First, I want to point you to this article, which is the most upvoted thing on the EA forum: http://effective-altruism.com/ea/9s/effective_altruism_is_a_question_not_an_ideology/ You can still call yourself an EA even if you disagree with most of “mainstream” EA thinking on how to do the most good, and arguably these EAs are among the most valuable since they increase the diversity of perspectives in EA. (I think I’d probably put myself in this category of “EAs who disagree with most mainstream EA thinking”, FWIW.)
“Rather than trying to convince people to take these jobs and then try to do good, shouldn’t he be trying to convince people to not take these jobs in the first place?”
All it takes is 1 person to take the job and it will get done. Getting everyone to avoid taking a job is a difficult coordination problem. Salaries will go up until someone is willing to take the job. (I suspect that petroleum engineering is the highest paid engineering specialty because it contributes to global warming, and many people avoid it for this reason.)
And the flip side is that if nonprofits have a boatload of money, they will likely not have a hard time finding quality people. Even if this “disaster” were to occur, it’d be easy for the EA community to change its recommendations at that point.
Overall I think this issue is a complicated one and there are EAs on both sides, see e.g. http://effective-altruism.com/ea/sz/being_a_tobacco_ceo_is_not_quite_as_bad_as_it/ https://80000hours.org/2015/11/why-you-should-focus-more-on-talent-gaps-not-funding-gaps/
The broad principle here is to think on the margin. The way to do the most good is not to ask “what would the world be like if everyone was to copy my exact actions?” It’s to ask “how can I as an individual do the most good for the world on the margin?” And the right answer to this second question will change depending on how others are behaving and what areas are neglected.
And yes, thinking on the margin implies, for instance, that it’s more valuable to vaccinate the last person might catch a disease than the first.
There’s also been some pushback against using marginal evaluation within the EA community: https://80000hours.org/2015/11/take-the-growth-approach-to-evaluating-startup-non-profits-not-the-marginal-approach/
Re: quantification. Yes it’s difficult, but I’d argue that attempting to quantify something is almost always going to get you a superior result than not even making an attempt. https://80000hours.org/2011/11/estimation-is-the-best-we-have/
Re: time scales. IMO one has to strike a balance between short and long term interventions. I lean towards longer-term interventions myself. But I’d argue that something like an art museum is probably pretty low impact even if considered as a long term intervention.