As the demotivated person you referenced, I appreciate this! But I think the case for de-motivation is a bit stronger than presented in this calculation, so I’ll try to steelman it.
First of all, consider the framing: you’ve assumed that if I don’t get this fellowship I’ll continue with the job I already have, and that if I don’t apply I’ll just have leisure time. In reality many applicants will be looking for a new opportunity, and if they don’t get this they’ll probably take something else. They might (as I am) be making as many applications as they have energy for, such that the relevant counterfactual is another application, rather than free time.
Here are some more specific points:
I agree that the base rate is a poor guide to your personal probability. But I would expect that the range of probabilities across applicants is likely to be skewed, such that a number of applicants each have a very significant chance, while the majority of others have next to no chance at all. This would be the case if the hiring team are willing to accept only exceptional candidates, and there are more exceptional candidates than positions, yet the majority of applicants are not exceptional. I expect this skew to be stronger the more applicants there are.
If you are going to consider the intangible value of the fellowship, then you also need to consider the intangible value of the counterfactual. This will either be that of your current work, or another job you might take (which is hard to estimate, but you could guess at it).
Similarly to the above, if you’re going to consider the information value of applying, then you also ought to consider the information value of applying for whatever else you might have spent the time on (if that’s what you would have done). Since many fellowships receive thousands of applications, I presume that most applications get not feedback (as in my own experience). The information value is therefore likely to be higher for opportunities with higher chances of success. Interviews give a greater information value still, so this effect is quite strong. This might even be the main reason why I prefer not to apply for positions where my chance seems so slim.
A related advantage of focusing on higher-probability positions is that you might be lucky enough to receive multiple offers around the same time. That lets you compare options after learning more about them during the process, and choose the one that seems like the best personal fit. If most of your applications are very low-probability, you’re more likely either to accept the first offer you get, or to pass on a good offer in the hope of a better one later, simply because the relevant information arrives at very different times.
Lastly, there is a psychological cost to firing out many low-chance applications. Rejections are, after all, demotivating. Getting interviews or offers is highly motivating, even if you end up declining the offers.
This leads to a much much more complicated equation! I asked GPT, to try to link this all together, and well… I’ll just link its answer here.
So I remain pretty unsure as to whether it’s worth it for myself personally, but I might put in an application anyway :)
One general point: My rough guess is that acceptance rates have stayed largely constant across AI safety programs over the last ~2 years because capacity has scaled with interest. For example, Pivotal grew from 15 spots in 2024 to 38 in 2025. While the ‘tail’ likely became more exceptional, my sense is that the bar for the marginal admitted fellow has stayed roughly the same.
They might (as I am) be making as many applications as they have energy for, such that the relevant counterfactual is another application, rather than free time.
The model does assume that most applicants aren’t spending 100% of their time/energy on applications. However, even if they were, I feel like a lot of this is captured by how much they value their time. I think that the counterfactual of how they spend their time during the fellowship period (which is >100x more hours than the application process) is the much more important variable to get right.
you also need to consider the intangible value of the counterfactual
This is correct. I assumed most people would take this into account (e.g. subtract their current job’s networking value from the fellowship’s value), but I might add a note to make this explicit.
you also ought to consider the information value of applying for whatever else you might have spent the time on
I’m less worried about this one. Since we set the fixed Value of Information quite conservatively already, and most people aren’t constantly working on applications, I suspect this is usually small enough to be noise in the final calculation.
there is a psychological cost to firing out many low-chance applications
I agree this is real, but I think it’s covered in the Value of Your Time. If you earn £50/hr but find applying on the weekend fun/interesting, you might set the Value of Your Time at £5/hr. If you are unemployed but find applying extremely aversive, you might price your time at e.g., £200/hr.
I also have the impression that these fellowships are getting more competitive each year. Do you share that perspective? If so, that would require a further adjustment to the calculation.
As the demotivated person you referenced, I appreciate this! But I think the case for de-motivation is a bit stronger than presented in this calculation, so I’ll try to steelman it.
First of all, consider the framing: you’ve assumed that if I don’t get this fellowship I’ll continue with the job I already have, and that if I don’t apply I’ll just have leisure time. In reality many applicants will be looking for a new opportunity, and if they don’t get this they’ll probably take something else. They might (as I am) be making as many applications as they have energy for, such that the relevant counterfactual is another application, rather than free time.
Here are some more specific points:
I agree that the base rate is a poor guide to your personal probability. But I would expect that the range of probabilities across applicants is likely to be skewed, such that a number of applicants each have a very significant chance, while the majority of others have next to no chance at all. This would be the case if the hiring team are willing to accept only exceptional candidates, and there are more exceptional candidates than positions, yet the majority of applicants are not exceptional. I expect this skew to be stronger the more applicants there are.
If you are going to consider the intangible value of the fellowship, then you also need to consider the intangible value of the counterfactual. This will either be that of your current work, or another job you might take (which is hard to estimate, but you could guess at it).
Similarly to the above, if you’re going to consider the information value of applying, then you also ought to consider the information value of applying for whatever else you might have spent the time on (if that’s what you would have done). Since many fellowships receive thousands of applications, I presume that most applications get not feedback (as in my own experience). The information value is therefore likely to be higher for opportunities with higher chances of success. Interviews give a greater information value still, so this effect is quite strong. This might even be the main reason why I prefer not to apply for positions where my chance seems so slim.
A related advantage of focusing on higher-probability positions is that you might be lucky enough to receive multiple offers around the same time. That lets you compare options after learning more about them during the process, and choose the one that seems like the best personal fit. If most of your applications are very low-probability, you’re more likely either to accept the first offer you get, or to pass on a good offer in the hope of a better one later, simply because the relevant information arrives at very different times.
Lastly, there is a psychological cost to firing out many low-chance applications. Rejections are, after all, demotivating. Getting interviews or offers is highly motivating, even if you end up declining the offers.
This leads to a much much more complicated equation! I asked GPT, to try to link this all together, and well… I’ll just link its answer here.
So I remain pretty unsure as to whether it’s worth it for myself personally, but I might put in an application anyway :)
Thanks a lot for engaging!
One general point: My rough guess is that acceptance rates have stayed largely constant across AI safety programs over the last ~2 years because capacity has scaled with interest. For example, Pivotal grew from 15 spots in 2024 to 38 in 2025. While the ‘tail’ likely became more exceptional, my sense is that the bar for the marginal admitted fellow has stayed roughly the same.
The model does assume that most applicants aren’t spending 100% of their time/energy on applications. However, even if they were, I feel like a lot of this is captured by how much they value their time. I think that the counterfactual of how they spend their time during the fellowship period (which is >100x more hours than the application process) is the much more important variable to get right.
This is correct. I assumed most people would take this into account (e.g. subtract their current job’s networking value from the fellowship’s value), but I might add a note to make this explicit.
I’m less worried about this one. Since we set the fixed Value of Information quite conservatively already, and most people aren’t constantly working on applications, I suspect this is usually small enough to be noise in the final calculation.
I agree this is real, but I think it’s covered in the Value of Your Time. If you earn £50/hr but find applying on the weekend fun/interesting, you might set the Value of Your Time at £5/hr. If you are unemployed but find applying extremely aversive, you might price your time at e.g., £200/hr.
I also have the impression that these fellowships are getting more competitive each year. Do you share that perspective? If so, that would require a further adjustment to the calculation.