October 19, 2016

AI for Developers: Enabling the Next Economy with Machine Intelligence

Last week, I had the opportunity to speak at the O'Reilly Next:Economy Summit in San Francisco. I was on a panel of startups working to empower and augment the human workforce alongside the new technologies that will define our future. 

One of those new technologies, almost certainly the most important, is artificial intelligence. With the continued march of AI, it has become apparent that intelligence is going to become a core part of every software and hardware application. In practice, this means it is now viable to automate non-routine work that has previously resisted automation, enabling us to scale our expertise. 

I spoke about the disconnect between those who build teachable systems and those who have something to teach.  On the one hand, we have data science and machine learning experts who excel at making sense of vast troves of data and uncovering the intelligence contained therein. On the other hand, we have subject matter and domain experts who already have intelligence we want to bring to bear. In-between, we have developers and engineers who codify and build the systems that power our businesses. 

The key for the next economy is to unlock AI for our developers and engineers. As it stands today, there are more than 21 million developers in the world but only 19,000 data science experts capable of building AI at the lowest level - that's three orders of magnitude more developers than data scientists worldwide. We must empower the developers that are already in our organizations to scale our expertise and automate our work. 

So how do we do that? By recognizing a simple fact - today's machine learning technologies are all about learning, which is a fancy way of saying they're about making better students. What we need instead is to look at the other side of that coin - to look at teaching and crafting the technologies and tools that allow us to codify how to teach learning systems the intelligence we actually want to apply. At Bonsai, we've built a platform that enables exactly this. We shift the level of abstraction up so your teams can focus on building, teaching, and using intelligence models without getting mired in the low-level mechanics of machine learning. 

Using Bonsai, developers focus on the concepts they want to teach the AI model and how to teach them. The low level details are left to the platform, similar to how databases let us focus on using our data for the business problem we want to solve instead of managing the database itself. By working at a higher level, a problem that formerly required handwritten advanced deep learning network topologies can now be solved with just a couple dozen lines of code (you can see the difference in the slides below).

Ultimately this gives you a host of benefits: increase the scope of what you were doing and achieve it in less time; understand what was going on so that you can debug and iterate more effectively; reuse and share work more effectively; take advantage of improvements in the levels below without having to rebuild all of our work. This enables us to harness the power of AI in our businesses, bringing our unique and collective intelligence to bear as we build the next economy. 

I've posted slides from my talk below. You can also learn more about Bonsai and apply for the private beta at bons.ai!

Always. Be. Learning.

Stay up-to-date on our latest product news, AI industry highlights, and more!