I’m shocked, and I’m flattered.
I’ve noticed that certain posts I’ve made on this blog are routinely read, or at least looked at, over and over again. It’s surprising, and my popular posts are never the ones I expect to be a hit. Those posts are: English to Cantonese Translation, Analyzing Justin Trudeau’s Twitter Sentiment, The Math Behind Bitcoin, Match Your Resume with a Job on Glassdoor, and my most popular post, Fluid Dynamics = Financial Mathematics, also known as “How I am Convincing Myself that my Master’s Degree was Useful.” (Just kidding. I really value my Master’s and enjoyed it a lot).
A lot of people have read my fluid dynamics post, and even commented on it, which I really appreciate. It’s quite humbling to hear that my ideas have resonated with people. What’s not humbling though, is having someone plagiarize my work. Rather than calling it humbling, I’d call it violating. It’s not a nice feeling. And quite frankly, it makes for quite an emotional experience.
My (sort of) novel idea
The idea of comparing the Navier-Stokes to just about any other model out there is something that only someone with a background in fluid dynamics and/or physical oceanography would make. No one else cares enough about the Navier-Stokes to even think to compare it to something else. Not when magical machine learning and deep neural networks are out there to obsess over! And I doubt many people would compare the Navier-Stokes to the Black-Scholes, of all equations. I made that comparison in my post because first of all, I studied physical oceanography as part of my Master’s, and second, I have experience working in financial markets. It’s not impossible for someone else to make this comparison, and a quick Google search will show you that I’m not the only one to have done this. But it’s rare.
So when, after a quick Google search, I came across a spring 2018 class presentation by a PhD student at a top tier American university titled “From Navier-Stokes to Black-Scholes,” I was intrigued. As soon as I opened the slides, it felt a bit off. Reading through it, it was all too familiar. A quick search of his name led me to his LinkedIn profile. No mention of a finance background, or fluid dynamics study. Just mechanical engineering. It didn’t make sense to me that, given his background, he could draw an analogy between the two models that was so close to mine.
A quick side by side comparison of his slides and my post showed that he copied my work. Word for word.
Fortunately, I contacted the professor of the course, and he promptly took the presentation down. I also contacted the academic integrity office at the university, and they took the matter very seriously.
Whose IP Is It Anyway?
The PhD student copied ideas from my personal blog- not exactly an academic resource. It’s not clear what legally is my intellectual property. Is anything I post owned by me? Is it owned by WordPress? It’s a bit of a grey area.
But we’re moving into an age where things get even more grey. My experience reminded me of the famous case where three French art students used DCGAN to create a painting that sold for $432500. But the DCGAN implementation was taken from 17-year-old Robbie Barrat’s GitHub, and he didn’t receive even a penny for his work. But Ian Goodfellow came up with GANs. So who owns the rights to painting? Who should the money go to?
Sure, Goodfellow could’ve patented his work, and he could’ve cashed in big time. GAN-style algorithms pop up all over the place nowadays, as it’s moved from cutting edge to common knowledge in the machine learning community. Does that mean we have to pay him every time we use some form GANs? And if the inventor of the algorithm doesn’t own it, what about the implementer? If we fork some code on GitHub, how many changes do we have to make to it before we can call it our own?
The real problem with all these machine learning algorithms and implementations is that they have the possibility to make thousands, if not millions of dollars, as shown by the case of the painting. It’s important to give credit where credit is due.
In my case, I didn’t lose out on millions of dollars. But for creators of machine learning models, and engineers who implement them and open source their code, they could be.
As a good friend said to me, “Just paraphrase, cite, and move on, like the rest of the world. Lol.” But we need to figure out how that translates into our new age, where there’s real money on the line.