Part Two of our Jensen Huang decoding gets to AI safety, where Jensen gives the labs some simple advice: if your products aren’t safe, don’t release them! There we go. Glad we got that sorted.
Luckily Jensen knows a lot of CEOs and he’s confident they want to do the right thing. Good CEOs and good engineers, doing good work. Reassuring stuff, although Volkswagen presumably had a few of those too when it was cheating on emissions tests. We’re slightly less inclined to treat Jensen’s confidence in his rich friends as a safety mechanism, particularly when he does rather well out of selling them chips.
Though to be fair, he seems a little more open to independent auditors and new regulations than some accounts of the interview suggest. Matt also offers some qualified reasons for optimism about AI safety research. So there are some things to discuss beyond whether we should trust all the nice billionaires.
We also cover his "interesting" environmental argument. To save the planet, Jensen explains, we’ll first have to do it quite a lot of damage. It’s like surgery: an ENORMOUS amount of pain now, with the promise that things might possibly get better afterwards. In this case, perhaps some more solar plants. For operations usually they give you anaesthetics, but a lovely new park next to the data centre will maybe do the same job.
In short, there’s plenty of out-of-touch billionaire optimism here, and a convenient overlap between what’s good for the future/planet and what’s good for Nvidia sales. But even with all that, Jensen remains a bit short on the guru flair we usually encounter. C'est la vie!