Image via Fcb981/Wikimedia

Do you know how to invest globally? Guild Investment Management has decades of proven experience investing in emerging growth opportunities all over the world. Subscribe to the Guild Investment’s Market Commentary today to access exclusive actionable insights with no borders.


At the beginning of the year, we wrote about Alphabet’s (GOOG) acquisition of a British artificial intelligence (AI) startup called DeepMind. DeepMind made headlines last year when its AlphaGo program defeated the champion go player Lee Sedol of South Korea, then the second-ranked go player in the world. It has since defeated the world’s top player, although only in informal online play.

AlphaGo showcases DeepMind’s transformative approach to AI: machine learning that emulates human learning. Traditional approaches to computer chess, for example, rely on brute force — working out all the permutations of possible future moves to identify the most promising next move to make. That approach is limited because the number of permutations gets so large as the program looks more and more moves ahead. It’s even more limited with the game of go, because there are so many more possible moves than there are in chess. The “decision tree” quickly gets far too large to exhaustively examine and evaluate. That’s why until last year, most experts thought it would be decades before a computer program could compete with world-class go masters. They were waiting for new hardware that could crunch the numbers faster. But DeepMind delivered with revolutionary software instead.

A New Model For Deep Learning

How did AlphaGo do it? By emulating human learning. DeepMind’s “machine learning” emulates the process of human learning — learning by experience, rather than trying to have everything “baked in” from the beginning. It also emulates the neurological structure of human learning, constructing artificial “neural nets” that are formed and persist in the same way as analogous structures in the human brain.

That neurological modelling means that one key weakness of AI is being overcome — the risk of what has been called “catastrophic forgetfulness.” In less dramatic language, this just means that when an AI machine begins a new task, it starts from scratch instead of leveraging previously acquired knowledge of different but related tasks, as humans do. By making “neural nets” more persistent, DeepMind is letting its programs that have gotten good at one video game, for example, carry some of that knowledge over when they begin trying to play a similar game. So far, the programs that are “single minded” are still better, but the programs applying neural nets acquire broader expertise much faster.

Now DeepMind is beginning to apply the same technology to other problems besides gaming. The company has entered into talks with the UK’s National Grid (NGG), the company responsible for electricity transmission in the UK and a descendant of the Thatcher-era breakup of the country’s state electricity company. NGG’s job has gotten more challenging as intermittent and unpredictable renewables have risen in the mix of generation technologies. DeepMind’s CEO commented last week that the application of AI could cut the country’s electricity usage by 10% with no new infrastructure — just by optimization and demand balancing.

Managing Public and Regulatory Sentiment

Another — and more controversial — partnership started last year, between DeepMind and the UK’s National Health Service (NHS). DeepMind and the NHS initially portrayed their collaboration as a limited and focused effort to deliver actionable clinical data on kidney disease to doctors using a new smartphone app. However, it turned out that the data provided to DeepMind was much more extensive than initially disclosed. Watchdog groups complained that medical data privacy was being compromised, and DeepMind’s subsequent smaller collaborations have been more constrained and transparent.

Medical applications will be one of the most significant areas for AI deployment, and the potential benefits are very great. Doctors will use AI analysis to make better, earlier diagnoses and to choose more appropriate care for patients, with great potential savings in public health expenditures. Huge data sets will have to be leveraged to get these benefits. There will be public discussion along the way — some of it potentially disruptive, particularly when the private firms involved loom as large in public and regulatory consciousness as GOOG.

However, the power of AI to augment human decision-making power and lead to better and more efficient outcomes will inexorably drive the adoption of AI technologies, even if the process is accompanied by occasional outcry and controversy. Competition will be fierce. Differentiation among the competitors will come not just from technology, and not just from the acquisition of top talent. It will come from the ability to negotiate access to large public data sets — and the ability to maintain public and regulatory goodwill in the process.

Investment implications: Alphabet, Inc’s DeepMind subsidiary showcased its revolutionary artificial intelligence technologies last year, when its AlphaGo program confounded the experts and defeated the world’s best players of go. Now it is making deals to leverage massive public data sets — in energy and health care — and use its machine learning to deliver improved outcomes and potentially big cost savings to the public. Some critics say DeepMind’s use of these big data sets has not been transparent enough, or satisfied data privacy laws. We think that the potential benefits of AI are so great that its adoption will go forward inexorably, sometimes with public and regulatory concern and pressure. The competition to develop AI and deploy it will be intense. The companies that succeed will be not just those with the best technology and the top talent; they will also be the ones who can maintain the goodwill of the public and of regulators while gaining access to the huge public data sets that will allow their technology to demonstrate its true potential. Although many upstart tech companies have a reputation for arrogance, it may be the companies who can best adapt to being in the public eye who will ultimately be the most successful.


Want to invest with the experts? Subscribe to Equities Premium newsletters today! Visit http://www.equitiespremium.com/ to learn more about Guild Investment’s Market Commentary and Adam Sarhan’s Find Leading Stocks today.