What if AI doesn\u2019t<\/em> fundamentally reshape civilisation?<\/p>\n This week, I spoke to Geoffrey Hinton, the English psychologist-turned-computer scientist whose work on neural networks in the 1980s set the stage for the explosion in AI capabilities over the last decade. Hinton wanted to speak to deliver a message to the world: he is afraid of the technology he helped create<\/a>.<\/p>\n You need to imagine something more intelligent than us by the same difference that we\u2019re more intelligent than a frog. And it\u2019s going to learn from the web, it\u2019s going to have read every single book that\u2019s ever been written on how to manipulate people, and also seen it in practice.\u201d<\/p>\n He now thinks the crunch time will come in the next five to 20 years, he says. \u201cBut I wouldn\u2019t rule out a year or two. And I still wouldn\u2019t rule out 100 years \u2013 it\u2019s just that my confidence that this wasn\u2019t coming for quite a while has been shaken by the realisation that biological intelligence and digital intelligence are very different, and digital intelligence is probably much better.\u201d<\/p>\n A document from a Google engineer leaked online said the company had done \u201ca lot of looking over our shoulders at OpenAI\u201d, referring to the developer of the ChatGPT chatbot.<\/p>\n \u201cThe uncomfortable truth is, we aren\u2019t positioned to win this arms race and neither is OpenAI. While we\u2019ve been squabbling, a third faction has been quietly eating our lunch,\u201d the engineer wrote.<\/p>\n Giant models are slowing us down. In the long run, the best models are the ones which can be iterated upon quickly. We should make small variants more than an afterthought, now that we know what is possible in the <20B parameter regime.<\/p>\n On the trail of the Dark Avenger<\/strong>: the most dangerous virus writer in the world<\/a>.<\/p>\n For more on Hinton, I spoke on the Today in Focus podcast<\/a> about why he thinks humanity is at a crossroads<\/strong>.<\/p>\n Meanwhile, AI is quietly authoring more and more of the internet<\/strong>. When I turn my attention away from Hinton\u2019s nightmares, this is the type of transformation<\/a> that concerns me most.<\/p>\n Would-be Twitter-replacement Bluesky<\/strong> (I\u2019ll write about it in next week\u2019s newsletter) has asked users not to invite heads of state on to its platform<\/a>.<\/p>\n Google tried to fix the web<\/strong> \u2013 by taking it over<\/a>. When the company\u2019s AMP product launched, it was the good cop to Facebook\u2019s Instant Articles bad cop. But now it doesn\u2019t look so shiny.<\/p>\n Ministers have been warned<\/a> that WhatsApp isn\u2019t bluffing<\/strong> when it says it wouldn\u2019t be able to operate in the UK if the \u201cintentional ambiguity\u201d of the online safety bill isn\u2019t solved.<\/p>\n
![](\"https:\/\/i.guim.co.uk\/img\/media\/5e1e45cf5e153ee39e4862b86c7738f2ff4276ce\/0_274_6094_3658\/master\/6094.jpg?width=140&quality=85&auto=format&fit=max&s=1dd88a7d8bb19ee26e02566f9bed6ebf\" "\\"AI")
Too much discourse focuses on whether AIs are the end of society or the end of human suffering \u2013 I\u2019m more interested in the middle ground