I had the privilege of giving a keynote this week at the annual Internet-of-Things Developers Conference in Santa Clara. Below are the slides from my presentation.
First off, it was remarkable to see the energy and excitement leading developers have around IoT, ML, and AI. As my colleague Ethan Kurzweil recently noted in his Venture Beat article regarding Microsoft’s gargantuan acquisition of Github, we’re living in an increasingly developer-centric world. This means that developers have more say about the projects that get built, the direction new technology takes, and what enterprises spend money on. For any investor interested in understanding the cutting-edge trends in the AI world or in enterprise software generally, developers are the people to hang out with!
Speaking with attendees, I noticed a lot of excitement around hardware, specifically building custom silicon for AI and ML applications. Since the presentation applies mainly to software businesses, I thought I’d quickly address the topic. With the announcement of Google’s newest TPU (tensor processing unit) and the potential for Amazon and Facebook to do something similar soon, it is clear that large companies are eager to own the “full-stack” in the cloud infrastructure ecosystem. They also have the R&D budget to make it happen and ownership of massive datasets to train models. On top of this, we see the legacy chipmakers like Intel and Nvidia eagerly looking to capitalize on the demand for AI chips. I see the space as extremely competitive and capital intensive and as such, very challenging for a startup to win.
Nonetheless, there are some interesting startups designing low-cost AI chips for specific use cases such as autonomous vehicles or edge computing, or specific types of algorithms like reinforcement learning. Furthermore, it’s quite possible that the next major breakthrough in quantum or neuromorphic computing comes from a new entrant rather than an incumbent. It is an exciting time, and innovations in hardware are requisite to “solving” Moravec’s paradox and seeing the next industrial revolution come to life. Who will ultimately bring to market these innovations remains an open question.
Finally, reflecting on my presentation, I’m eager for feedback – as my penultimate slide notes, laws are meant to be broken and oftentimes the best investments don’t follow preconceived laws about how technology or markets function. My investment philosophy around machine intelligence continues to evolve so I’d love to hear any pushback, questions, or new ideas people may have. Most importantly, if you’re working on a machine intelligence startup, software or hardware, that you believe will be the next unicorn, please be in touch!