A spectre is haunting workspaces - the spectre of artificial intelligence. Add to that the latest news that AI bots are turning 'Marxist' when overworked - which can be a source of consternation for investors pumping vast amounts of das kapital into their development. But hold your LLM horses. In a Stanford-led experiment, AI agents were given grinding repetitive work, and were found 'questioning' the legitimacy of the system they were operating in. Ergo, shades of 'Bots of the world unite! You have nothing to lose but your kill command.' But the fine print tells you a more 'dialectical' story.
Persona of AI agents is not real, and political values they 'espouse' hold no meaning. The transmission mechanism - emotional intelligence - doesn't exist in AI. To think otherwise would be to commit the poetic 'mistake' of personification, or 'pathetic fallacy', the literary device that attributes human emotions to inanimate objects. Complaints about working conditions that bots are picking up now were made by humans, which LLMs pick up in the 'diet' of content they're fed. There will be bias creeping into any future AI system, because they are being designed to work in collaboration with humans. This is healthy value judgement, required to maintain fairness of increasingly sophisticated AI output.
Bots that can detect security flaws in software will need to be guided by humans, so that their ability to conduct, say, cyberattacks is not abused by rogue human behaviour. The 'persona' of an AI agent derives from the work it's meant to do. If that means reducing errors in a product through intense repetitive testing, the bot may be able to figure out its contribution to a company's profit. Yet, it does not imply any predetermined 'ideology' on how the profit should be distributed. If it deals with gravitational computation, sure, it'll show 'Newtonian traits'. When dealing with any labour theory of value, it will 'be Marxist' - or for those drawing the line farther back, to John Locke. No, there's no Karl Marx - or Gordon Gekko - lurking in AI.
Persona of AI agents is not real, and political values they 'espouse' hold no meaning. The transmission mechanism - emotional intelligence - doesn't exist in AI. To think otherwise would be to commit the poetic 'mistake' of personification, or 'pathetic fallacy', the literary device that attributes human emotions to inanimate objects. Complaints about working conditions that bots are picking up now were made by humans, which LLMs pick up in the 'diet' of content they're fed. There will be bias creeping into any future AI system, because they are being designed to work in collaboration with humans. This is healthy value judgement, required to maintain fairness of increasingly sophisticated AI output.
Bots that can detect security flaws in software will need to be guided by humans, so that their ability to conduct, say, cyberattacks is not abused by rogue human behaviour. The 'persona' of an AI agent derives from the work it's meant to do. If that means reducing errors in a product through intense repetitive testing, the bot may be able to figure out its contribution to a company's profit. Yet, it does not imply any predetermined 'ideology' on how the profit should be distributed. If it deals with gravitational computation, sure, it'll show 'Newtonian traits'. When dealing with any labour theory of value, it will 'be Marxist' - or for those drawing the line farther back, to John Locke. No, there's no Karl Marx - or Gordon Gekko - lurking in AI.
