Artificial Intelligence will bore us to death before it kills us

Now that MI5 has been called in to protect us from rogue AI systems and Anthropic has opted not to release its Claude Mythos model to the public over safety concerns, it seems that talk of AI erasing humanity is not entirely overblown. But before the world ends, Artificial Intelligence poses another grave threat to humanity: it risks boring us to death. Lawyers are the first victims of the wave of AI slop that risks drowning us all Lawyers are among the first victims of the wave of AI slop that risks drowning us all. The Financial Times reports that clients are bombarding their lawyers with letters and emails generated by AI chatbots. One partner at a US firm told the paper he had received so many AI-generated emails that he could not keep up with the ‘barrage’. As ever, it’s hard to have too much sympathy for overpaid lawyers, not least because some firms have already laid off trainees and paralegals over the supposed efficiency benefits AI was going to deliver for them. But I know how these lawyers feel. My day job analysing geopolitics has also been made harder, rather than easier, by AI. Normally, I spend hours a day wading through papers on subjects like migration routes or the international traffic in illegal goods. It can be pretty fun if you like reading and hearing about exotic places. Or at least it was until the dawn of ChatGPT. Now many of these papers have tripled in length and increasingly consist of pages of rather dense text which obscure rather than illuminate the subject. It’s no longer the fabled em dash which is the giveaway of an AI model; if you come across a piece of writing that is overly long, and which contains fatuous and quite bossy bullet points to compensate for the confusion, the chances are that it was probably the machine what wrote it. As well as wasting the readers’ time, those who use AI to write are cheating themselves: cutting and pasting a computer-generated answer surely misses the fundamental point of writing: to impart knowledge, through both demonstrating and communicating a deep understanding of a given subject. What is the point if you shortcut the bit of actually having the knowledge in the first place and rely on a bot which spouts out generic gobbledygook? The headaches and boredom caused by reading AI writing isn’t the only problem. There is also a striking feeling of being cheated after wading through a document and realising the mismatch between your effort of reading it and the lack of effort which went into writing the thing. I’m now beginning to wonder if our appreciation of something stems from the intuitive knowledge that years of skill have gone into something ephemeral. It is the essence of Picasso’s point, “It took me four years to paint like Raphael, but a lifetime to paint like a child.” Music is surely more than playing notes in the right order. Writing is more than just “Words, words, words.” The pernicious effects of the deluge of AI slop go beyond the arts. In December, Nature raised the alarm on studies using AI generated peer reviews to give reports a false mark of authenticity. Researchers are also becoming increasingly worried about the effect of AI-brain. A recent study showed that ChatGPT use in one group reduced brain activity by 55 per cent. Instead of AI offering a future like The Matrix, where people can instantly upload all manner of exciting skills instantly into our heads, our AI future increasingly looks to be one of cognitive decline. In my attempts to understand this brave new world that venture capitalists, Silicon Valley nerds and supine governments are foisting on us, I have read through what I initially considered to be the thoughtful – if not a bit overly long – essays of Anthropic’s CEO, Dario Amodei. Wading through 19,000 word essays with portentous titles like ‘Machines of Loving Grace’ and ‘The Urgency of Interpretability’, is not something I’d recommend. Buried in there was a warning that AI could make it easier for criminals or terrorists to get their hands on killer bioweapons. Perhaps this was why Mrinank Sharma, Anthropic’s head of AI safety, decided to give up on staving off the apocalypse to read poetry instead. Sharma’s decision is telling. If brevity is the soul of wit, then pages and pages of auto-generated content would indicate that AI ain’t got no soul.

View Original Article