Overall investment thesis to bet on:
- research (R&D) - heavy businesses
- teams with strong domain expertise
- going after unique use cases where AI plays a critical role in solving them
But really, theres potential for all flavors of startups:
- Algorithm developers: new algorithms are typically very short-lived. If you serach ArXiv, hundreds of papers describing new techniques are published daily. That being said, there are various examples of successess, including DeepMind and Whetlab. For example, DeepMind never had an actual product but was a leader in deep learning / reinforcement learning algorithms, and was eventually acquired by Google for over $500 million. H2O is another example of an (open source) suite of algorithms, which has raised over $33 million and gained trememdous user traction.
- Algorithm / application solvers with deep domain expertise: a team may have several AI specialists (e.g., PhDs in AI / Computer Science), as well as members with deep domain expertise who have experience / are tech-savvy enough to understand how to use AI. One of the best examples is in self-driving cars - Cruise is an example of a company recently acquired by GM and currently testing fleets of their vehicles in San Francisco. In the AI for healthcare landscape, Cardinal Analytx Solutions, started by two physicians including an MD/PhD in CS (Nigam Shah).
- Algorithm / application solvers with proprietary data: this is unique and difficult to replicate. Success depends on multiple iterations of acquiring data, building a product that leverages such data, getting users to adopt the product and thus generate more data. Having data can also enable a company that traditionally isn’t a machine learning/AI company, to start doing AI and potentially become very disruptive. An example is Propeller Health, which started as a typical medical device company, but is able to work on AI initiatives now that they have a large user base. An example that the average consumer can relate to is Google - those CAPTCHA’s asking you to tag images of cats or street signs, are allowing Google to improve image tagging and search capabilities.