One of the biggest phenomena of the past decade was “Software is eating the world”. Today, there is almost no industry which is not run on software and delivered as an online service. This has resulted in new online giants such as Uber, AirBnB, Amazon etc. But now that so many businesses are leveraging software, I believe we are about to witness the next disruptive wave in the form of Artificial Intelligence – in which software will not only be used to ‘run’ services, but will also be able to ‘understand’ what service it is running, making it that much more powerful. This will be one of the most exciting opportunities that will generate a new wave of transformative companies.
Artificial Intelligence (AI) is not a new thing. It has been researched for decades. But even the best AI systems were always highly tuned for specific problems and required many ‘rules’ to operate successfully. However, in the past few years, a specific AI technique called Deep Learning is taking the world by storm, and many researchers are abandoning the classical AI methods and switching to it. Deep Learning relies on simulating large, multilayered webs of virtual neurons, which enable a computer to learn to recognize abstract patterns (somewhat similar to the way a human brain functions). It can be used to solve any general-purpose pattern-recognition problem, which means that any activity that has access to large amounts of data can find it useful. When you combine this with the fact that much more data is available today due to the proliferation of online services that log every transaction, improvements in storage, advancements in GPU and computing power, development of cheap sensors, and the growth of the Internet of Things (IOT) – this becomes even more interesting.
All the big software companies are investing heavily in building Deep Learning Capabilities: Google recently acquired London-based AI-startup DeepMind, which recently beat a Go champion in a historic moment for AI evolution. The company is now deploying Deep Learning in almost all of its products (everything from search, speech recognition, and Google photos), Facebook built a dedicated AI research laboratory, headed by Yann LeCun, a star researcher from NYU, and last year Baidu even poached one of Google’s top AI gurus to head up a new Silicon Valley-based lab of its own.
However, what’s more interesting is that these companies are not only pushing AI and Deep Learning for internal use, but they are doing their best to advance the entire industry by releasing their software frameworks and libraries: Google has recently announced that it’s open sourcing its latest AI system, TensorFlow. Facebook is releasing, for free, the designs of a powerful new server intended to run AI software. IBM open-sourced their Machine Learning code, SystemML. Elon Musk and others founded OpenAI, a nonprofit AI research group. And there are many other examples. What this means is that startups can benefit from the vast internal research done by large software institutions.
As a result, we are seeing a new crop of startups that leverage Deep Learning to solve really interesting and complex problems which were unsolvable only a few years ago. This can be anything from performing medical diagnosis based on vast amount of clinical data, detecting fraudulent transactions in a large data set, predicting manufacturing failures based on sensory data, understanding human voice and recognizing human intentions based on context, identifying objects in a moving image, and much more.
I think we are going to see many interesting startups that build cutting-edge Deep Learning expertise. Many of which will get acquired by the large software giants that want to improve their AI capabilities. But over time, I believe the most interesting opportunities for new startups is to build a data network effect on top of Deep Learning capabilities. This is what it looks like:
What this means is building a large and unique data set that allows to better ‘train’ the Deep Learning algorithms which results in better insights. These insights should translate to a better product and hence more customers. If customers share their data with the company, this will improve the quality and size of the data set, resulting in a positive virtuous cycle. New competitors will have a classic chicken and egg problem: It will be impossible to match the Deep Leaning insights without the data, but they would not be able to get customers’ data without having better insights (and product). This results in a natural monopoly which tend to generate large companies.
Startups that not only build top notch AI expertise, but also leverage it to build a positive data network should become extremely valuable over time. This is what we are looking to invest in.