What is the ambition of DeepMind, led by artificial intelligence development genius Demis Hasavis, and the battle with Google's parent company?
by
Born and raised in London, Demis Hasavis has had a career in game development, and in 2010 has been CEO and CEO of DeepMin, a company at the forefront of AI research. DeepMind was acquired by Google for $ 500 million (about 55 billion yen) in 2014 and attracted attention, and in 2016, AI 'AlphaGo' developed by DeepMind beat the top players in Go and has a great response I called. Hal Hodson, a writer for the British 1843 culture magazine, reviews Mr. Hasabis's personal image and ambitions, as well as the battle between DeepMind and his parent company, Google.
DeepMind and Google: the battle to control artificial intelligence | 1843
https://www.1843magazine.com/features/deepmind-and-google-the-battle-to-control-artificial-intelligence
In August 2010, Mr. Hasabis made a speech on General Purpose Artificial Intelligence (AGI) at the annual Singularity Summit ( Singularity Summit ) . Unlike AI, which specializes in a specific field, AGI, which can solve various tasks like humans, limits intelligence only to available processors, unlike human brains, which are limited to brain capacity.
Mr. Haservis argued that it is important to understand AGI's realization as a software of the brain, not consisting of controlled programs or physically copying the human brain. Researching how the brain processes information and activities using MRI, etc. will help develop AI and AGI, Mr. Habasis said at the summit.
In fact, at this summit, Mr. Hasabis made his first public announcement about DeepMind. 'Hasabis's talk at the Summit was an appeal to Silicon Valley investors, and it was aimed to get an investment in DeepMind,' Hodson notes that DeepMind is about 2 million (approximately Succeeded in raising investment of 290 million yen). And in 2014 Google acquired DeepMind for $ 500 million (about 55 billion yen), and has succeeded in getting a 50-fold return from the initial investment.
Early childhood Hasabis is one of the world's leading chess players, and at the age of 13, he ranked second in the world by age group rating. However, Mr. Hasavis, who was certified as a high school graduate at the age of 16, started his career as a game developer at
Having left Bullfrog Productions and moved to Cambridge University in 1997 to study computer science, Mr. Hasabis participated in a student go tournament hosted by Charles Matthews, a Go champion at Cambridge University. Mr. Hasabis, who was still poor at Go, won the beginner tournament and then beats the student who won the experience tournament as well. Mr Matthews says that he was shocked by the fact that he had been a beginner, and that he had beaten the experienced.
Mr. Hasabis has had a very strong interest in 'games' since that time, and his games have greatly stimulated intelligence. Hodson speculates that he might start wondering if he could do the same with computers as he had accumulated his experience and raised his gaming skills. In 1998, Mr. Hasabis established a game company called Erikser Studio based in London. After the release of games such as ' Republic: The Revolution ', which is a political simulation game, since Erickser Studio was shut down in 2005, Mr. Matthews should have Erickser Studio just to gain management experience. He believes he has established it.
by
After that, Mr. Hasabis will start research in neuroscience. Research on neuroscientific fields has been essential for Mr. Hasabis who thought it important to understand the human brain in the development of AI. Mr. Hasabis who had not mentioned biological fields up until now, but some of the papers that 'the patient who damaged hippocampus not only cause memory loss but can not imagine himself to have a new experience' While being questioned by researchers, it has received a very high evaluation, such as being cited 1000 times or more.
DeepMind, founded in 2010 by Hasabis, attracted many talented people from all over Europe, and in DeepMind, talented people from all over Europe focused on machine learning and did pioneering research. This machine learning method is to collect information about the environment and accumulate experiences repeatedly, and it is based on the human learning method by accumulating memories and experiences supported by Mr. Hasabis. As a result of this, it is said that AlphaGo achieved the result of defeating the Go champion in 2016.
Google launched the acquisition of DeepMind in 2014, and it was said that at this time Mr. Hasabis's emphasis was 'protecting DeepMind's independence.' DeepMind was also attracted by Google's funding, but Mr. Hasavis had already developed a game and had plans for getting research funding. So, when we made an agreement that 'Google can not manage DeepMind's intellectual property unilaterally', and DeepMind created a technology that leads to AGI, a contract that the ethics committee, which was established exclusively, will manage AGI technology It was also done. Mr. Hasabis seems to have emphasized DeepMind's independence, even after being acquired by Google, to make ethical monitoring work.
Also, Hodson points out that the creative project that Mr. Hasabis gives to his employees and the research environment freed from the annoyance of the conference strengthens the relationship between DeepMind and his employees. A kind of loyalty directed against Mr. Hasabis is working strongly in DeepMind.
Being independent, DeepMind has been sending out a lot of research results, but on the other hand, the research costs charged to Google are huge. However, when Google is publicly criticized for privacy issues etc., the result of DeepMind's AI research with the intention of improving the world will lead to the evaluation of the parent company Google. Another positive factor for DeepMind is that Larry Page, who is a co-founder of Google and now CEO of Google's parent company, has shown an understanding of AI.
However, in recent years there have been concerns that DeepMind's independence could be threatened by Google. DeepMind Health, the health department of DeepMind , has developed an application called 'Streams' to notify doctors and nurses when the patient's health has deteriorated, and has cooperated with NHS , a UK state-owned medical service business. It has been operated.
Because of the nature of the app, Streams accesses sensitive patient information, and DeepMind has been cautious about handling sensitive information. However, in November 2018, Google declared the establishment of its health care department, Google Health, and announced the consolidation of DeepMind Health five days later. On this move, DeepMind says that a hospital contract signed by DeepMind Health does not automatically transition to a contract with Google, but employees of DeepMind Health are showing anger, Hodson says You
Right groups alarm that medical data of 1.6 million hospital patients are now held by Google-GIGAZINE
Also, Hodson points out that while it is very attractive for AI developed by DeepMind, AIs that have repeatedly learned only grow their skills only under the conditions originally set up. Even if it is a change of a subtle condition that human beings can cope with immediately, AI will take over learning so far and it will be difficult to cope.
And what has been emphasized in the development of AlphaGo, etc., is a system that sets AI to reward better results for complex games, and trains AI to get higher rewards. is. Remuneration to rules is an important element in AI development, but in the real world where rules do not move like rules, complex conditions are intertwined, so it is necessary to develop an AGI that addresses various issues. Hodson thinks it is very difficult.
If DeepMind was able to develop an AGI that solves all the problems of this world as Mr. Hasavis hoped, Mr. Hasavis will be able to clear the more difficult games than any of the games he has ever tried. Hodson said.
Related Posts:
in Note, Posted by log1h_ik