Artificial Truth

It’s hard to believe we’re just fifteen months past the launch of ChatGPT; the day generative AI went mainstream. The tech behemoths of the world now race to make these systems ever more powerful, with little regard for safety & few, if any, guardrails in place. This has made lots of people very rich; Microsoft alone gained nearly $1 trillion in market value in the year following ChatGPT’s launch. Indeed, huge sectors of the economy will soon be handed over to AI; printing gold for profit-making actors who have long since externalised the risks. Of course, at this stage, we are all aware of AI’s negative impacts. We know it has eliminated jobs – and will eliminate many more. We know large language models inherit biases from the data they are trained on. We recognise the significant privacy & copyright concerns in terms of how this data is collected. We feel scared about a further reduction in meaningful human interaction & the ensuing social isolation. For today, let’s think more deeply about navigating truth in the coming months, review the recent New York Times lawsuit & briefly discuss the extinction scenario.

The concept of truth will soon be reshaped. This is not conjecture; it is fact that AI will usher in a new era of deception, where reality may be manipulated and distorted with ease. This past month, OpenAI launched Sora: a text-to-video model capable of generating convincingly realistic videos up to a minute long. Yes, there are flaws; they will be gone in a year or two, so let’s discuss the implications now. What happens when you hand humanity the power to create visual evidence of events that never occurred? Simply put, these tools will undermine the very fabric of society. Misinformation already proliferates at unprecedented rates through social media. We see our elections corrupted & far-right conspiracies spread. Once these bad actors may generate video at will – things will get so much worse. As socialists, we must reject the notion that truth is a commodity to be manipulated & exploited for personal or political gain. We must advocate for a comprehensive approach that combines technological innovation, regulatory oversight, education, and grassroots activism. We must raise awareness around the dangers of deepfakes, promote media literacy, and empower individuals to resist the corrosive influence of AI-misinformation. The time to raise awareness is now. I would encourage all of you to speak to those in your personal life & help them to traverse this digital age.

The New York Times is the first major media organisation to sue Microsoft and OpenAI for copyright infringement, accusing them of ‘unlawful copying and use of The Time’s uniquely valuable works.’ They show evidence of GPT-4 reciting NY Times articles word-for-word. It’s no small secret that the business model of AI is built on stealing content. This is how the systems learn: by taking the art of others & reproducing it. Backers of OpenAI will argue copyright slows development, hampering the technology, and harming their position as a global leader. We are left with a situation where the AI models set to replace journalists, media, and artists must first ingest their work in order to exist. Time will tell who wins out.

Many of us dismiss the extinction narrative as overhyped, but it is important to recognise the recursive nature of AI development. Recursive self-improvement refers to the process by which an early AI system may enhance its intelligence without human intervention, leading to an explosion in capability. Deep learning is uniquely good at this, but it is also uniquely unexplainable. AI models learn highly intricate relationships within their datasets & complex behaviours such as image or language generation emerge, but the inner workings of how these relationships are formed are ultimately opaque to human understanding. The rate at which this technology improves will soon become uncontrollable. There are deep political & social questions at the very heart of this, but we are not engaged in solving them. We are poking at a petri dish and landing on outcomes based on a set of goals written by big tech. Do you trust Meta & Google to get it right? I trust trillion-dollar companies to ruthlessly optimise for profit, super-charge late-stage capitalism, bypass government regulation, and eradicate what remains of social fabric. We are not ready.