I wrote this post up and circulated it among my rationalist friends. I've copied it verbatim. I figure the more rationally inclined people that can critique my plan the better.

--

TL;DR:

* I'm going to commit to biomedical engineering for a very specific set of reasons related to career flexibility and intrinsic interest.
* I still want to have computer science and design arts skills, but biomedical engineering seems like a better university investment.
* I would like to have my cake and eat it too by doing biomedical engineering, while practicing computer science and design on the side.
* There are potential tradeoffs, weaknesses and assumptions in this decision that are relevant and possibly critical. This includes time management, ease of learning, development of problem solving solving abilities and working conditions.

I am posting this here because everyone is pretty clever and likes decisions. I am looking for feedback on my reasoning and the facts in my assumptions so that I can do what's best. This was me mostly thinking out loud, and given the timeframe I'm on I couldn't learn and apply any real formal method other than just thinking it through. So it's long, but I hope that everyone can benefit by me putting this here.

--
So currently I'm weighing going into biomedical engineering as my major over a major in computer science, or the [human-computer interaction/media studies/gaming/ industrial design grab bag] major, at Simon Fraser University. Other than the fact that engineering biology is so damn cool, the relevant decision factors include reasons like:

  1. medical science is booming with opportunities at all levels in the system, meaning that there might be a lot of financial opportunity in more exploratory economies like in SV;
  2. the interdisciplinary nature of biomedical engineering means that I have skills with greater transferability as well as insight into a wide range of technologies and processes instead of a narrow few;
  3. aside from molecular biology, biomedical engineering is the field that appears closest to cognitive enhancement and making cyborgs for a living;
  4. compared to most kinds of engineering, it is more easy to self-teach computer science and other forms of digital value-making (web design or graphical modelling) due to the availability of educational resources; the approaching-free cost of computing power; established communities based around development; and clear measurements of feedback. By contrast, biomedical engineering may require labs to be educated on biological principles, which are increasingly available but scarce for hobbyists; basic science textbooks are strongly variant in quality; and there isn't the equivalent of a Github for biology making non-school collaborative learning difficult.

The two implications here are that even if I am still interested in computer science, which I am, and although biomedical engineering is less upwind than programming and math, it makes more sense to blow a lot of money on a more specialized education to get domain knowledge while doing computer science on the side, than to spend money on an option whose potential cost is so low because of self study. This conjecture, and the assumptions therein, is critical to my strategy.

So the best option combination that I figure that I should take is this:

  1. To get the value from Biomedical Engineering, I will do the biomedical engineering curriculum formally at SFU for the rest of my time there as my main focus.
  2. To get the value from computer science, I will make like a hacker and educate myself with available textbooks and look for working gigs in my spare time.
  3. To get the value from the media and design major, I will talk to the faculty directly about what I can do to take their courses on human computer interaction and industrial design, and otherwise be mentored. As a result I could seize all the real interesting knowledge while ignoring the crap.

Tradeoffs exist, of course. These are a few that I can think of:

  • I don't expect to be making as much as an entry level biomedical engineer as I would as a programmer in Silicon Valley, if that was ever possible; nor do I believe that my income would grow at the same rate. As a counterpoint, my range of potential competencies will be greater than the typical programmer, due to an exposure to physical, chemical, and biological systems, their experimentation, and product development. I feel that this greater flexibility could help with companies or startups that are oriented towards health or technological forecasting, but this is just a guess. In any case that makes me feel more comfortable, having that broader knowledge, but one could argue that programming being so popular and upwind makes it the more stable choice anyway. Don't know.
  • It's difficult to make money as an undergraduate with any of the skills I would pick up in biomedical engineering for at least a few years. This is important to me because I want to have more-than-minimum wages jobs as a way of completing my education on a debit. While web and graphic designers can start forming their own employment almost immediately, and while programmers can walk into a business or a bank and hustle; doing so with physics, chemistry or biology seems a bit more difficult. This is somewhat countered by co-op and work placement, and the fact that it doesn't seem to take too much programming or web design theory and practice before being able to start selling your skills (i.e. on the order of months).
  • Biomedical Engineering has few aesthetic and artistic aspects, the two of which I value. This is what attracted me to the media and design program in the first place. Instead I get to work with technologies which I know will have measurable and practical use, improving the quality of life for the sick and dying. Expressing myself with art and more free-wheeling design is not super urgent, so I'm willing to make this trade. I still hope to be able to orient myself for developing beautiful and useful data visualizations in practical applications, like this guy, and to experiment with maker hacking.

There is still the issue of assuring more-than-dilettante expertise in computer science and design stuff (see Expert Beginner syndrome: http://www.daedtech.com/how-developers-stop-learning-rise-of-the-expert-beginner). I am semi-confident in my ability to network myself into mentorships with members of faculty [at SFU] that are not my own, and if I'm not good at it now I still believe that it's possible. In addition, my dad has recently become a software consultant and is willing to apprentice me, giving a direct education about software engineering (although not necessarily a good one, at least it's somewhat real).

There are potential weaknesses in my analysis and strategy.

  • The time investment in the biomedical engineering faculty as SFU is very high. The requirements are similar to those of being a grad student, complete with a 3.00 minimum GPA and research project. The faculty does everything in its power to allay the burden while still maintaining the standard. However, this crowding out of time reduces the amount of potential time spent learning computer science. This makes the probability of efficient self-teaching go down. (that GPA standard might lead to scholarship access which is good, but more of an externality in this case.)
  • While we're on the conscientiousness load: conscientiousness is considered to be an invariant personality trait, but I'm not buying it. The typical person may experience on average no change in their conscientiousness, but typical people don't commit to interventions that affect the workload they can take on either by strengthening willpower, increasing energy, changing thought patterns (see "The Motivation Hacker") or improving organization through external aids. Still, my baseline level of conscientiousness has historically been quite low. This raises the up front cost of learning novel material I'm not familiar with, unlike computing, of which I have a stronger familiarity due to lifelong exposure; this lets me cruise by in computing courses but not necessarily ace them. Nevertheless, that's a lower downside risk.
  • Although medical problems are interesting and I have a lot of intrinsic interest in the domain knowledge, there are components of research that interest me while others that I don't currently enjoy as much as evidenced from my current exposure. I can seem myself getting into the data processing and visualization, drafting ergonomic wearable tech, and circuit design especially wrt EEGs. Brute force labwork would be less engaging and takes more out of me, despite systems biology principles being tough but engaging. So there's the possibility that I would only enjoy a limited scope of biomedical engineering work, making the major not worth it or unpleasant.
  • Due to the less steep learning curve and more coherent structure of the computer science field, it seems easier to approach the "career satisfaction" or "work passion" threshold with CS than for BME. Feeling satisfied with your career depends on many factors, but Cal Newport argues that the largest factor is essentially mastery, which leads to involvement. Mastery seems more difficult to guage with the noisy and prolonged feedback of the engineering sciences, so the motivations with the greatest relative importance might be the satisfaction of turning out product, satisfying factual curiosity or curiosity about established/canon models (as opposed to curiosity which is more local to your own circumstances or you figuring things out), and in the case of biomed, saving lives by design. With mathematics and programming the problem space is such that you can do math and programming for their own sakes.
  • Most instances of biomedical engineering majors around the world are mainly graduate studies. The most often reported experience is that when you have someone getting a PhD in biomedical engineering, it's in addition to their undergraduate experience as a mechanical engineer, an electrical engineer or a computer scientist. The story goes that these problem solving skills are applied to the biology after being developed - once again a case of some fields being more upwind than others. By contrast, an undergradute in bioengineering would be taking courses where they are not developing these skills, as our current understanding of biology is not strongly predictive. After talking to one of the faculty heads, the person who designed the program, he is very much aware of problems such as these in engineers as they are currently educated. This includes overdoing specialization and under-emphasizing the entire product development process, or a principle of "first, do no harm". He has been working on the curriculum for thirty years as opposed to the seven years of cases like MIT - I consider this moderate evidence that I will not be missing out on the necessary mental toolkit over other engineers.
  • In the case where biomedical engineering is less flexible than I believed, I would essentially have a "jack of all trades" education meaning engineering firms in general would pass over me in favor of a more specialized candidate. This is partially hedged against by learning the computer science as an "out", but in the end it points to the possibility that the way I'm perceiving this major's value is incorrect.

So for this "have cake and eat it to" plan to work there are a larger string of case exceptions in the biomedical option than the computing options, and definitely the media and design option. The reward would be that the larger amount of domain specific knowledge in a field that has held my curiosity for several years now, while hitting on. I would also be playing to one of SFU's comparative advantages: the quality of the biomedical faculty here is high relative to other institutions if the exceptions hold, and potentially the relative quality of the computer science and design faculties as well. (This could be an argument for switching institutions if those two skillsets are a "better fit". However, my intuition is that the cost for such is very high and probably wouldn't be worth it.)

Possible points of investigation:

  • What is hooking me most strongly to biomedical engineering were the potentials of cognitive enhancement research and molecular design (like what they have going on at the bio-nano group at Autodesk: http://www.autodeskresearch.com/groups/nano). If these were the careers I was optimizing towards as an ends, it might make more sense to actual model what skills and people will actually be needed to develop these technologies and take advantage of them. After writing this I feel less strongly about these exact fields or careers. Industry research still seems like a good exercise.
  • I will have to be honest that after my experience doing lab work for chemistry at school, I was frustrated by how exhausted I am at the end of each session, physically and mentally. This doesn't necessarily reflect on how all lab work will be, especially if it's more intimately tied with something else I want to achieve. And granted, the labs are three hours long of standing. It does make me question how I would be like in this work environment, however, and that is worth collecting more information for.
  • To get actual evidence of flexibility in skillset it would be worth polling actual alumni from the program, to see if any of the convictions about the program are true.

--

Thoughts, anyone?

New Comment
14 comments, sorted by Click to highlight new comments since:
[-]lmm90

While we're on the conscientiousness load: conscientiousness is considered to be an invariant personality trait, but I'm not buying it. The typical person may experience on average no change in their conscientiousness, but typical people don't commit to interventions that affect the workload they can take on either by strengthening willpower, increasing energy, changing thought patterns (see "The Motivation Hacker") or improving organization through external aids.

I worry that you're deluding yourself. What evidence have you examined?

I knew several extremely smart people who planned to study something else at the same time as their university course. All of them gave up these plans in a matter of weeks. There was one term where due to an organizational cock-up I had to do 25% more courses than normal. I survived, but it wasn't pleasant. University is hard, probably the hardest thing I've ever done. (It was certainly much harder than any job I've had since).

Remember that older people are in charge of most hiring. It's politically unacceptable to say that certain subjects are outright better than others, but they are (or at least will be treated that way when hiring), and a "funny" course like this will make it harder to get jobs than a more traditional subject - even jobs for which this course is perfectly suited (explanation: even a more spread out degree won't give you the actual skills the job needs. But a more focused degree is proof that you can focus on a subject and learn it deeply)

Your career will force you into specialization even if your degree isn't specialized; after your first job or two people are much more interested in your experience than your degree. If you switch into another field after 10 years that means going back to being treated like a fresh graduate - very bad for your salary.

I thought you had something about motivation, but I can't find it now; my experience is that doing something you believe in is not sufficiently powerful motivation when you're working on something unpleasant. You need to find a job you can be satisfied first, and one that's doing the right thing second; you won't be productive if you're unhappy.

So my advice is:

  1. Do the most traditional, focused degree you can - in your case Computer Science (and do it at the most reputable university you can)
  2. The prior odds of you successfully studying something else while at university are vanishingly low. Plan accordingly.
    1. If you still want to follow this plan, I'd suggest you start the learning computer science part now. You're not going to be much smarter next year, and you are going to be much busier. This should also give you some information on how much work you can do in practice.

I'm probably underweighing more conservative assessments like this, so I appreciate it.

motivation and self-delusion

I have not collected evidence the directly contradicts statistical assessments regarding the conscientiousness trait. Instead I'm making an inference based off a collection of evidence that I can name. I don't think I've given much consideration to evidence strength yet so working through this will be a good exercise.

For example:

Historically my conscientiousness has been quite low in part due to depression. I've been coming out of that depression recently, and have improved in my ability to keep on task even when I'm discouraged. Oftentimes psyching myself out was the reason why I haven't instigated behavioral change, because when I fall off the bandwagon I don't get back on. This change towards optimism makes me feel comparatively more competent and willing to explore my alternatives for support and skills.

Though, as a counterpoint: I am not experiencing mania, but the fact that I've recently acquired and optimistic temperament that has not been subject to calibration by the new action-space means that I might still be overestimating my abilities instead of underestimating them.

But given that I am strongly interested in doing things that successful people do that I couldn't before:

  • Nick Winter's assessments in his book "The Motivation Hacker" make me believe that there exists low hanging fruit when it comes to motivation that I have not yet picked. I would guess the same for typically surveyed people due to the recency of prescriptive motivation literature like "The Procrastination Equation".

  • Successful students and learners follow regular patterns of behavior that can be turned into habits. The particular examples would be the writings of Cal Newport, Scott H Young, in addition to consulting my academic advisors and the successful students themselves. Needless to say I probably haven't been using those patterns, which include precommtiments, oicking a good study environment and using it regularly, processing textbooks in a way that produces reviewable notes, and using office hours.

Twin and developmental studies might make me eat my dust on this if I'm directly challenging claims about a personality trait. I'm feeling a bit of resistance to looking them up but I should probably push through it and get it over with.

There are other conditions by which the amount of work and stress that someone can take on goes up, like joining the military; yes, I'm considering it. But there are also less extreme options like just having good health and being more organized, taking up a martial art or doing a sport. Not all of these are going to take off and most certainly I won't be doing all of them at once. So one obstacle I need to consider is the timeframe towards orienting myself properly for success in biomedical and whether the value is greater or lower than lost wages or other measures of opportunity cost.

I have also experimented with nootropics, which I know believe are overrated but still a useful tool in the toolkit. Finally I am beginning to use Anki, which might be a good way of managing larger volumes of knowledge.

At this point I would like to get answers to my questions on actual working conditions, hiring practices, and future work opportunities. Grabbing all of the experiences with the largest decision-relevant information:cost ratio possible could help me resolve whether this plan will work out. This is unless all of the evidence from current models is substantial enough to outweigh the potential evidence from empiricism.

computer science and self-study; old people

There are at least two components here: the actual studying and skill acquisition, and the judgement made by the hiring practitioner.

I read on Less Wrong in this popular PSA that a handful of people have managed to get programming jobs through self-study. Although it seems reckless - would it be possible to define a satisficing case for the amount of practice that I would do towards the profile of skills of what a hiring person would want from their employee? This would help resolve the following:

  • whether or not the idea of studying is even feasible for the target skill level and time constraints
  • if you control for skill level, and add the condition of whether I have a compsci major or don't have a compsci major, what do the probabilities of being hired look like? If for a person with a major at the expected skill level I will have has a largely dominating probability, then yeah, I would want to reconsider.

I could talk to HR people or other software engineers at developer meetups, or at career fairs, to get a clearer picture on this. But if like you claim this is a political factor, then maybe I won't be getting the evidence I need.

You need to find a job you can be satisfied first, and one that's doing the right thing second; you won't be productive if you're unhappy.

I'll keep this in mind. It does seem safer.

[-]lmm20

Getting the actual programming skills is easy if you're smart. Getting the evidence that will lead people to hire you is harder. Large companies tend to go by the book; you will need the qualifications or something unusual like a personal recommendation from someone in the company. Startuppy places it's more about fitting in with the culture and talking/coding well in interview. If that's the kind of job you're after you'll probably be fine as a self-taught programmer if you can perform under interview pressure and you conform to the right stereotype. (it's possible I'm being excessively cynical here)

The interest thing on your list is that neither Nick Winter, Cal Newport or Scott Young have jobs at some company. If you take those people as your role model, are you sure you want to focus on the goal of getting a job?

I might be a bit biased but I think it easier to do a startup when you can do computer programming.

That's a good point. How mutually exclusive is the optimization path for being highly employable versus self-employing or bootstrapping? Is it just a question of efficiency of time spent or is there more to it?

How much computer science knowledge is necessary for startups, do you think? I can program and have worked on software modules and have written my own utilities, but I still have a lot to learn conceptually and I still need to survey a wider range of technologies, especially related to databases and web development in the front and back end. That's even excluding some of the trendier hotspots like semantic web, NLP and machine learning.

That's a good point. How mutually exclusive is the optimization path for being highly employable versus self-employing or bootstrapping? Is it just a question of efficiency of time spent or is there more to it?

There are companies that you can't start via bootstrapping. I think a lot of expensive medical equipment design is in that class. I would also think that bio/nano tech is in that class.

I can program and have worked on software modules and have written my own utilities, but I still have a lot to learn conceptually and I still need to survey a wider range of technologies, especially related to databases and web development in the front and back end.

I have taken a semester worth of course on data bases and they didn't tell me anything useful about them. It was mostly impractical theory. The most disturbing thing was that the TA didn't know that a prepare statement in Java prevents you from SQL injections.

When it comes to databases the things you have to know are:

1) Try to never query the database directly in a way that allows for SQL injections.

2) Create indexes possible. It can make sense to experiment around with indexes to get optimal speed.

3) There something like transactions. In some settings a database automatically updates when you send it data, in other settings you have to commit or end the transaction.

Take a look at Nick Winters startup Skritter. He's doing a spaced repetition learning software for learning Japanese and Chinese Kanji. In contrast to Anki his software allows you to draw the Kanji. As far as cognitive enchancement goes I think learning Kanji is in the ballpark.

How much computer science knowledge does that need? Not that much. You need to know how to use a webframework like Django. You need to know javascript, probably something like JQuery, html, css. Some framework for iPhone/Android apps.

That's a bunch but you can learn as you go along. It also isn't deep computer science like machine learning and NLP.

In Nick Winter case it's interesting that he's a Asian studies minor. That's where he learned that the world needs a better way to learn Kanjis. That's where he felt the pain needed to focus on the idea. I feel similar to the biochemistry that I learned while studying bioinformatics.

If you want to produce medicial technology and are already able to program I don't think Biological Engineering is necessarily a bad choice. But I would recommend you to put the knowledge directly into practice.

An Arduino lilypad is cheap. Design the hardware with it and program it. Think about the kind of data you can measure and what to do with it.

This is important to me because I want to have more-than-minimum wages jobs as a way of completing my education on a debit

If you join any research labs during your career (and you should) you will probably end up programming quite a bit out of necessity in BME regardless of whether you set out to do so (although I suppose being a proficient programmer isn't the same as knowing computer science) and I think the fact that you can program at all should open up 12$-15$ / hour pre-graduation jobs. My experience among my friends is that Engineering / Math / Physics majors generally end up learning to program at some point (to the extent that people just generally expect them to be able to write basic stuff) so you've got a head start on that front.

I don't expect to be making as much as an entry level biomedical engineer as I would as a programmer in Silicon Valley, if that was ever possible; nor do I believe that my income would grow at the same rate.

Is this specific to silicon valley? Nation wide, things seem roughly similarly optimistic in terms of prospects for those majors...

Forbes

Payscale

What is hooking me most strongly to biomedical engineering were the potentials of cognitive enhancement research and molecular design (like what they have going on at the bio-nano group at Autodesk: http://www.autodeskresearch.com/groups/nano). If these were the careers I was optimizing towards as an ends, it might make more sense to actual model what skills and people will actually be needed to develop these technologies and take advantage of them. After writing this I feel less strongly about these exact fields or careers. Industry research still seems like a good exercise.

I'm guessing that computer science majors can often pursue these biomedical-ish sorts of careers, but the reverse is not true (Biomedical Engineers typically don't pursue computer science-ish careers).

Also, don't forget that some colleges allow you to do independent majors, so depending on the level of flexibility in the institution you can potentially hybridize the two.

I'm guessing that computer science majors can often pursue these biomedical-ish sorts of careers, but the reverse is not true (Biomedical Engineers typically don't pursue computer science-ish careers).

I am strongly interested in figuring out if this is true. Do you have any thoughts on how I would do this?

[-]oooo00

To do this your best bet is to talk to large numbers of biomedical engineering alumni. As a data point, you mentioned before that SFU has one of the most respectable biomed engineering programs. As another data point, University of Toronto doesn't allow general stream undergraduate engineers to choose certain specialties requiring that extra bit of intellectual horsepower unless you are able to enter (and survive) the more theoretical Engineering Science program. Biomed Engineering is one of the specialties that falls in this category.

I feel the reason that most biomed engineers don't pursue CS-ish careers is because many of them feel that their additional knowledge, training and suffering should be used for more "important" pursuits (grad school, designing life-saving medical devices, etc.). Combined with the general engineering school attitude that their education is more rigorous or harder than probably any other major in university (other than perhaps actuarial), and you have a situation where most engineers freshly graduated (barring Computer Engineers) would view pursuing a CS-ish career as a major step back.

However, given your stated interest in other goals (e.g. cognitive science, human cybernetics/enhancements/augmentation), this may not be a bad path to take provided you are mindful of and can navigate the immediate post-graduation job interviews.

As others suggested in this thread, it seems that you're probably much more geared towards a startup culture, in which case if you've chosen your electives correctly in 3rd and 4th year you would hopefully have had the chance to focus in on data visualization and/or bioinformatics and show an impressive body of work.

If you are motivated enough you may also try to take CS & math courses in the summer, or work on design projects to build up a body of work. Ideally summers would also be taken up with internships also, but at least the studying intensity would be somewhat reduced to allow you to get ahead on other credits/courses/knowledge/portfolio.

You may want to apply to 80,000 hours for a career coaching session. I think that it may have significant value for you since you currently have more of an ability to do a hard pivot than you will in the future.

I have first-degree friends who have worked with 80K and they've said it's unlikely that they would prioritize interviewing me, due to me not directly optimizing for earning-to-give (something which I made clear). I think it's still worth a shot to try and be put in their candidate pool, and I could see if I could get an off-the-record conversation with some of the staff. So we'll see.

Hi, I'd like to clarify that we prioritise people who are optimising around positive impact, not earning to give. If someone takes earning to give seriously, then we view that as a good indicator, but we speak to lots of people who aren't considering earning to give careers.

I started writing a response, but decided it would be better to summarise my general thoughts on degree choice and post them on our blog. So see our latest thoughts on how to pick a degree.

Insofar as this particular situation goes, I haven't thought about it much, so take this with a pinch of salt. My gut reaction is that CompSci is slightly more impressive than bio engineering, and if it helps you learn to program better, then the skills will be more generally useful. You also say that bio engineering is a major time sink, which I'd see as a count against it. So, my highly uncertain impression is that I'd prefer CompSci. On the other hand, if you'll find it easier and more motivating to study bio engineering and you'll get better grades, then I'd rate that pretty highly (especially if aiming to continue into research).

I'm personally studying bioinformatics. While I think the bit of labwork I did had some value I think you don't lose that much by instead learning through books & Anki.

As far of my skill in thinking about cognitive enhancement go, taking biochemistry couses had an interesting effect. The content of the knowledge it pretty usefless of cognitive engineering. On the other hand I have gokt a much better understanding about cognition by trying to get that kind of knowledge into my brain.

By contrast, biomedical engineering may require labs to be educated on biological principles, which are increasingly available but scarce for hobbyists; basic science textbooks are strongly variant in quality; and there isn't the equivalent of a Github for biology making non-school collaborative learning difficult.

That's not really true. If I wanted to build a better device for measuring lung function measuring I gained little of the relevant knowledge through lab work. The ideas I have about improving it would rather come from using the device for a long time for QS. The skills that I would need are skills about building hardware. That's probably something I could learn through internet research and become a resident at a hackerspace.

In the case where biomedical engineering is less flexible than I believed, I would essentially have a "jack of all trades" education meaning engineering firms in general would pass over me in favor of a more specialized candidate.

Firms often don't hire for specific skills. Provided you are smart and have skills a firm can hire you and teach you the domain specific skills that they need.

Note that I still have a timeline of 2-5 months before this plan can fully propogate to my actions. So that's the amount of time I have to research decision-relevant information and be able to pull through towards making my choice.