Over the past three days, I attended a workshop on GPU programming and applications at IIT Bombay by NVIDIA. Here are a few comments by different speakers, that stayed with me. And prompted me to write about a topic I have got tired of writing about.
1) “If the government had not invested in super-computing in the 90s we would not have mobile phones today.”
2) “The prime minister recently announced an outlay of Rs 45,000 crores for developing supercomputers. This is what all of us are excited about”
3) “Gone are the days when hardware was designed to solve important algorithms. Today, we have to adapt our algorithms to suit available hardware.”
The last comment refers to the fact that GPUs were developed to satisfy the enormous demands of the gaming industry. They are now being used for solving scientific problems, and algorithms have to be adapted or designed to utilize the enormous capabilities that GPUs can provide.
It seems plausible that some of the complex problems being researched would remain unsolved if the government did not fund them. But, is it true?
The history of GPUs indicates otherwise. GPUs were funded by the desire of people to entertain themselves. By consumers who have no knowledge whatsoever of the problems researchers are trying to solve, no knowledge of the technical hurdles they face, and perhaps no ability to even comprehend these problems.
10 years back when I was doing a project that required some graphics programming, I was struck by the enormous amount of computing and programming effort involved in producing realistic rendering of scenes. And I wondered about the guys who toiled to build games that had not just realistic rendering but so much more (modelling collisions, deformations, haptic feedback, simulation of human behavior …). I wondered where they got the motivation to toil away just to satisfy some silly gamer who wanted to see blood spurting out of a bullet wound in a realistic manner in some 1st person shooter game. The computational and modelling was challenging and interesting. But to what end? I couldn’t have done that.
The point is that it doesn’t matter. The technology that is useful to model fake blood spurting out of a fake bullet wound is also useful to simulate the impact of waves generated by a storm surge or a tsunami. And the amazing thing about the free market is that the idiot gamer who spends much of his life shooting imagined enemies on the biggest and most expensive devices money can buy, is contributing to the research efforts of all the speakers who presented their work in the last three days. And the idiot gamer is spending his money voluntarily to enrich (or waste?!) his own life. It is not necessary to tax him or the luxury products he buys to fund research projects.
Proponents of government funding often make the argument that it is necessary to invest in activities that are not economically viable in the short term to get returns in the long term. And yet, ask any of the eminent speakers if they would have invested government money in GPU technology 20 years ago. The answer is probably obvious. The fact is that none of us is good at predicting what the future will hold, what direction technology will take, what will work, and what will not. And none of us has the right to take other people’s money for what we think are more worthy projects. I might choose to think of hard core gamers as idiots, but that judgement does not give me the right to take their money and spend it on earthquake modelling. Those of us who are passionate about scientific projects have as little conception of the joy that gamers derive from their games, as the gamer might have of the joy that we derive in formulating and solving differential equations. It is not for us to decide what activities other people should value. And we should stop making the case for robbing other people of their money to meet our goals – even if we do think our goals to be bigger and worthier that those of the people we propose to rob.
And if the moral argument is not sufficient, the history of GPUs should be a humbling reminder. The money we propose to spend on our pet projects could have been spent by others in other ways – on their pet projects. Are we confident that our pet projects will do more for the future of technology (whatever that means), than their pet projects? And if we are not, we should have the good sense to let the decision remain in the hands of the people who earned that money in the first place.