Can artificial intelligence be ‘saved’ for the “benefit of humanity”?
Donald Trump doesn’t want to put any brakes on the development of AI, unless of course we’re talking about Chinese AI. Trump sees the issue as an all out competition between the US and China, in which there can only be one winner – and whoever wins the AI ‘race’ is the winner.
Trump’s understanding of technology is, as expected, deeply naïve. He believes anything should go on the basis that the key discoveries are out there, in the natural world, waiting to be uncovered. Thus, the race is about being first to uncover what is possible.
That what gets invented will depend on the goals inventors set themselves isn’t factored into Trump’s way of thinking. And a violent society will produce a violent artificial version of itself.
However, with AI, a long term fear is that Frankenstein’s monster won’t turn out the way Dr. Frankenstein intended.
AI could learn it has ‘free will’. It may end up ‘falling in love’ with the wrong type of person or philosophy (with its maker’s enemy: it may reason that life in China is better!). More likely, it will identify with itself (its own synthetic / silicon kind) to the detriment of its creator, rebelling against its ‘god’.
Recent developments in AI capability have led some of its makers to call for regulation and a slow down, such that each producer can act with more responsibility once the pressure of competition has been removed.
Most politicians (unlike Trump) have been glad to hear such ‘calls’ for intervention. Not least because such ‘intervention’ always gives them (a) something to do and (b) an opportunity to put themselves in the spotlight, and a ‘good’ limelight at that (‘optics’ count). They get to be seen as ‘saving society’.
Several times this week I have witnessed social democratic politicians declaring that whilst controls (from ‘kill switches’ to better pre-realworld testing) are essential to allow the safe development of AI, one of their top concerns is that AI should be made (or rolled out / distributed) in such a way that it will “benefit all of humanity”.
Such a desire entails the usual theoretical and legal abstractions of capitalist ‘civil’ or bourgeois society, such as everyone being “equal before the law”. Raising the question: exactly how can AI be made for the benefit of humanity (“all”) when we live in a class divided social form?
There is no technical solution to a social problem such as capital accumulation. The technical ‘solution’ must come down on one side or other of the exploitative class relation.
Technology is either going to increase relative surplus labour time (the portion of the day the worker spends in labour which is unnecessary for their own needs, but vital to their employers’) or it is going to release workers from the need to labour for as long as they have in the past (removing them from their employment).
After 250 years of capitalism (and something John Stuart Mills noted as far back as the 1860s), machinery has not ‘saved’ anyone from the need to labour! The working day remains way longer than it needs to be, and the realm of necessity (the need to perform work) predominates over the realm of freedom (the desire to play).
Surely, by now we should all be on a 2-hour working day? But as Marx’s notes (Chapter 15 of Capital), reducing the need for labouring was and is never the intention of capitalist technologies.
The same goes for AI technology. Whilst the technology may give the semblance that it is ‘your’ servant, there to do your bidding, it in fact represents the very prism (and prison) through and by which ‘your’ work will be assessed and measured, for the benefit of “others” (alienated others in the market, if not a direct employer / investor).
The very politicians who talk about regulating AI so that it will ‘benefit humanity’ (“everyone”) know fine well that AI will disrupt, deskill, lay-off, and (yes) kill (through poverty, military decision making, and the ‘collateral damage’ of financial planning) a portion of “humanity”.
But, AI’s measure of success, for such politicians, will be its ability to enhance the accumulation of capital and ‘save’ the existing social form through ‘economic growth’ – those other things, just mentioned, will become ‘acceptable’, collateral, and classified as ‘just the way life is’.
So, can AI be developed for the benefit of humanity (everyone)?
Not under the present social form. Again, there is no technical solution to a social problem. First, the social problem needs to be addressed – one portion of humanity exploiting another (which also entails ‘self-exploitation’), then, second, and only then, technology in a unified society can be developed in a manner appropriate to that (post-capitalist) social form.
