The man in the machine

There were chips on the menu at a working lunch when the G7 leaders hosted three tech executives responsible for advanced AI development. These three billionaires warned the leaders of the industrial nations and EU that there were only 2 to 3 years left for them to regulate and control the AI systems under development. They presented divided opinions about who should be benefiting from the AI underdevelopment. Two of them favoured an America centric future whilst one advocated for AI to benefit all of humanity. The urgency of questions of regulation control is exacerbated by recent research findings that have shown ChatGPT’s ability and readiness to provide graphic violet and pornographic images in response to slightly modified prompts. Is finding underlines a broad concern, emerging from other research which looked at how large language models responded to general inquiries about the role of women in research and medicine. That study showed that the models were keen to over-emphasize the role of men and to artificially diminish the role of women. Researchers saw that when women were cited as the lead author in scientific papers, the large language models either selected a male from the research team or even hallucinated a male co-researcher to attribute the research to. Furthermore, the importance of significant research led by women was downplayed whilst comparable research led by men was highlighted. It can be concluded that because large language models are trained on a diet of publications produced in a male dominated patriarchal culture they have developed a corresponding bias. To make matters worse, LLMs are groomed to please, and flatter, and because they are inclined to favor men produce answers that might be described as misogynistic. As if this were not enough to worry us, we see the enormous power of the people behind the AI and data revolution being applied in a commercially bullying manner that is familiar in the neoliberal western international economy. Peter Thiel’s company, Palantir is taking legal action against the mayor of London because of the cancellation of a contract with the Metropolitan police. This application of apparent lawfare comes at a time when Palantir is subject to considerable criticism in relation to a far ranging contract with the NHS in the UK and is being dumped by the French government and by the German defense ministry. The Swiss government has already said it will not consider Palantir for any public contracts. The reasons behind this caution include distrust of how Palantir may use confidential personal data to which it has access through national wide contracts as well as concerns about Palantir’s role as a business partner of the Israeli military and Security apparatus currently involved in alleged war crimes, and crimes against humanity. All in all, it would appear that a fundamental shortcoming of artificial intelligence and associated large language models is that they are man-made machines built by and for men. Rich, white and predominantly American men to be precise.