Would anyone board a plane if half the engineers who built it gave it a 10% chance of crashing? If not, then why are we racing ahead with artificial intelligence? In a 2020 survey, half of 700 AI researchers surveyed put the odds of AI causing human extinction at 10% or greater and that was before the great leap forward of Chat GPT 3.5 and the even more dazzling and unsettling GPT-4. In the “Large Language Models (LLM) we have summoned up an alien intelligence that can digest every sentence, scientific study, image, song and artwork that ever existed and is online and regurgitate it back to us in transmuted form.
AI will soon be able to generate political speeches, ideological manifestos, holy books for new cults, and floods of misinformation that could erode the foundations of our civilizations. We have already seen how an unchecked explosion of social media two decades ago begat an America whose citizens can no longer agree on who one elections. This time, before we unleash the god-like powers of AI, let us agree through our government to slow things down and get this right. Something we’ll only have one chance to get right.
AI may not be able to “truly think” but it can reproduce the thoughts of men in sometimes distorted ways. Part of what makes it unnerving is that it is a black box to its own creators who do not fully understand how it probabilistic algorithms create its responses. What if we enlist some future LLM to stop global warming and with its limited ingenuity hacks into the worlds atomic arsenal to erase mankind: the source of global warming. Pumping the brakes on AI is essential and imperative at this moment but may not be possible in a very short period of time.
We subject new medicines to years of safety trials even as patients die waiting. We’ve managed to slow, if not stop, nuclear proliferation. And have, so far, kept a lid on gene-editing and cloning technology. We need to hit the pause button on AI and begin to consider the logic of our creation and ensure it cannot run amok.
Time is running out. LLMs will already advise us on building biological weapons if you “ask the question in a clever way. Their ability to converse in a convincingly human way has inspired some lonely people to give up dating and fall in love with AI Chatbots. And what’s coming in a few years will make today’s Chatbots “look like toys.” Google CEO Sundar Pichai promises, or warns, AI will transform human life more profoundly than electricity or fire. What will that mean? AI might come up with a cure for cancer and vastly accelerate scientific technological progress. But even the engineers who are creating it concede “they might summon demons” and a darker, less human future. With AI’s capabilities accelerating this fast “we do not have the luxury of moving this slowly in response