Press "Enter" to skip to content

Category: AI

Power of GPU: 20 Years Ago and Now

When I got my first job at Sony, I was assigned to work on the graphics library development of PlayStation2 gaming system.

My job at that time consisted of writing code with assembly-like programming language called microcode, packing that code together with polygon and texture data, sending them to Vector Units (aka GPU) via DMA, and letting GPU do the rest of work while minimizing involvement of CPU.

I had to adopt every technical tweak in order to get the best performance and achieve the highest frame rate out of that gaming machine and GPU. That’s simply because the library was intended to be used by so many developers that make blockbuster game titles and they relied on it.

Time passes and now GPU is used for more generic purposes including computation of neural network. If I were a software engineer who just graduated from university today, I would jump straight into the world of neural-network-on-chip bandwagon without thinking much.

I didn’t realize that my knowledge and experience of GPU programming had to do anything with AI back then. I believe NVIDIA didn’t realize it neither.

To me, it’s quite interesting to see young software engineers learning how to write code against GPU without CUDA these days, especially ones trying to use Raspberry Pi as a deep learning accelerator. Special thanks to Broadcom for making its VideoCore specification public.

At the same time, it’s sad to see that Japan is a bit behind of this movement. The country used to host many GPU engineers. We had Sony, Sega and Nintendo. Every gaming system had such a sophisticated graphic library that pulled nearly 100% performance from its GPU.

I cannot stop wondering what if the engineers behind these gaming machines were given a chance to work on today’s neural network chip development. Maybe Japan could be in different position in AI industry by now.

a16z Podcast: Taking the Pulse on Bio

As a CEO of digital health startup, I could not miss this one.

Bioengineering can also benefit from unsupervised learning approach as discussed in another a16z podcast by letting AI show us the unique features that lead to a theory (or CRISPER) rather than having a human ‘guess’ it. There is no doubt that quantum computing will play an important role here in the near future.

I’m 100% certain that new type of entrepreneur who understands both bioengineering and computer science (perhaps, quantum computing) will attract so much attention from VCs as a bio industry needs this kind of talent.

Well, I gotta go back to college and study bioengineering again. :)

a16z Podcast: AI, from ‘Toy’ Problems to Production

I like this “end of theory” approach. Some businesses might not need a theory about what’s correct outcome will look like. In other words, AI deployment in the future will always start with unsupervised learning, let the AI tell us unique features, and then define a problem to solve from what the algorithm told us.

Plus, it’s no more about how to implement AI or what technique to use given that the technologies like TensorFlow are available for everyone today. As the podcast says, a couple of data scientists can start a real AI business within few days using these ready-to-use technologies.

It’s about how you pick which business problem to solve in terms of ROI. If you apply a cutting-edge technique to a problem that does not have a meaningful business impact at the company you are talking to, then the whole project is considered as ‘failure.’ You are simply solving a wrong problem.

So my conclusion is that as AI technologies are commoditized the startups whose strength is only technology will eventually go out of business. Instead, the startups that can tell a customer which problem to solve and contribute to the customer’s ROI will stay in business. Sometimes it results in denying what the customer wants to do with AI.

To do that, what you need is not a data scientist or software developer. You need a new type of business person who can take advantage of unsupervised learning and lead a customer in the right direction.

I also agree that a key differentiator will be domain expertise and not AI technology itself, meaning that a startup with the specific vertical focus such as medical imaging will stand out from others. I think this is true for both supervised and unsupervised learning.