UK spy chief warns of Al danger
The landscape of AI is one that is rapidly expanding. Dating back to Ancient Greece and Hephasteus’ power of creation, the existential threat and possibility of a matrix-style dystopia remain a part of current discourse. However, the nuances of intelligence systems are wide-ranging. They vary from self-driving cars and specialised healthcare to generative videos of cats riding an electric scooter.
The benefits of AI are numerous. In the field of healthcare we can see increased accuracy and safety in surgery and a greater emphasis on personalised, patient-focused treatments with less time being spent on administrative work. Jobs which are deemed dangerous: in construction or mining industries for example, are being taken over. Accessibility for people with mobility problems has risen with virtual assistants. These are only a few of the advantages which AI has to offer its relatively early stage of development.
While the potential for creating a more ideal society with AI remains, inequality is ingrainedly built into its usage. We see such discrimination manifest in areas of gender, race, and class. A study conducted by Nataliya Kosmyna of M.I.T., found that those who used Large Language Models to write essays had a weaker understanding of their arguments and displayed a reluctance to claim ownership of work they produced, stunting educational growth. In a climate where paid work is becoming a necessity to students, AI usage is crucial with resulting time pressures. The ‘malevolent seduction: excellence without effort’ as posed by David Brooks of ‘The New York Times’, will become all the more enticing. Inequality present in accessibility to education is further widened with private schools being three times more likely to have explicit policies on AI usage.
Due to the presence of human bias in information used by intelligence programs, algorithms perpetuate racism, despite McCallum, Director General of MI5 stating that, ‘artificial intelligence may never ‘mean’ us harm.’ When 28 members of the U.S. Congress were incorrectly matched with mugshots from a criminal data set, 40% of those mistakes made were of people of colour. China’s use of AI to identify Uyghurs - who they have been accused of committing a genocide against - have displayed a “persistent anti-Muslim bias.” Trends in discriminatory sentiment infiltrate the intelligence systems used on both local and national scales.
Arguments such as Sir Keir Starmer’s, believing that “no person’s substantive time should be spent on a task where digital or artificial intelligence can do it better” without the careful introduction of AI into the workplace, and legislation being put in place to protect workers from AI-related job loss, has resulted in up to 8 million UK jobs being put in jeopardy as found by the IPPR, with women being markedly more at risk. Again, and again we see the most vulnerable in society targeted by systematic oppression.
Who do we blame for these systematic failures? Who needs to be held accountable?
The opacity of artificial intelligence leads to a growing mistrust in autonomous AI systems. In the absence of transparency the mistakes, biases, connections, and networks of information used to solve obscure problems remain incomprehensible. However, the simplification that would be needed for a human to decipher solutions would mean less advanced and thus less useful AI-generated algorithms. This leads to increased exposure to hacking, thus breeding deeper mistrust and fear of uncontrollable privacy violations and autonomous systems (such as weapon mechanisms) being given violent instructions, leading to a less than favourable situation that is impossible to control.
These security risks posed by AI have led to rising anxiety surrounding potential privacy violations. In her study Dr Anna Maria Mandalari revealed AI assistants which continued to record online activity upon transitioning to a private space, validating relevant fears surrounding data collection.
Deepfakes escalate mistrust, impermeably altering the political sphere by creating a capacity for chaos and an escape from accountability. The near “serious disorder” that arose from an AI generated audio of Sadiq Khan - eliciting anger amongst the far right - or Trump’s widely disputed birthday message to Epstein, demonstrate the extent to which AI damages the politics, privacy, and reputation. Arguably, better relationships could be encouraged between politicians and voters, to increase trust in our leaders before fake news is spread. Overall, continued advancements in AI will only heighten the inability to differentiate between the real and the fake.
In conclusion the inevitable integration of AI usage should be approached with flexibility, empathy, and practicality. Both the UK Government’s current ‘AI Playbook’ and existing intelligence models should be regularly reviewed. Constant human evaluation is paramount in emphasis on responsibility, fairness, beneficence, and privacy. In this way, with careful and continuous regulation AI can become a crucial tool in creating a more inclusive and safe society.
By Amelia Owczarczyk
