Comment by Jeremie Harris

Yeah, well, I think it’s clear that we’ve taken some significant steps toward human-level AI, particularly in the last three years. I think that’s so much the case that, as we’ve just seen, many of the world’s top AI researchers — including two of the three earliest pioneers of modern AI — are wondering aloud whether we might have fully cracked the code on intelligence. I wouldn’t personally go that far. But this is genuinely being discussed through that lens, looking at whether the techniques we have today — simply by scaling them with more data and more processing power — could achieve something like human-level or even superhuman AI. If that’s anywhere close to true — even within the realm of possibility — then we need to contemplate some fairly radical shifts in the risk landscape this technology exposes us to, and what that means for society more broadly. There are many dimensions to that risk. One key dimension is malicious use. As these systems become more powerful, the destructive capacity of malicious actors who use them grows accordingly. We’ve already seen cybersecurity exploits and malware generated by people who don’t know how to code. That represents a fundamental shift in the landscape. Unverified source (2023)