It has been said that “the underlying purpose of AI is to allow wealth to access skill while removing from the skilled the ability to access wealth.” This is the most fundamental problem we need to address with AI.
We are seeing this years-old quote become more and more true as corporations try to replace labour with artificial intelligence systems, with varying degrees of success. Do not mistake their short term failures for a lack of desire to make this work.
Much of the commentariat is predicting a market collapse for over-hyped AI companies. 25 years ago, the “dot-com” bubble burst, laying waste to over-valued websites and crashing the short-term value of some survivors. What it did not do was end the dot-com’s revolution of our economy and society. It acted more as an extinction event, allowing only the economically fittest to survive, serving to corporatise rather than democratise the internet.
Similarly, even if AI causes a multi-year market crash, it will not kill AI; it will make it vastly more efficient. Overhyped players will collapse. Ones with solid fundamentals will survive. It may not be the big ones we see, it may not be in the form we expect, but AI is here to stay. The environmental problems around energy usage and the vast amounts of water required to cool the immense data centres at the core of AI technology will be solved if the alternative is the total failure of the AI economy. The same applies to the problems of recursive learning, where training material is increasingly itself AI-generated, causing AI-produced material to be, for want of a better term, inbred.
Large language models (LLMs) are destroying peoples’ ability to think, to read, to write. These are the systems that have taken much of the available written material in the world, including visual arts, and used it to train on the structure of language and imagery to generate correct-sounding responses to queries. According to internal documents from Microsoft, it is the largest theft of labour in human history, using lifetimes of copyright material to come up with plausible answers and information.
These AI models are not built around accuracy. Because it works on the basis of “this query should generate a response that sounds like this,” we encounter significant problems of so-called hallucinations. The fundamental engineering structure is not geared toward “this is a factually accurate response to the request.” Rather than being credible, it merely sounds credible, and offers a false sense of confidence to its users.
AI is being used by teachers to generate assignments and examinations and by students to complete them. A large segment of our education system is collapsing into who can write the best prompts, rather than who can perform the tasks themselves.
Employers are using AI to filter job applicants, interact with customers, and replace employees in a wide variety of tasks. Entry level employees are being replaced with AI systems. Senior staffers who rise from the ranks of junior staffers will eventually cease to exist because those entry level staffers were replaced by AI systems and were never hired and developed. Employees themselves are using AI for shortcuts. Everyone thinks they are being clever and saving money.
In the long term, people capable of thinking without AI assistance will become less common. University degrees will be even more meaningless than they have already become. As AI replaces genuine learning, even many people in the STEM fields will not have the skills their degrees ostensibly represent.
Corporate boards will realise that their most expensive employees, CEOs, senior executives, divisional vice presidents, and other high priced decision-makers can be replaced with AI and probably far more efficiently so than lower level employees. AI systems can synthesise data and make decisions with greater efficiency than any chief executive ever could, and can map out a future and analyse share price trends with far more granular detail than any person ever will. Yes boss, they’re coming for you, too.
In other words, there is a lot of pain ahead. But AI, which is not limited to LLMs, is going to become a permanent part of our economy. It is not that we are not already familiar and comfortable with it. We have been using primitive proto-AI systems like spell and grammar check in our word processors for decades.
A generation ago, the United States Navy stopped teaching celestial navigation to its officers, believing it to be obsolete thanks to modern technology like the Global Positioning System. About 10 years ago, the Navy realised that this was a terrible idea, and began, once again, teaching navigation by the stars. The realisation that GPS could fail, that technology in a war setting may not function as expected, has forced the world’s most powerful navy to go back to teaching the basics. Our education system will get there, too, eventually.
All of this begs the question: if AI can replace vast swaths of the economy and human productivity is no longer fundamental to economic productivity, what will happen to the economy, and to us as participants?
There are only two long-term outcomes possible. They are mutually exclusive.
On the road on which we are now traveling, we will fulfil the ultimate goal of unrestrained capitalism: all of the world’s assets will be owned by a very small number of people who control the means of production which, in turn, will be controlled by the automated systems of AI. The rest of humanity will live in increasingly dilapidated shanties, their labour no longer even being required for the comfort of the ruling class. In this dystopian world, AI will be militarised to control the people who no longer have an economic function. AI is not inherently violent, but it is certainly capable of violence. Like the people it seeks to emulate, AI systems will behave according to their upbringing.
It is beyond even the feudalism that the oligarchs have dreamed of and have been pursuing through a half-century of neoliberalism. The only people needed by the oligarchs in such a world are a small band of working class people whose job it is to keep control of the AI systems. Everyone else will simply be superfluous, and the oligarchs do not see a problem with that.
That is the very manifestation of unrestrained capitalism’s end-game.
The other option is the one that we need to pursue, and it will require more than a few essays written without the help of a large language model to achieve.
With the decoupling of productivity from labour, AI has to be treated and pursued as the great democratiser. AI is not inherently bad, but it is only as good as the ethical boundaries within which it lives.
Artificial Intelligence cannot coexist with the full human population in the long term in the absence of Universal Basic Income (BI). If AI can take care of all of our actual needs, then the productivity of the AI-based economy must be shared with all of its participants. We are currently headed for the former, but very much away from the latter. If work is entirely elective because everything is done by machine, then the results of that productivity must be distributed among the population.
Those who choose to work, who continue the skills of our species, who opt to learn the celestial navigation over the GPS as it were, can do so in that reality. In a world of AI backed by a world of BI, the automated systems will augment our society and our quality of life rather than destroying it. Those who wish to pursue purpose or riches could do so, but those who merely want to exist could do that, too.
We are rapidly approaching a crossroads. To survive the oligarchy in a world where productivity is decoupled from labour, we cannot have AI without BI.