You're missing the second part of his argument. Serendipitous discoveries can come from anywhere, so it does make sense to pay attention to cost. Which would yo

You're missing the second part of his argument. Serendipitous discoveries can come from anywhere, so it does make sense to pay attention to cost. Which would you think would be a better bet for a random win, 100k separate experiments, or just one large experiment that's a scaled up version of something we already have? If you say the second, you have severe scope neglect: 100k is a huge number. Of course, one expensive experiment may be the right call in some situations, but only if you have good reason to think it'll pay off. Either because you're certain it will, or because the expected utility of unknown results is larger than the cost (expected utility being probability times value). This very much isn't the case here.