Deepmind recipient of the Royal Society’s Mullard Award and

Deepmind Technologies is a British Artificial Intelligence Company formally based in
London, England. It was co-founded by Dr. Demis Hassabis, Mustafa Suleyman, and Dr. Shane
Legg in 2010. In 2014, the company was acquired by Google and is now a part of their parent
company Alphabet Inc. Deepmind now has an applied team in Mountain View, California and
research centres in two Canadian cities: Edmonton and Montreal. CEO, Demis Hassabis, is
former chess prodigy and began his technology career making award winning video games. His
later research into neural mechanisms was listed as one of the top scientific breakthroughs of
2007 by Science. He is also a,” 5-times World Games Champion, a Fellow of the Royal Society
of Arts, and the recipient of the Royal Society’s Mullard Award and the Royal Academy of
Engineering’s Silver Medal” (Deepmind Technologies). Head of Applied AI, Mustafa Suleyman,
is accountable for the company’s commitment to solve large-scale world issues. At the young
age of 19, Mr. Suleyman, established a telephone counselling service which became one the
largest mental health services in England. Later, he started Reos Partners, an international
consultancy that aimed to solve complicated problems. Recently, he launched DeepMind Health,
a technological partnership with the U.K National Health Services helps these health institutions
run more efficiently, and ultimately saves lives. Chief Scientist, Shane Legg, spent his higher
education in Switzerland becoming a master on models of intelligent machines and human
decision making. After receiving his PHD, he spent two years at the Gatsby Computational
Neuroscience Unit at UCL studying how the brain relates to algorithms.
Tasong 6
Purpose
Artificial Intelligence according to Merriam-Webster Dictionary is,” the capability of a
machine to imitate intelligent human behavior”. However, the reality is the programs that
continue to be regularly updated prove that machinery can and will outperform humans
capability. Within the last ten years, the artificial intelligence community has significantly
expanded and project funding has skyrocketed. Therefore, the discoveries and developments
within this time are occurring rapidly and without precedent. Deepmind Technologies main
objective is to develop artificial intelligence programs that can self-solve complicated issues. It
believes that the use of AI can speed the development process and create efficient solutions. The
company’s website states,
If we’re successful, we believe this will be one of the most important and widely
beneficial scientific advances ever made, increasing our capacity to understand the
mysteries of the universe and to tackle some of our most pressing real-world challenges.
From climate change to the need for radically improved healthcare, too many problems
suffer from painfully slow progress, their complexity overwhelming our ability to find
solutions. With AI as a multiplier for human ingenuity, those solutions will come into
reach. (Deepmind Technologies)
As a “multiplier for human ingenuity” the success of much of Deepmind’s research can be
attributed to their nuanced examination of the human brain itself. The application of the brain’s
various mechanisms of learning are utilized to further machine information processing. Their
premier focus on game playing was the best possible activity to perfect AI complexity on. It
represents the capability to solve critical and complex dilemmas in the near future.
Tasong 7
Structure
Deepmind Technologies is split into two general sectors: Research and Deepmind
Applied. Research is organized through combining,” long-term thinking and interdisciplinary
collaboration of academia along with the relentless energy and focus of the very best technology
start-ups” ( Deepmind Technologies). This sector tackles a large amount of scientific problems,
many of which are prone to shift rapidly. However, their main focus is always centered on
progressing computer ability to mimic the human brain. According to Senior Research Scientist
Raia Hadsell believes the sectors,” combination of research into games, neuroscience, deep
learning and reinforcement learning as a unique proposition that could lead to fundamental
breakthroughs in AI” (Deepmind Technologies).
Deepmind Applied is focused on the application of their successful research into the real
world. It focuses on collaboration with other businesses and experts to support the betterment of
humanity. This sector is split into three subsections: Deepmind Health, Deepmind for Google,
and Deepmind Ethics & Society. Deepmind Health was formed in order to provide advanced
technology that could support doctors, nurses, and patients. The research and mobile tools are
forming a more efficient test and treatment system. The Deepmind for Google focuses on energy
efficiency for Google Data centres and the further development of Google services. The Ethics &
Society section researches the implications of various AI developments. The goal is to only
create technology that has a positive social impact. This sector has fellows who serve as
independent advisors that provide guidance and feedback. Their current fellows are: Professor
Nick Bostrom (Oxford University), Professor Diane Coyle ( U of Manchester), Professor Edward
Felten (Princeton University), Professor Jeffrey Sachs (Columbia University), Christiana
Tasong 8
Figueres, and James Manyika. Their partners are: The AI Institute at NYU, Article 36, The
Institute for Policy Research at University of Bath, The Institute for Public Policy Research,
Leverhulme Centre for the Centre for Intelligence, Oxford Internet Institute’s Digital Ethics Lab,
Partnership on AI, The Royal Society, and The Royal Society for Arts, Manufactures &
Commerce.
Research
Deepmind’s research has had internationally renowned success, and most of it has
occurred at public events for the world to see for themselves. Go, is an ancient Chinese board
game that has been played for an estimated 3,000 years (Knapton). It is an extremely complex
game that, according to Deepmind has,” … an astonishing 10 to the power of 170 possible board
configurations – more than the number of atoms in the known universe – making Go a googol
times more complex than Chess” (Deepmind Technologies). This classic game has been a
standard challenge for AI programs and for decades they could only play on an amateur level.
The creation of the AlphaGo program by this company has forever changed this long studied
game. This program was the first to ever win against a Go professional and then against Lee
Sedol, known to be the greatest player of the decade. The match took place in Seoul, South
Korea in March of 2016 with over 200 million viewers. This was a major breakthrough for the
science community and the world itself. Since then, Deepmind has continued to further develop
their Go programs and now has the AlphaGo Zero program which surpasses all the previous
ones. It’s high efficiency is because it learned from playing itself instead of data from other
games and players. This concept is referred to as tabula rasa, the program had no preconceived
notions and essentially became its own teacher. According to CEO and Co-Founder Demis
Tasong 9
Hassabis,” the program was so powerful because it was no longer constrained by the limits of
human knowledge” (Knapton).
The Deepmind Research sector has also had great success with Deep Q-networks, termed
DQN, and continues to develop differential neural computers (DNC). DQN is a deep
reinforcement learning system that is able to master Atari 2600 games on an extraordinary level.
Atari 2600 is a classic game console that was the precursor to popularized consoles like
Nintendo and Playstation. The DQN when able to defeat all human players when tested with 49
diverse games. It was able to perfect its strategies without any human intervention. This is
considered progress towards potential General AI. General artificial intelligence is the ability for
a machine to perform any cognitive exercise a human can. The developing DNC system uses a
neural network and memory to solve complex data. The network uses memory to produce
answers and optimisation enables it to self-improve. Essentially it is,” a learning machine that,
without prior programming, can organise information into connected facts and use those facts to
solve problems” (Deepmind Technologies). Currently, the system is quite slow and but has a
solid efficiency. However, in the future it could prove groundbreaking for computer science,
neuroscience, and cognitive science.
Deepmind Health’s research is focused on a technique called deep learning. Their
algorithms analyze visuals and attempts to interpret their meaning. Currently, it is being
developed with: head and neck scans at University College London Hospitals NHS Foundation
Trust, eye scans at Moorfields Eye Hospital NHS Foundation Trust, and mammograms with the
Cancer Research UK Imperial Centre. It is learning to identify potential health risks within the
scan photos, then eventually recommend treatments to health professionals. In collaboration with
Tasong 10
the UK National Health Services, an app called Streams has been created to address failure to
rescue. This is a concept that address how efficiently healthcare providers can respond and
prevent complications. Currently, Streams is geared towards helping patients with acute kidney
injury (AKI). Treatment for this condition costs the NHS an estimated one million euros. In the
UK every year, approximately 40,000 AKI related deaths occur, 25% of which the NHS claims
are preventable. The app holds patient information such as test results, vital signs, and pre
existing conditions. It can alert clinicians remotely in the case of urgent assistance as well. Right
now, the app is only in use at the Royal Free London NHS foundation. Nurse there have reported
the app was,” saving them up to two hours each day” (Deepmind Technologies). The app is still
in the development stages and Deepmind is working to add more features that benefit a larger
range of patients and clinicians.
Deepmind for Google has made the Google company more sustainable and environment
friendly. Industrial sites, like Google data centre, are sources of large energy consumption.
However, the one of the company’s key focuses is sustainability and Deepmind has made major
breakthroughs in the race for green energy. They applied machine learning and,”trained the
neural networks on the average future PUE (Power Usage Effectiveness), which is defined as the
ratio of the total building energy usage to the IT energy usage” (Deepmind Technologies). The
AI company has reduced Google data centres cooling bill by 40%. This is through the use of
water to cool their centres as opposed to air conditioning systems. The utilize sea water, natural
evaporative cooling, and recycle the water. Another large success for Deepmind was their
WaveNet network that was implemented by Google Assistant. It is a neural network that
produces audio waveforms that produces realistic speech patterns in comparison to alternative
Tasong 11
networks. Alternative network use live actor recordings, referred to as concatenative TTS, or
computer generated voices that mimic grammar rules, known as parametric TTS. In contrast,
WaveNet is a,” deep generative model that can create individual waveforms from scratch, one
sample at a time, with 16,000 samples per second and seamless transitions between individual
sounds” ( Deepmind Technologies). This system is referred to as a convolutional neural network
because it is based on actual brain processes.
The Deepmind Ethics & Society sector bases their research on six themes: privacy,
transparency, and fairness, economic impact, governance and accountability, managing AI risk,
AI morality and values, AI and the world’s complex challenges. This is the newest unit of the
company and it focuses on interdisciplinary research methods. These researchers must abide to
five core principles: social benefit, rigorous and evidence-based, transparent and open, diverse
and interdisciplinary, collaborative and inclusive. All of their research is based on the postulation
that all forms of artificial intelligence should be beneficial to all of society and remain under
human control.
Publicity
Deepmind has had large media coverage due to their numerous research projects and
groundbreaking successes. The company is a firm advocate for open source information. This
means a program’s source code is open to the public. They have released,”… open source
environments, datasets, and codes to support and accelerate research in the wider AI community”
(Deepmind Technologies). The company’s website has hundreds of published research papers
that are available for download. Many of which were accomplished in collaboration with
academia, a priority of the company. In addition to these feats, they have also been the topic of
Tasong 12
controversy. The Deepmind Health sector collaborates closely with the National Health
Services. During the AI company’s Stream trials, the Royal Free NHS Foundation neglected to
comply with the Data Protection Act. Therefore, the estimated 1.6 million people’s personal
information given to the company lacked patient consent. This was confirmed by the Information
Commissioner’s Office conducted a year long investigation