DeepMind: the AI Revolution

Artificial intelligence is the ability of a computer to do tasks usually carried out by humans, by mimicking human thinking to learn how to improve at tasks independently. Though there have been recent breakthroughs in Artificial Intelligence, especially in the automotive and robotics industries, the field is still in its infancy. Although the software that has been developed is extraordinary at finding patterns in data, it lacks the ability to evaluate this. However, firms are focused on making the comprehension ability of these programs more human-like, with one pioneer in the industry being DeepMind.

DeepMind was founded in 2010 by neuroscientist Demis Hassabis, a former child chess prodigy, with the aim of advancing science and humanity through developing artificial intelligence. The firm is working towards bridging the gap between neuroscience and machine learning, to produce a “computer brain” that can think and generate new ideas independently. It was acquired by Google in 2014 for more than 500 million USD, signalling Google’s willingness to invest to remain at the cutting edge of artificial intelligence. 

DeepMind’s revenues are entirely derived from applying its technology to commercial Google projects and they have found success in improving Google’s voice recognition and the accuracy of estimated arrival times in Google Maps. They have also produced ground-breaking results in protein structure prediction, using the software APLHAFOLD2 to predict the shapes of every protein in the human body to a very high degree of accuracy, solving one of the hardest problems in biology. DeepMind has made the software and code for this project available to the public, allowing pharmacy companies to access and use the technology to help discover new targets for drugs, and cutting the time taken to discover new innovative treatments.  

DeepMind has also found success in using Artificial Intelligence to produce solutions to open ended coding problems and has produced AlphaCode that can perform to the level of an average participant in coding competitions. However, the program cannot replicate the tasks of human programmers yet, being only able to write short snippets of code reliably. There is also the fear that letting Artificial Intelligence write code could inadvertently lead to bugs and errors in the software.

Artificial Intelligence is a competitive industry, with other software giants making ventures into the industry, such as Microsoft who are working to develop their own technology by acquiring the firm OPENAI for one billion USD. This has resulted in the cost of talent acquisition rising exponentially, and some leading researchers in this field have been offered ten times their salary to work in the private industry. This has led to DeepMind’s wage expenses being very high, contributing to the firm making multi-billion-dollar losses until very recently, resulting in questions over the firm’s potential profitability for Google. 

However, DeepMind has not been without its controversies, and its co-founder Mustafa Suleyman was relocated from the firm to Google’s headquarters after complaints about his aggressive management style. In addition to this, Google has struggled to implement checks and balances to its Artificial Intelligence technology. Attempts to set up an independent ethical body have been met with failure after staff protests at the potential appointees, and two heads of artificial intelligence ethics have left the firm.

Furthermore, Artificial Intelligence relies heavily on acquiring a large amount of data, and the sources of data that DeepMind utilized have been controversial. In 2017, DeepMind received access to the medical records of 1.6 million NHS patients without explicit permission, resulting in fines for breaches of privacy. Although this data could be used to improve diagnostics such as the reading of x-rays, there are serious ethical concerns regarding its misuse.  

Google clearly can achieve great things with DeepMind, such as assisting Doctors in the diagnostics of patients and improving our ability to problem solve in the field of software engineering. However, there are increasing concerns over how software giants such as Google access and use the data of their customers, and Google’s need for DeepMind to become a money-making venture may lead to it using its technology to simply develop more profitable ways of manipulating personal data. The attempts to make sure DeepMind is behaving in an ethical fashion have not proved reassuring so far. Therefore, it will be interesting to see whether Google can become more transparent with its use of data in the near future, especially relating to DeepMind. 

By Alex de Souza

Sector Head: Dylan Buckley