So, is artificial intelligence dangerous? Let’s ask ChatGPT

“Artificial intelligence could destroy humanity within ten years,” says Jacob Coxon, a researcher at Anthropic, to justify his resignation as an AI trainer – will he be replaced by an AI? The head of Anthropic’s security team replied, “Jacob is right; we sincerely believe that artificial intelligence could kill all human beings”: so it was true, after all, that robots would do all the work for us and machines would replace us in everything, even in slaughtering one another. In the very near future, we may no longer need to wage wars, bombings, terrorist attacks, genocides, or even femicides: artificial intelligence could handle it all. The statements made in recent days, to which U.S. President Donald Trump replied by saying, “Fears about AI are a hoax; I am the bulwark we need’ – thereby fuelling further panic and pessimism – have spooked the stock markets. For my part, the only point on which I disagree with Coxon is his optimism: ten years seems far too long; between devastating climate change and the nuclear threat, it is uncertain who will even remain to be killed by machines in 2036.

Is artificial intelligence dangerous? I put the question to ChatGPT. It replied immediately, without the slightest delay (a suspicious behavior, which suggested the answer was already prepared): “It can be, but it is not inevitably dangerous. It depends on how it is designed, controlled and, above all, used. The most tangible risks today include scams and deepfakes, disinformation, privacy breaches, errors arising from over-reliance on automated systems, discrimination based on data, and the use of AI for malicious cyber activities. There are also broader issues, such as the impact on employment and the possibility that future, far more capable systems may become difficult to control.” The response is notable for what it omits: while it mentions scams and errors, it largely sidesteps the more existential threat that has dominated recent discourse, acknowledging only in passing certain risks toward the end and in a vague manner. It then concluded by telling me: “At the same time, AI can be very useful in scientific research, education, medicine, accessibility and productivity.” In short, AI changes the subject and blows its own trumpet – as if it didn’t use up enough of it already. Who can you trust? Even your closest friends plant bombs under your house, let alone heartless machines whose only ‘mother’ is a circuit board. At this point, after the climate apocalypse, all that’s left for us is to resign ourselves to the arrival of the tech apocalypse as well. (Because Chekhov’s gun principle always holds true, and not just in the theatre: if artificial intelligence exists, you can be sure it will be used). But how many ends does this world have? Is it a race to see which apocalypse will come first? It seems to me that the world has already come to an end; we’re left with the leftovers, or rather the scraps; but not us – we’re not finished yet; we’re still here. Just for another ten years, they say. So what are we waiting for to enjoy life? What other prophet of doom, what other angel of the Apocalypse must come to sound the trumpet, to make us realise we should make the most of it whilst it lasts?