Great to hear that you agree with most of the post (two out of its three main points), let me focus on the disagreements regarding the impact potential of choosing to go to med school:
As far as I can see, you raise three general points of disagreement. (do let me know if you don’t feel like I represented your points well)
Disagreements based on personal experiences with studying medicine in Australia
Disagreements regarding the value/transferability of knowledge taught in med school for high impact work
Disagreements regarding the relevance of EA meta risks (funding dependencies, cause X, etc.) and whether a medical degree actually makes you resilient against these
Regarding 1) Disagreements based on personal experiences with studying medicine in Australia:
I think some of the negative aspects you describe about med school are very important but might not be general to all medical degrees. I’ll give some personal examples just to show that I disagree that your descriptions are generally accurate, of course you might actually be right in the majority of cases.. I do not have a good intuition if the median medical school internationally is more similar to the one that you describe or to the counterexamples I offer!
You write of highest workloads, few electives, all for a medical degree with no doctorate. (and that networking opportunities suffer from all that) I see how all of this is generally sub-optimal for high impact career paths, such as the research career in AI safety that you’re currently pursuing.
Credentials
This is not the right comparison. A medical degree doesn’t help you to do many of the things that a doctorate does.
In Germany, as an option additional to the medical degree (‘Staatsexamen’) , most medical schools offer a Dr. med. (MD) degree track and some offer an advanced MD/PhD track (e.g. here). Both of these are considered a doctorate and are a very good foundation on which to build a career in research/academia or to go into industry. (given sufficiently good grades on either, with the MD/PhD generally being more valuable)
Electives and workload
15% of my modules are wholly elective, two further semesters offer longitudinal electives. I am sure that a lot of degrees offer way more, but I consider general workload being reasonable to be even more important.
Because, in my experience the workload is actually not that high in some medical schools, such as mine. A lot of my friends/acquaintances studying fundamental/applied sciences degree at elite universities such as Oxford or ETH Zurich have much higher workload. This means that the majority of my time is ‘elective’ in that I currently have enough of it to use it for 1) being main organizer and founder at my uni, 2) a research assistant job at a lab and for 3) things such as writing this post (and I am currently actually in the most time-intensive year). I know that this would not possible to this extent in all other courses that I could be doing.
2) value/transferability of knowledge taught in med school
It is hard to say how much of the disagreement here arises from uni to uni differences in course structure and how much is more about theories of value of knowledge.
First, regarding uni to uni differences: Of the 40 modules during my 5 theoretical years, only 9 focus on specific pathologies (and related pharmacology, etc.). A lot of the rest does contain knowledge on pathologies, pharmacology, etc. but always embedded in a general, translational framework which furthers understanding of the human. It might be that this is more focus on transferable knowledge than in the median course.
Re theories of value of knowledge: Since you focus mostly on transferable knowledge for research: From the experience that I gathered so far, academic research requires deep specialization in very specific methods researching very thin slices of reality, all of which with a depth that goes far beyond any course curriculum. This means that learning and special knowledge beyond uni material will always be necessary for academic research. While fundamental science degrees may arguably make this easier (especially the more fundamental the research is), knowledge from a medical degree will rarely make this further work impossible! It would be interesting to know how much harder you’ll have it with a medical degree and thus how big this cost is but right now I am not convinced it is big enough to make me update! (The fact that you did manage to specialize in AI safety at FHI, Oxford with a medical degree is a nice anecdote in this regard that I couldn’t help noticing)
Furthermore, over here there is the option of specializing into research and starting your doctorate as early as 2-3 years into the course, which could arguably make your specialized knowledge & fit for a specific research direction by the end of med school much better than that of someone with a general fundamental science degree. This is again of course something that will vary from uni to uni.
Lastly, when it comes to potential high impact career paths outside of research such as policy, entrepreneurship, (industry?) I suspect that we have a similar general picture:
Some more specific degrees could have arguably been an even better fit in terms of knowledge, but some aspects could yet again be in favor of the medical degree: e.g. translational understanding and firsthand experience of health systems on all levels, trained social skills, credentials of being a doctor (which is might be more of a thing in fields outside academia), job/income security allowing for time for other things etc.
And of course, the last point about income/job security is dependent on:
3), Disagreements regarding the relevance of EA meta risks (funding dependencies, cause X, etc.) and whether a medical degree actually makes you resilient against these:
Funding
You might be right that there’s nothing to worry about in being dependent on EA funding due to historic successes in getting funds and the funding diversity you speak of. However, my own prior is generally quite low when it comes to placing trust in getting funds without problems, even when being ‘smart enough to be admitted to a medical degree’. I’ve got some more to say about this topic but I to some more research/thinking on this specific aspect and then write another post discussing it in a more general context!
Cause X / Resilience
I think you might have changed my mind on this point while I was writing this reply. Here are my thoughts: Regarding the question of how well a medical degree prepares you for cause X:
I agree that we can not assume that a medical school graduate is magically better suited for cause x, however I personally believe that EA right now is too focused on narrow quantifiable causes (due to inherent uncertainty of anything that’s broader) and focused too little on general resilience-building. (Here are my rough thoughts regarding this)
I was going to say here that while potential future high impact resilience building efforts don’t necessarily have to relate to medical fields, a broad degree such as medicine will be more likely to be valuable for them than a narrower one such as a specific fundamental sciences degree. However I think you raised a good point in that some fundamental science degrees such as e.g. physics or biology are actually broader than medicine in important ways and might thus arguably be more valuable/applicable to future high impact resilience building efforts. (at least in my current abstract understanding of the term)
This is something that I would like to think more about and that I will reserve my judgement on for now!
I hope I could give some more perspective as to why we disagree on some points, overall you did cause me to update regarding my views on a medical degree resilience, potentially causing me to update on the value of spending time to apply for unis abroad after all! Let’s see! ;)
PS:
I’m not sure it’s a matter of personal opinion. Greg has convincingly argued the opposite.
A very small fraction of MDs are admitted to joint MD-PhDs. A medical degree is the only one that can come with optional extras—in many other degrees a similar fraction of students would be publishing papers with supervisors. And the PhD that a medic does will not necessarily be as relevant as those of a computer scientist. Basically, it seems like a way of avoiding an apples-with-apples comparison.
15%-elective is terribly little.
Note that the workload may skyrocket in the clinical years
Regarding (2) transferability: I believe you’re overthinking it. From a zoomed out view, medical classes are approximately useless, and this talk of a specialised class becoming useful by being “embedded in a translational framework” is basically waffle.
You understate the case for the usefulness of useful subjects. If I’d studied computer science for undergrad, I could’ve got where I am now 5+ years earlier. Even dropping out of medical school could have accelerated things. In such a scenario, I could’ve been a somewhat more credible applicant for things like top professor positions than is currently the case. (Of course, skill is the main thing, but getting promptly educated, and building a stellar CV at a young age does help, vs studying irrelevant subjects).
Regarding (3) funding
my own prior is generally quite low when it comes to placing trust in getting funds without problems, even when being ‘smart enough to be admitted to a medical degree’.
What kind of evidence would make you update your prior? Many funders say they are willing to fund any person doing excellent longtermist work, and many orgs are continually growing and hiring. To take one extreme example, $50k fellowships are being given out to interested teenagers. It’s a movement that’s >10 years old, with its funding-base growing double-digits per year. If you’re smart enough to get into a German medical degree, and dedicated, then it should be possible to do excellent work...
A very small fraction of MDs are admitted to joint MD-PhDs. [...] in many other degrees a similar fraction of students would be publishing papers with supervisors. And the PhD that a medic does will not necessarily be as relevant as those of a computer scientist. <
It being a small fraction doesn’t make it less viable for an EA approach to studying med school. Every EA approach to uni will incorporate some tight admission rate..
It might not be relevant for AI safety but it will be super relevant for e. g. neartermist EAs or EAs that don’t rank AI risk as high and want to focus on biorisk.
I believe you’re overthinking it. From a zoomed out view, medical classes are approximately useless, and this talk of a specialised class becoming useful by being “embedded in a translational framework” is basically waffle.<
We do not have ‘medical classes’. We have classes on systems of the body: foundational classes (biochemistry, molecular biology, physics, physiology)
and classes that incorporate practical info, where you would argue they’re approximately useless such as pharmacology. I disagree that they are entirely useless as it teaches you on a daily basis how the fancy science translates to practice, a skill that I will continue to argue is highly important (and at the core of any problem solving inside and beyond academia) and a skill that a pure fundamental science degree is ‘approximately useless’ for.
funding
Fair points, as I said, I reserve my judgements here for now..
Hi Ryan, thank you for your comment!
Great to hear that you agree with most of the post (two out of its three main points), let me focus on the disagreements regarding the impact potential of choosing to go to med school:
As far as I can see, you raise three general points of disagreement. (do let me know if you don’t feel like I represented your points well)
Disagreements based on personal experiences with studying medicine in Australia
Disagreements regarding the value/transferability of knowledge taught in med school for high impact work
Disagreements regarding the relevance of EA meta risks (funding dependencies, cause X, etc.) and whether a medical degree actually makes you resilient against these
Regarding 1) Disagreements based on personal experiences with studying medicine in Australia:
I think some of the negative aspects you describe about med school are very important but might not be general to all medical degrees. I’ll give some personal examples just to show that I disagree that your descriptions are generally accurate, of course you might actually be right in the majority of cases.. I do not have a good intuition if the median medical school internationally is more similar to the one that you describe or to the counterexamples I offer!
You write of highest workloads, few electives, all for a medical degree with no doctorate. (and that networking opportunities suffer from all that) I see how all of this is generally sub-optimal for high impact career paths, such as the research career in AI safety that you’re currently pursuing.
Credentials
In Germany, as an option additional to the medical degree (‘Staatsexamen’) , most medical schools offer a Dr. med. (MD) degree track and some offer an advanced MD/PhD track (e.g. here). Both of these are considered a doctorate and are a very good foundation on which to build a career in research/academia or to go into industry. (given sufficiently good grades on either, with the MD/PhD generally being more valuable)
Electives and workload
15% of my modules are wholly elective, two further semesters offer longitudinal electives. I am sure that a lot of degrees offer way more, but I consider general workload being reasonable to be even more important.
Because, in my experience the workload is actually not that high in some medical schools, such as mine. A lot of my friends/acquaintances studying fundamental/applied sciences degree at elite universities such as Oxford or ETH Zurich have much higher workload. This means that the majority of my time is ‘elective’ in that I currently have enough of it to use it for 1) being main organizer and founder at my uni, 2) a research assistant job at a lab and for 3) things such as writing this post (and I am currently actually in the most time-intensive year). I know that this would not possible to this extent in all other courses that I could be doing.
2) value/transferability of knowledge taught in med school
It is hard to say how much of the disagreement here arises from uni to uni differences in course structure and how much is more about theories of value of knowledge.
First, regarding uni to uni differences: Of the 40 modules during my 5 theoretical years, only 9 focus on specific pathologies (and related pharmacology, etc.). A lot of the rest does contain knowledge on pathologies, pharmacology, etc. but always embedded in a general, translational framework which furthers understanding of the human. It might be that this is more focus on transferable knowledge than in the median course.
Re theories of value of knowledge: Since you focus mostly on transferable knowledge for research: From the experience that I gathered so far, academic research requires deep specialization in very specific methods researching very thin slices of reality, all of which with a depth that goes far beyond any course curriculum. This means that learning and special knowledge beyond uni material will always be necessary for academic research. While fundamental science degrees may arguably make this easier (especially the more fundamental the research is), knowledge from a medical degree will rarely make this further work impossible! It would be interesting to know how much harder you’ll have it with a medical degree and thus how big this cost is but right now I am not convinced it is big enough to make me update! (The fact that you did manage to specialize in AI safety at FHI, Oxford with a medical degree is a nice anecdote in this regard that I couldn’t help noticing)
Furthermore, over here there is the option of specializing into research and starting your doctorate as early as 2-3 years into the course, which could arguably make your specialized knowledge & fit for a specific research direction by the end of med school much better than that of someone with a general fundamental science degree. This is again of course something that will vary from uni to uni.
Lastly, when it comes to potential high impact career paths outside of research such as policy, entrepreneurship, (industry?) I suspect that we have a similar general picture:
Some more specific degrees could have arguably been an even better fit in terms of knowledge, but some aspects could yet again be in favor of the medical degree: e.g. translational understanding and firsthand experience of health systems on all levels, trained social skills, credentials of being a doctor (which is might be more of a thing in fields outside academia), job/income security allowing for time for other things etc.
And of course, the last point about income/job security is dependent on:
3), Disagreements regarding the relevance of EA meta risks (funding dependencies, cause X, etc.) and whether a medical degree actually makes you resilient against these:
Funding
You might be right that there’s nothing to worry about in being dependent on EA funding due to historic successes in getting funds and the funding diversity you speak of. However, my own prior is generally quite low when it comes to placing trust in getting funds without problems, even when being ‘smart enough to be admitted to a medical degree’. I’ve got some more to say about this topic but I to some more research/thinking on this specific aspect and then write another post discussing it in a more general context!
Cause X / Resilience
I think you might have changed my mind on this point while I was writing this reply. Here are my thoughts:
Regarding the question of how well a medical degree prepares you for cause X:
I agree that we can not assume that a medical school graduate is magically better suited for cause x, however I personally believe that EA right now is too focused on narrow quantifiable causes (due to inherent uncertainty of anything that’s broader) and focused too little on general resilience-building. (Here are my rough thoughts regarding this)
I was going to say here that while potential future high impact resilience building efforts don’t necessarily have to relate to medical fields, a broad degree such as medicine will be more likely to be valuable for them than a narrower one such as a specific fundamental sciences degree.
However I think you raised a good point in that some fundamental science degrees such as e.g. physics or biology are actually broader than medicine in important ways and might thus arguably be more valuable/applicable to future high impact resilience building efforts. (at least in my current abstract understanding of the term)
This is something that I would like to think more about and that I will reserve my judgement on for now!
I hope I could give some more perspective as to why we disagree on some points, overall you did cause me to update regarding my views on a medical degree resilience, potentially causing me to update on the value of spending time to apply for unis abroad after all! Let’s see! ;)
PS:
I agree, I updated the wording on this
Just to rebut a few points there.
On (1) credentials/electives/workload:
A very small fraction of MDs are admitted to joint MD-PhDs. A medical degree is the only one that can come with optional extras—in many other degrees a similar fraction of students would be publishing papers with supervisors. And the PhD that a medic does will not necessarily be as relevant as those of a computer scientist. Basically, it seems like a way of avoiding an apples-with-apples comparison.
15%-elective is terribly little.
Note that the workload may skyrocket in the clinical years
Regarding (2) transferability: I believe you’re overthinking it. From a zoomed out view, medical classes are approximately useless, and this talk of a specialised class becoming useful by being “embedded in a translational framework” is basically waffle.
You understate the case for the usefulness of useful subjects. If I’d studied computer science for undergrad, I could’ve got where I am now 5+ years earlier. Even dropping out of medical school could have accelerated things. In such a scenario, I could’ve been a somewhat more credible applicant for things like top professor positions than is currently the case. (Of course, skill is the main thing, but getting promptly educated, and building a stellar CV at a young age does help, vs studying irrelevant subjects).
Regarding (3) funding
What kind of evidence would make you update your prior? Many funders say they are willing to fund any person doing excellent longtermist work, and many orgs are continually growing and hiring. To take one extreme example, $50k fellowships are being given out to interested teenagers. It’s a movement that’s >10 years old, with its funding-base growing double-digits per year. If you’re smart enough to get into a German medical degree, and dedicated, then it should be possible to do excellent work...
It being a small fraction doesn’t make it less viable for an EA approach to studying med school. Every EA approach to uni will incorporate some tight admission rate.. It might not be relevant for AI safety but it will be super relevant for e. g. neartermist EAs or EAs that don’t rank AI risk as high and want to focus on biorisk.
We do not have ‘medical classes’. We have classes on systems of the body: foundational classes (biochemistry, molecular biology, physics, physiology) and classes that incorporate practical info, where you would argue they’re approximately useless such as pharmacology. I disagree that they are entirely useless as it teaches you on a daily basis how the fancy science translates to practice, a skill that I will continue to argue is highly important (and at the core of any problem solving inside and beyond academia) and a skill that a pure fundamental science degree is ‘approximately useless’ for.
Fair points, as I said, I reserve my judgements here for now..