who you gonna believe, trump or your lying eyes?
(Originally published in the Western People on 2024-10-22)
AI-generated_Donald_Trump_wading_through_floodwaters, Wikipedia (unknown source, Facebook 30 Sep 2024)
Donald Trump’s face is set in concentration, wading through almost hip-high floodwaters on some overcast, unknown street. He appears to be listening to his companion, likely discussing the extent of the inundation. Both are wearing orange life-jackets as they trudge past stranded cars in the background. This photograph swept through social media, with an accompanying wave of support for the former president who — unlike his nemesis in the White House — was heroically on-the-ground helping with the effort to save lives in the states affected by Hurricane Helene.
There is only tiny one problem with all of this. The actual image is a complete fraud, created entirely using an A.I. (artificial intelligence) software programme. In fact, a close inspection of the ‘photograph’ reveals that Trump’s right-hand is malformed, looking more like a pig’s trotter. Yet, this didn’t stop it shared hundreds of thousands of times on Facebook, before eventually being flagged as ‘an altered photo’ — but still staying on the platform. Indeed many Republican-leaning accounts then claimed authorities were trying to censor the former President.[1] This image is but one of many generated by A.I. which is flooding the online space, with Trump himself pushing A.I.-generated ‘photos’ of women wearing "Swifties for Trump" t-shirts in a daft claim that the followers of mega-singer Taylor Swift were supporting him.[2] An Uber-driver (who was rather quickly taking me to an audition in West Hollywood) informed me that these ‘bullshit photographs’ are easily debunked with anyone in possession of half a brain – yet the eyes will see what the eyes want to see, especially as they hurriedly doom-scroll their mobile phone contents before falling into a fitful, night-time sleep invaded by monsters, zombies and Kerry football supporters (but that could be just me).
A.I. technology is set to have one of the most revolutionary impacts on our society’s development, alongside that of the internet itself, smartphones and the Mary Robinson centre. Yet, we are nowhere ready for it, with many in denial of how ubiquitous it will become or how detrimental / beneficial it will be to humans (A.I. I mean, not the Mary Robinson centre). But it is in the ignorance of this technology that perhaps poses the greatest threat. If you don’t know it exists, how are you aware of what it can do?
In very simple terms, Artificial Intelligence is the development of computer systems that can perform tasks typically requiring human intelligence, such as learning, reasoning, and understanding language. Meanwhile, Large Language Models (LLMs) are a specialised type of A.I. that focuses on understanding and generating human-like text through pattern recognition from analysing billions of human-made files of text, video, audio and images. Chat-GPT is probably the best known example of a ‘chatbot’ application allowing humans to interact with an advanced LLM and produce news articles and social media posts, create human-like support agents for websites, mix never heard before music tracks or invent images of a smiling Donald Trump with his arms around a group of shiny black women at a party, which was shared on Facebook by a prominent conservative radio host, Mark Kay.[3]
This image, created by a radio host and his team using AI, is one of dozens of fakes portraying black Trump supporters (BBC News website, 3 March 2024)
My Uber-driver had insisted that he would always able to tell the difference between what is real or created online, even as swore in Russian at the honey-voiced navigation system who directed his Tesla onto the wrong street, leaving us perched rather precariously facing oncoming traffic. Meanwhile, as I searched for ‘what is the least I should tip an Uber driver’, the google algorithm shoved me videos of ‘real’ smiling faces, claiming that we are on the cusp of A.I. robots replacing humans’ mundane work (while also suggesting I should watch more cat videos).
The sheer pace of A.I. advancement is staggering. Dario Amodei, CEO of Anthropic (a rival to Chat-GPT’s owner’s, OpenAI), believes in A.I.’s power to radically transform our world, while cautioning against threats like authoritarianism and job displacement. ‘I think that most people are underestimating just how radical the upside of AI could be, just as I think most people are underestimating how bad the risks could be’.[4] Yet does anyone really know what A.I. is? A recent study by a team of Apple researchers concluded that ‘that current LLMs are not capable of genuine logical reasoning; instead, they attempt to replicate the reasoning steps observed in their training data’. Essentially A.I. is not able to think like a human and stunning displays of A.I. ‘intelligence’ is not much more than pattern recognition, from being ‘trained’ on vast amounts of human-generated data.[5] In fact, the more complex the questions being asked of the A.I. the greater the LLMs' performance deterioration. But could I tell the difference? Could you? And what does it actually mean to think like a human?
‘Deep Thought’ is a fictional A.I. supercomputer from Douglas Adams' The Hitchhiker's Guide to the Galaxy. Created by hyper-intelligent, pan-dimensional beings, Deep Thought was designed to calculate to answer ultimate question ‘to life, the universe and everything’. But after forcing its creators to wait seven and a half million years to think on it, it finally gave the disappointing answer of “42”, while complaining that it wasn’t a real question in any case. Meanwhile at the end of Monty Python’s 1983 absurdist comedy The Meaning of Life, a female presenter glibly gives the answer to the title’s question as "It’s nothing very special: Try and be nice to people, avoid eating fat, read a good book every now and then, get some walking in, and try and live together in peace and harmony with people of all creeds and nations”.
Yet, when I asked the ‘Perplexity’ A.I. chatbot ‘what is the comedy sketch where a deity in statue form is asked the secret of the universe and gives a one word answer?’ It confidently replied: “The sketch you are referring to is from Monty Python and the Meaning of Life, specifically the segment titled "The Meaning of Life". In this sketch, a character approaches a statue of a deity and asks for the secret of the universe. The statue responds with a single word: "Six." This absurdly simple answer humorously contrasts with the profound nature of the question, typical of Monty Python's comedic style. The sketch plays on the idea that profound questions about existence can have unexpectedly trivial answers”. Indeed. It’s an unexpected bullshit answer, but sounds authoritatively true, until fact-checked, just like a photo of Donald Trump saving survivors from the post-hurricane floodwaters.
We are not educated or prepared for any of this and bad actors (no, not me) will take advantage of A.I. to radically alter our trust in democratic systems and poison belief in our government institutions, academia, scientific community, news-media and even each other – which was already at record lows with severe cultural and political polarisation. The Russian government waged a sophisticated disinformation campaign in the run-up to the 2016 Presidential election, creating fake social media accounts and pages to spread divisive content on topics like Black Lives Matter, immigration, and gun control, while running political ads criticising Hillary Clinton. It may have cost her the election. While there is no evidence of coordinated external use of A.I. to similarly target the 2024 campaign, someday, very soon, Donald Trump or Kamala Harris will be on video dramatically saying or doing something that they didn’t say or do — and no pig trotter hand will give it away.
[1] https://www.usatoday.com/story/news/factcheck/2024/10/03/ai-image-trump-hurricane-helene-fact-check/75483588007/
[2] https://www.cbsnews.com/news/trump-shares-fake-swifties-for-trump-images/
[3] https://www.bbc.com/news/world-us-canada-68440150
[4] Amodei, D. (2023). Machines of Loving Grace. Retrieved from https://darioamodei.com/machines-of-loving-grace [2024-10-14]
[5] Mirzadeh, S. I., Alizadeh, K., Shahrokhi, H., Tuzel, O., Bengio, S., & Farajtabar, M. (2024). GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models. arXiv preprint arXiv:2410.05229.