from us is definitely not the direction artificial intelligence may take us over from.
Having got that clear, there are two broad directions from where artificial intelligence does pose a significant challenge to us. The first of these is in the matter of jobs, and certainly artificial intelligence does threaten jobs on a scale hitherto unprecedented in human history.
The use of artificial intelligence can aid in what economist Thomas Piketty sees as a factor which might potentiality decrease inequality: “diffusion of knowledge and investment in training and skills” (Piketty, 2017). It could be used to amplify/modify human labour’s skills, training and ability. But AI goes beyond such technology as aided in the amplifying of skills in the twentieth century. Existent factors of production, as we know from Smith, Ricardo and Marx, are ‘capital’ and ‘labour’. Artificial intelligence, though, has the potential to be a capital/labour ‘hybrid’, as a study by Accenture Strategy has argued (“Why Artificial Intelligence is the Future of Growth”, Accenture). It could replicate labour activities at a scale and speed that would mean a paradigm shift in labour.
This would seriously affect the distribution of income, wages, and inequality. A study hosted by Bruegel.org informs us that the low-skill, low-wage side of the labour market will be ravaged by new advances in artificial intelligence as “tasks previously hard to computerise in the service sector become vulnerable to technological advance” (https://www.bruegel.org/2014/07/the-computerisation-of-european-jobs/). It is likely that labour relations will drastically change themselves, with more frequent job changes and a surging rise in the levels of precarious work, self-employment and contractual work. This would undoubtedly weaken the avenues workers themselves have for lobbying for enhancement of their economic conditions: worker’s rights mechanisms and trade unions.
An increased demand for highly skilled workers who have expertise in artificial intelligence will likely see money being channelled into wages for them. This will reduce finances available for any social security schemes that governments might otherwise be willing to deploy to absorb low-skilled workers, causing severe wage cuts and rampant unemployment. Some might think this could be solved by ‘upskilling’. But the problem is governments might likely be tempted to favour national investments being utilised to further artificial intelligence above the upskilling of labour that otherwise does not have the economic means to upskill itself.
All of this would only exacerbate the fundamental problem that Thomas Piketty, for example, sees in the existent system of global capital. Put simply, Piketty’s argument is that ours is an economy where the rate of return on already put-in investment far outweighs the rate of economic growth. As such, for those whose families already have not amassed wealth by the previous century, life is likely going to be very difficult in the 21st century. This is to say, that that “rich kids can swan aimlessly… while the poor kids sweat into their barista uniforms — is not an accident: it is the system working normally” in this century (“Thomas Piketty’s Capital…”, The Guardian).
Artificial intelligence is going to exacerbate this issue on two counts. For one, jobs such as the barista ones are in themselves going to be taken over by robots who can perform such manual tasks easily. Secondly, the system will only develop further concretised and more firmly entrenched manifestations of family-owned superstar companies and sectors. This is because by dint of already having the capital to be able to harness the power of, or perform research and development in, AI, these will be best placed to leverage it.
However, we aren’t in any idealistic or utopian world and the genie of AI cannot be put back into the bottle. Once having been unleashed, it is now out and it is here to stay.
The Accenture study mentioned earlier performed analysis on 12 developed economies of the world to conclude that artificial intelligence has the potential to double their growth rates by 2035.
The first of three ways that Accenture sees artificial-intelligence based economic growth happening is through intelligent automation. A possible example of an environment where this could be practical is a dangerous warehouse or mill where so far, the implementation of technology has been unable to help with the basic human requirement of dexterity, as in manoeuvring through a complex space/site while avoiding equally complex obstacles. Robots could now help with this.
While this is task specific, the second implication of intelligent automation could be applicable across a diverse set of industries and job requirements. This is machine learning. An artificial intelligence platform that has read through a vast number of, say, technical manuals –something it could do much faster than the normal human– can now quickly diagnose problems entered into it and suggest solutions. Indeed, Accenture informs that an AI platform known as Amelia has learnt the one hundred and twenty questions most frequently asked by mortgage brokers and has been deployed in a bank to handle financial queries, something that would naturally make economic transactions less labour-intensive processes.
Secondly, artificial intelligence can enable augmentation. Without replacing existent labour, AI can enable labour to be re-routed to those avenues which are the most value-adding parts of their roles or job descriptions. In the hospitality industry, routine tasks such as room service being taken over by robots can enable staff to expend more efforts on maximising customer satisfaction.
The third way artificial intelligence is most fundamentally going to have an impact is through direct innovations. One may here see the wide range of eventualities stemming from just one innovation, driverless vehicles. Driverless vehicles, relying on a combination of GPS, cameras, radar, computer-vision and machine learning algorithms, can significantly cut costs per mile. By reducing wait times on commutes, they would also enable an increase in a workforce’s productivity. In spill-over effects, individuals freed from the onerous task of driving long distances would likely possibly be spending more time on the internet. This would lead to avenues of economic growth such as in an increase in the packs demanded from mobile service providers, also leading to ancillary possibilities like more advertising opportunities for the said providers and potential selling opportunities for their retail partners. Companies that offer streaming services too would benefit in this scenario.
Consequently, most people aren’t going to want artificial intelligence to be gone. Rather than indulging in wishful thinking, arguing a point that becomes moot and academic, or making unreasonable demands, therefore, it would probably be more sensible to try to bring pressure on our representatives such that they don’t get tempted by easy lures but instead draft policy and procedure to ensure protection of the rights of workers, implementation of a concrete and binding international legal framework for grievance redressal that workers could avail of, and establishment of individual government regulations. In addition, or at the least alternatively, something like the concept of universal basic income (the sort of idea Elon Musk vociferously promotes) needs to come into being.
Meanwhile, there’s another issue.
Someone told me the other day that replacing New York’s fleet of 13,000 yellow cabs with 9,000 driverless (i.e. self-driven) cars could cut costs per mile by nearly 88% and wait times down from 5 minutes to just 36 seconds during rush hour. And there wouldn’t be any cases of drunk driving either. Fine. But does that mean we can heave a sigh of contentment and let the skill of driving become a thing of the past in the way that the profession of the scribes went the way of the dinosaurs after the introduction of technology like the printing press?
I would argue the cases are not comparable. This is because risks of collision in the real world usually involve high speed vehicles moving at fraction-of-a-second velocities, and there may well arise many cases where instant lateral thinking skills and a flexibility of approach would be needed to avoid an onrushing vehicle- something that only a human driver could possess. Besides, if a glitch developed in the self-drive algorithm, you’d also quite simply need to have the skill of driving to be able to get back home from work (or the deserted wilderness you had driven to for some peace and quiet).
Robots and computers are not going to wrench control away from us, but we must take care to ensure that we do not become so drugged by the easy lure of artificial intelligence solutions that we relinquish all control of our own volition.
Without naming names, an artificial intelligence technology that called up someone and did all the talking and communication necessary for you is potentially something risky in an era where people are already so withdrawn and technology-using that countless statistical surveys show a gradual increasing lack in basic communicative ability and soft skills right across the board, from corporate atmospheres to teaching environments. We cannot afford to let AI tech be an opium that will deaden being intelligent and articulate.
One could also consider here scenarios where something life-threatening is involved. Suppose, say, an older adult not in full control of their faculties for whom you were a caregiver had a serious disease and it was up to you to ensure that they got pills of such and such a variety each day. Now further suppose that you bought a robot that could go over to a medication dispenser, get the appropriate pills, and bring them over to the loved one. The robot performed this task well for a long period, and you were satisfied. At the end of this robot’s utility period, you decided to purchase another robot. However, the next robot had a glitch in its software or algorithm. Erratic programming might very easily cause it to give inaccurate pills, resulting in potentially dangerous consequences.
This does not mean artificial intelligence cannot be allowed to help in life-and-death situations. Persons diagnosed with disorders of consciousness may, for example, using artificial intelligence tech like the brain/computer interface, be able to process information and make end-of-life decisions for themselves. Some persons who are diagnosed with DOC die as a result of end-of-life decisions, which may be made by family members who sincerely feel this is in the patient’s best interests. However, given the new prospect of allowing these patients to themselves provide their views on this decision, there would seem to be a strong imperative to further research on this aspect of AI tech and provide it with funding, to guarantee that DOC patients are at least given an opportunity to decide whether they want to live. Furthermore there are applicable benefits of the brain/computer interface in motor skills recovery after a stroke or injury, and in functional brain mapping as well. (There are ethical considerations here just as in other realms of therapy, but I believe that the pertinent standard protocols are well-equipped and can be implemented to ensure ethically sound informed-consent procedures.)
To what degree we should pass over control to artificial intelligence technology, then, must depend on every individual, particular, specific, unique situation. AI is not a threat to humanity itself, but the question of whether the benefits of artificial intelligence outweigh the risks has no easy, pat answer. What one can proffer as an attempt at a response is that as long as we retain the intelligent capacity to determine the outer limit to which we will agree to give up control in the co-existent partnership with AI, we will be alright. But if we aren’t careful, if we lose the faculty of basic intelligence and relinquish too much, then we could lose our own humanity- and then the risks of AI tech could very well end up outweighing the benefits.
Accenture. (2017, May). Why Artificial Intelligence is the Future of Growth. https://www.accenture.com/_acnmedia/PDF-33/Accenture-Why-AI-is-the-Future-of-Growth–Country-Spotlights.pdf
Mason, P. (2017, November 25). Thomas Piketty’s Capital: everything you need to know about the surprise bestseller. The Guardian. https://www.theguardian.com/books/2014/apr/28/thomas-piketty-capital-surprise-bestseller
Piketty, T. (2017). Capital in the Twenty-First Century. Digital.