Original article was published on Deep Learning on Medium
DeepMind: Advances in Artificial Intelligence — Part I
The most significant advances in Artificial Intelligence in public are being made by the alliance that exists between DeepMind and Google.
A machine with artificial intelligence was trained to play the ancient game of Go, called Alpha Go, defeated world champion Lee Sedol.
This event generated the first glimpse into a world where an entity can imitate or possess human intuition.
Artificial intelligence is advancing in time.
The advances that we are seeing today, specialists in the field did not see that they could be seen until after 2030.
DeepMind’s greatest achievement with AlphaGo is that this entity was able to learn and evolve its knowledge in a similar way to that of a human being would.
Computational capacity, logic in data analysis processes and research, are significantly accelerating the development of artificial intelligence.
How is DeepMind different?
One of the previous methods to provide knowledge to a machine, was to give it all the possible combinations or scenarios that could be presented.
An example of this is when playing chess, an artificial intelligence with great computing power, could deduce what move to make based on the result of a huge database of possibilities.
On the other hand, AlphaGo, apart from being an entity with the ability to retain knowledge, acquires it through reinforcement learning and neural networks.
This being a requirement to win in a Go game, since the number of possible movements is greater than the number of atoms in the universe.
So it is impossible to calculate or learn all possible movements on the board.
The objective of this combination of factors is to imitate as much as possible the learning process of the human brain.
What is DeepMind?
DeepMind is an Artificial Intelligence company based in England, created in 2010.
They grew significantly in terms of research and development of technology related to general purpose artificial intelligence.
It is important to differentiate that a general purpose intelligence is completely different from a General Artificial Intelligence.
It was acquired by Google in 2014, one of the conditions of the acquisition by DeepMind was the creation of an artificial intelligence ethics group.
This group would be in charge of validating the evolution and decision-making that will be made regarding the artificial intelligence developed by DeepMind.
According to its website, DeepMind’s goal is as follows:
“Solve intelligence” and use that discovery to make the world a better place.
They emphasize the use of machine learning in order to generate algorithms that have the consequence of generating intelligence.
Artificial Intelligence in DeepMind and Human Intelligence
DeepMind seeks to understand and reach a conclusion of what is intelligence.
This objective does not have the sole purpose of being able to generate General Artificial Intelligence, but also to understand how the human brain works.
It is a company that values knowledge and the logical process carried out by the brain in order to reach logical conclusions.
Understanding the limitations that have on accessing this human brain, we seek to extend intelligence by using the extensive computational capacity that grows year by year.
Does a machine really learn?
As previously mentioned, DeepMind is different in that the technology is based on evolutionary knowledge and not just stored.
These entities “train” with information or events on a recurring and massive basis, from hundreds to thousands of different trainings.
Based on machine learning technology and neural networks, this entity is able to gain experience, retain information, recognize patterns, and come to a conclusion.
This is the closest thing a human does to what we know today.
That is why the more complicated the training, the higher the quality of the knowledge acquired by artificial intelligence.
General Purpose Intelligence vs. General Artificial Intelligence
This is one of the biggest differentiators that we have to make when talking about the evolution and growth of artificial intelligence.
We have identified the characteristics that a general artificial intelligence would have. https://open.spotify.com/episode/27vhtTJQc1WOZA5HpMiBqf
A general-purpose intelligence has the ability to generate knowledge in a specific area.
In this case, DeepMind technologies are being tested in video games from the 90’s and 00’s with great success.
This technology, when tested and increasingly improved, could make the leap in sectors such as health, education or sports.
How DeepMind could change the Artificial Intelligence landscape
Through the accelerated use of general purpose artificial intelligence, the number of sectors that accept this technology could be increased.
We currently see that impact on our phones, computers and different sectors.
The processes could improve considerably, increasing the evolution of science and technology.
By increasing the “intelligence” of entities and people, one can think of creating new sciences and solving serious problems for humanity.
In fact, it is possible that through the increased knowledge generated by general purpose intelligence, we could be able to create a General Artificial Intelligence.
Controversies, curious facts
Elon Musk is an investor in DeepMind, his main intention being to have access to developments and to monitor possible damage to humanity through artificial intelligence.
After AlphaGo beat Lee Sedol, there was a surge of concern over the constant growth of artificial intelligence and its possible superiority over humanity.