Friday, May 27, 2016

Interview with Hitoshi Harada, CTO/Co-founder of Alpaca

- What is your name
Hitoshi Harada

- What is the name of your company?
Alpaca

- Who are you?
I'm a co-founder and CTO of Alpaca. I bring more than ten years of experience in database core engine technology and data-driven applications.

- What does your company do?
Alpaca's mission is to democratize the institution-level digital asset management in capital markets such as quantitative financial analysis and trading automation using Artificial Intelligence.  We currently provide a product called Capitalico, which helps traders build automated trading algorithms without any lines of programming code by simply highlighting the similar patterns in the historical chart.  It has been very difficult for computers to interpret the complicated technical charts at the level of human beings, but thanks to the evolution of deep learning, we are now able to capture human interpretation of charts.  Typically, traders have to sit in front of the computer just to wait for the winning chart pattern to come, our CEO who has over 10 years of trading experience identified this as a heavily burdensome inefficiency. We resolved to solve this problem and now with Capitalico traders can get the same results but with unprecedented efficiency.

- What caused you to start your company?
I have been interested in new technologies that change the world, especially things that have happened in Silicon Valley where I am now.  My career has been around database technology and I have observed how much of the so-called Big Data hasn't been utilized enough due to the lack of capability to understand complex data.  When I saw the recent AI technology evolution, it was obvious that this emerging technology is going to change the world.  Two of my longtime friends had already been working around the technology, and it was clear to me that I had to start something.

- What makes your company an AI startup?
Our company started around image recognition and used to provide AI technology to solve image problems to big companies in the game and internet area.  There is a lot of excitement around recognition performance improvements made by AI lately, but each use cases was different, so we built a service called Labellio which we later sold to Kyocera, one of the biggest Japanese ICT companies. We then pivoted to Fintech and started this Capitalico service with similar technology and similar concept, but much more focused in the financial trading sector. Our company's philosophy is "be a human being", meaning we want to use AI to free people from those nonsense tasks.  Computers have made a lot of things automated but we don't think it's enough, and believe people should spend more time on more "human" tasks such as creative work.  As a consequence, it was natural for us to build an AI startup.

- Describe your company's AI.
Capitalico is a tool for financial traders to build automated trading algorithms by understanding the chart patterns.  Users just need to find similar patterns from historical charts, highlight and click “add.”  It builds a different AI model for each user since there are infinite ways to read technical charts and there is no single answer.  It is a very important concept for us and is inherited from the previous image recognition service to enable our AI to build a dedicated model for each particular problem.  We don't believe a ‘God AI’ exists as some people may.  Instead, our approach is that we should build AI to understand very particular problems that help the productivity of human beings.

- How do you measure and communicate the quality of your AI?
Within our space, there are typically two aspects with regards to the quality of AI.  One is of course the accuracy of recognition and the other is trading performance itself.  While it is important for our AI to accurately understand the way our users interpret charts, at the end of the day, our users care more about how much money our technology could make them.  So, it is important that we always keep a focus on make it as transparent as possible in both recognition accuracy and trading performance. We accomplish this by showing the prediction score overlaying the chart and calculating the backtest results including those metrics such as Sharpe ratio and drawdown.

- Can you share something awesome that your AI has done or that you have been able to do with your AI that would surprise most people?
Chart pattern recognition has been one of the challenging problems in this space for more than two decades.  Some of the previous work has developed a hand-crafted way to find the known patterns of one time sequence value of closing price, but the real traders are looking at more complicated charts with many indicators on top of the raw prices, and the patterns are a combination of known simple patterns plus the market momentum.  Our AI has proven results to find similar patterns considering all of those complicated configurations.

- When you talk to investors, do they care about your AI or do they care that you are solving a business problem? How about customers? How about the media?
Investors are often less concerned about the actual technology, instead, they focus on how much we can make. However, it is also true that investors are knowledgeable in the value of solving important business problems.  In order to build a stable business around a startup, we do need the technology strength so that others cannot easily catch up.  That's how investors look at us.  Customers are similar in that they care whether we can solve their problems or not.  Customers don't even care if our business is going to be stable or not.  Sometimes the cool technology sells itself, but we know that's not the core.  But among everyone, it is the media that is most enthusiastic about our technology.
- Why is there so much excitement about AI startups?
Because AI will change every business for sure.  And also because it is one of the hardest problems in history.  It's not something everyone can do, but if you can do it, it will be huge.  It's not just about technology but it's about what kind of problems you solve using what kind of AI technology.  So more and more people jump into the space trying to be the one, and more and more people watch what's going on in this space, and that's why there is so much excitement.

- Where do you go to get information about AI startups?
Today's AI space is so hot that you can find general trends by reading news sites.  For more technology related knowledges, I follow a couple of active twitter and facebook accounts such as Deep Learning Hub and AI technology community.  The thing is, the community is growing so rapidly and new technology comes up literally everyday, so social media is a great tool to stay on top of the evolving environment.

- Make a prediction about the future of AI startups.
I have heard many people saying that AI will change every business, and I agree with that, but it's not clear yet how it will change.  Some people are trying to build a big AI system that understands everything, but there are more people like us who think that this isn’t the goal.  At Alpaca, we believe that AI should enhance the professional's skill to solve problems in particular domains, and the way AI solves each problem can and should be different.  In the human world, there is no single answer to life’s challenges.  AI will follow the same rule.

1 comment:

  1. Interesting topic for a blog. I have been searching the Internet for fun and came upon your website. Fabulous post. Thanks a ton for sharing your knowledge! It is great to see that some people still put in an effort into managing their websites. I'll be sure to check back again real soon.
    Silicon Valley

    ReplyDelete