I think I more or less agree with Taleb, so I will try to make it more plausible.

• Doing good is hard (cf Givewell. "Famine? Let's send free food! Oops, we bankrupted local food producers. Oh well, our hearts were in the right place.")
• Consider the infinite Platonic set of Interventions (in an economy, person... whatever). Throw a dart inside that set - are you more likely to hit a useful intervention, or a useless/harmful one?
• Further problem: a lot of harmful or useless interventions LOOK useful, or are useful for some parties but very harmful for the rest of us.
• Further problem: many harmful interventions are harmful on a truly spectacular scale, even - or especially - if they are really popular and seem really beneficial and are totally going to change the world for the better. (Fat tails.)
• Further problem: humans love power, and a great way to get power is via some grand intervention. The people in charge of such an intervention probably don't have skin in the game, so they aren't incentivized to care very much about REALLY getting it right.

This suggests the HEURISTIC that there is more to be gained from stopping people shooting themselves (or each other) in the foot than there is from promoting people's happiness.

I'm pretty sure Taleb would agree it is only a heuristic, and that bednets are a legitimate counterexample & are in fact pretty great.