Year 1: The Review

4 minute read


So I’ve just finished a year in Baltimore and have returned to Perth. Here are some thoughts gathered over that time.

Baltimore is a very nice place to live, once you figure it out. When I was considering the move, every person who had visited thought it a fine place; every person who hadn’t was terrified at the suggestion. In reality, the situation is not nearly as bad as media would make it out to be, and there are plenty of interesting neighbourhoods and attractions within city limits. I would say that my quality of life in Baltimore improved significantly over what I could afford in Sydney spending roughly the same percentage of income on housing - bigger apartment, better commute, nicer neighbourhood (maybe not in safety, but then it is America).

There are two conclusions. Firstly, people are generally terrible judges of the quality of a place. Baltimore suffers a rather unfair portrayal in media (see the Wire, Trump’s various tweets) which people equate to the entirety of the Baltimore experience. By way of contrast, people are routinely surprised when I tell them that I would much rather live in Baltimore over a city like New York. Having spent a nonzero amount of time commuting there, I can say that any commute short of walking would be worse than anything I experienced in Sydney. Anecdotally, rent is much higher than that of Sydney. But people tend to have a romantic notion of NYC, partly due to a rosy portrayal in films and such. Secondly, some people are poor optimisers of utility when it comes to housing. I have learned to maximise the quality of accommodation in terms of location, duration and difficulty of commute, and access to important amenities. This is because these features will be accessed multiple times daily. In contrast, accessing luxury amenities like the beach, movies, shopping, and other extras shouldn’t be prioritised as much, because it is unlikely that such features will be accessed more than once a month. Indeed, as great as these luxuries may be, it is to difficult to estimate the rate of access, and therefore the estimate of the utility contributed by these luxuries will have a high variance. When evaluated correctly, the utility of living in cities such as Canberra, Perth, and Baltimore can easily surpass that of Sydney, New York, and Hong Kong.

The PhD has turned out to be a little more theoretical than expected. It turns out that it is very difficult to execute research that is closely aligned with practice. This is because sometimes solving real problems requires nothing new, just clever application of existing methods. Unfortunately this does not qualify as publishable work in many areas of machine learning. Therefore, work ends up being either theoretical or applied, rarely both. But theory is interesting to work in because it is less beholden to the vagaries of a particular dataset. And sometimes a dataset can only say so much about a particular problem, despite the project’s original goals. It’s been relatively challenging balancing research with coursework, but I’m hoping that’ll shape up during the rest of this year.

I’m sure this is obvious, but in a PhD you are there to develop novel methods to solve a problem. Incentives are aligned around doing new things, because it becomes publications. In industry, no one cares about clever and novel solutions, it suffices to have something that works. The thing that works is usually tried, tested, and good enough for the job. We might call this the fundamental problem of translational research, and in my experience I’ve found that it is extremely hard to bridge the two. For now, I maintain a “research” and an “industry” hat, and am careful not to mix the two.

So far, no regrets. The next few years will be interesting, as my ability to play the academic game (in a fair and honest manner, but I think everyone is playing the game to some extent even if they insist otherwise) will determine what happens next.