As highlighted in our post 2016 Forecasts, and 2015 Reviews, for Canada's Startups, over the course of the year a noticeable number of venture capitalists have turned their attention towards artificial intelligence technology. One form of artificial intelligence that is catching the eye (and wallet) of many venture capitalists is machine intelligence1, where artificial intelligence programs are linked with existing data collection systems to solve problems and conduct work more efficiently. Machine intelligence technologies address a vast array of problems (from classification to natural language processing) and methods (from support vector machines to deep belief networks).
The AI and MI Startup Ecosystem
Shivon Zills, an investor at Bloomberg Beta, predicts that, "in the next five to 10 years, it's going to be very difficult to see any data-oriented piece of the world not fundamentally transformed by this technology". Zills describes recent development in machine intelligence in the many forms of autonomous systems, such as self-driving cars, autopilot drones, robots that can perform dynamic tasks without every action being hard coded.
Currently, the majority of Bloomberg Beta's investments are in machine intelligence companies based in the U.S., though some are headquartered in Canada. One such example is Toronto-based Deep Genomics, which applies machine intelligence systems to early disease detection to create targeted therapies. Brendan Frey, biomedical engineering professor at the University of Toronto, launched Deep Genomics in order to bring a machine learning technology that detects disease-causing mutations in DNA to the public at large. Frey likens Deep Genomics' technology to a Google search for genetic mutations: researchers can query a DNA sequence, and the system will identify mutations and tell them what's going to happen and why it might cause a certain disease.
Many artificial intelligence and machine intelligence technologies will transform the business world. However, development of these technologies can start in regulatory grey areas. To overcome the difficulties of entering areas where regulations may not be updated in line with the advancement of technology, many startups are resorting to unique strategies. According to Zills, startups are seeking global arbitrage opportunities, such as health care companies going to market in emerging markets and drone companies experimenting in the least regulated countries. Big companies like Google, Apple, and IBM are seeking out these unique opportunities because they have the resources to be patient and are the most likely to be able to effect regulatory change.
Startups in the artificial intelligence and machine intelligence realm should consider working with legal professionals who are familiar with navigating the regulatory environment. Startups should also be mindful of corporate governance and intellectual property issues that can arise during the expansion of their organization.
1 As Jeff Hawkins and Donna Dubinsky of Numenta describe, what was meant by AI in 1960 is very different than what is meant today. In their view, there are three major approaches to building smart machines: Classic AI, Simple Neural Networks and Biological Neural Networks.
The Classic AI system is highly tuned for a specific problem. IBM's Watson could be viewed as a modern version of a Classic AI system. It focuses on creating a sophisticated knowledge base on a particular issue. Although Watson doesn't rely on encoded rules, it requires the close involvement of domain experts to provide data and evaluate its performance. Classic AI has solved some clearly defined problems, but is limited by its inability to learn on its own and by the need to create specific solutions to individual problems.
Simple Neural Networks are developed to allow AI to learn from data without human supervision. Recent evolution has created Deep Learning networks, whose advances have been enabled by access to fast computers and vast amounts of data for training. Deep Learning has successfully addressed many problems such as image classification, language translation and identifying spam in email. The big advantage of Simple Neural Networks over Classic AI is that they learn from data and don't require an expert to provide rules.
Biological Neural Networks involve truly intelligent machines that require a program foundation upon which everything else depends and have capacity to store a primary sequence of patterns and exhibit behavioral learning and continuous learning. Such machine intelligence systems can automatically build individual models of normal behavior and then predict what would be normal and flag anything abnormal. Use of this technology can link machine intelligence with existing data sets in order to solve complex problems for a range of different industries, including agriculture, healthcare, transportation and logistics.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.