Monica Lam discusses how to turn cutting-edge computer science research into commercial products. She covers the gap between lab breakthroughs like custom processor cores and what the market actually needs, and the strategies required to make research viable at scale.
Chapters
The kind of spin out that I do are not the ones that I necessarily recommend to everybody here because we started these projects as computer science research projects and they're usually very disruptive. And so the first company tries to, what we did is that we let people have customers custom processor cores in ASIC chips. So it's a great idea, but it takes a while because it is so disruptive. And by the way, this is Tensor because they just sold the company to Cadence and it's successful, but it takes a while. And the second company was a company called Mocha 5 where we look at where we use use virtual machine technology for computer management and it's a huge problem again and it is very disruptive. And it also takes time. And if I look at what happened in those two times is like in research we try to solve the problem in the optimal way but the minute you leave the university you find that nobody is ready to absorb what you are trying to reach. So you have this go to market issue and that is that the first thing you have to do is to tick where you're going and reduce it down to these bite sized pieces, something that they can possibly absorb and then you build your way back up. So the work that we did kind of in the university is kind of like research for the next 7, 15 years depending on how you look at it and then you kind of work your way back up. And we were lucky because I think the problems we solve are real serious ones. They're not the fickle ones, they don't change. And even though we take such a heavy handed approach to them, it stood the test of time. We have the time to keep working and then it will come in after a while because we were looking, because we were focusing on the trends. So for example, in this one we're focusing on the trends that there will be a lot of distributed computing and this will be the right architecture as a result of that. I don't know when this is going to come in, but it is true that in the last three experiences it's getting faster and faster.
