AI will kill us all – and then do what, exactly?
In my short story AI.GOV, I posit that no intelligence will ever surpass that of its creator. Nothing ChatGPT or any other generative model has done since has changed my opinion. The pleas for legislation to limit AI development have far more to do with corporate politics and land-grabbing than with any genuine threat.
These models look brilliant and, sure, they ace the SAT and the bar exam. They’d better, because they were trained on all of that data, the equivalent of a student cramming with every tutor and every past exam available and then acing the test. The irony is that in spite of ingesting the sum of humanity’s written content, these same models still turn in dismal performance on other, comparably structured tests.
Here’s the framework I’d use to cut through the doomsaying, and it’s the CxO-relevant point. Any fair comparison needs to be unit-of-intelligence to unit-of-intelligence. One AI instance running on a single machine, disconnected from cloud infrastructure, against one human, both exposed to the same information. When people warn that AI will outsmart all of us, they’re quietly comparing one human against massive, internet-scale compute. That’s not a fair fight, and it’s not the fight that will actually happen inside your organization.
Three other data points worth holding onto. AI is trained on human-generated content, and as more humans lean on AI to produce that content, the quality of what it learns from will degrade, not improve. Humans are not standing still either. We’re already upgrading our own skills and understanding of how to work alongside and improving the systems we’ve built. And AI is likely to be the catalyst for genuinely personalized education-at-scale, personalized-medicine and other uses, which are a bigger unlock for human capability than most of the doom narrative accounts for.
I said on LinkedIn a while back that until AI makes the full journey from information to knowledge to experience to intuition, there’s no reason to worry about it taking over. It’s still at step on – processing information. Worth tracking how fast it moves and worth building your workforce strategy around that timeline rather than the fear mongering headline.