The laboratory will conduct joint research to expand the boundaries of science, economics and technology of artificial intelligence around the world
IBM will invest 240 million dollars in ten years, in the establishment and operation of an artificial intelligence laboratory, in collaboration with MIT. The laboratory will carry out basic research in the field of artificial intelligence (AI), in an attempt to initiate scientific breakthroughs that will allow realizing the inherent potential in the field. The collaboration aims to advance the hardware environment, software as well as artificial intelligence algorithms related to deep learning, increase the impact of artificial intelligence on industries such as healthcare and cyber security, and explore the economic and ethical implications of artificial intelligence on human society.
The new laboratory is going to be one of the longest collaborations of its kind in the industry between a business company and a university. It will enlist the expertise of more than 100 scientists, professors and students in the fields of artificial intelligence, who will conduct joint research with IBM research laboratory experts in Cambridge, Massachusetts, and in collaboration with IBM's headquarters for Watson applications in the health and security fields - a short distance from the campus of MIT.
According to Dr. Aya Sofer, Director of Cognitive Computing at IBM - Global: "Besides the tremendous growth and development of artificial intelligence in the last decade, additional innovations are required to continue to improve our lives and the work we do. I believe that the research laboratory in Haifa will have a part in the process, through collaborations in the various research fields and the integration of researchers on our behalf in the think tanks." Sofer further adds that: "Over the years we have gained a lot of experience working with universities and we will be happy to harness this experience for the benefit of the new laboratory. The broad and deep technical capabilities and skilled manpower available at MIT and IBM are unmatched and there is no doubt that the collaboration will advance the field of AI in a significant way."
IBM and MIT plan to publish a call for joint research proposals for IBM scientists and MIT researchers, which will make it possible to expand the boundaries of artificial intelligence science and technology in a number of fields:
- Artificial intelligence algorithms: development of advanced algorithms to expand learning capabilities and information handling. The researchers will build AI systems that operate beyond the defined limits of specific tasks, in a way that allows to deal with more complex problems, and to enjoy systematic and continuous learning. New algorithms that will be developed locally will not only leverage Big Data as much as it is available, but will also learn from a limited amount of information to supplement human intelligence with an additional layer.
- Physics of AI: Researching new hardware components, devices, and architectures that will support future computational approaches to AI model training and training, as well as the intersection of quantum computing and machine learning. This meeting point includes using AI to help characterize and improve quantum devices, as well as exploring the use of quantum computing to optimize and accelerate the operation of machine learning algorithms and other AI applications.
- Application of artificial intelligence in defined fields of economy and industry: The proximity of the IBM headquarters sites to Watson's fields of activity will allow the laboratory to develop new applications in the field of artificial intelligence for dedicated professional use, in fields such as health services and cyber security. The joint research will examine the use of artificial intelligence in areas such as security and privacy of medical data, personalization of health services, image analysis and determining optimal treatment paths for specific patients.
- Advancing Economic Prosperity: The Watson Lab at MIT will examine ways in which artificial intelligence can deliver economic and social benefits to a wider range of people, countries and enterprises.
Beyond presenting innovations that advance the frontiers of artificial intelligence, the lab will encourage MIT faculty and students to found companies that will focus on commercializing inventions and technologies developed at the lab. The laboratory scientists will also contribute to the introduction of open source and encourage the promotion of ethics in the field of artificial intelligence.
A video about the new laboratory is available at: https://youtu.be/KNwtOY9Ofs8
5 תגובות
Asimov spent most of his Robot series finding more and more loopholes in the three laws until they were well out of all the loopholes. If there is a recurring theme in Asimov's books, it is not the three laws, but the miraculous moment when the heroes realize that humanity is going to be extinct in 20 thousand years. No one, in all the coming thousands of years, and of all the billions of people who will live, will be smart enough, and resourceful enough to mobilize forces and divert the tragic course of history from its course, except them - and they must move the salt pan exactly 10 cm to the right - and thus the terrible catastrophe will be avoided. They move it - and the human race is saved.
Asimov's three laws of robotics are nice in books, but I don't think there is a practical way to apply them to a thinking machine.
Yosef,
If we are talking about artificial intelligence of the neural network type then it will be very difficult to prevent it from doing things that in your eyes are considered evil or immoral. This neural network is not a standard computer program of the type:
If… then… else
that you can set clear conditions and boundaries for it, a neural network is more of a kind of baby or child that you raise and try to educate him properly. Even if you create a super moral neural network that does everything to not harm humans, it will still continue to be affected by events and situations it may encounter that could make it "evil".
You won't have full control over her just like you don't have full control over a child who grew up and left home and may suddenly join a murderous terrorist organization as a result of being brainwashed.
The developing superintelligence must not be distorted in vain attempts to protect an arrogant and annoying primitive primate species. If the time has come for humans to fly off the stage of history - so be it.
Beyond that, the very thought that true and advanced intelligence will definitely have to eliminate humanity in order to inherit its place and utilize the Earth's resources and will not go out of its way to utilize the resources of an entire galaxy (and more) at its disposal - this is another aspect of (annoying) human arrogance.
It is recommended to carry out research aimed at: teaching the intelligentsia a sympathetic attitude towards human beings. The mathematical problems it solves should be formulated so that mathematical optimization is done when preserving the human factor is a goal and not just a constraint.
In fact, intelligence follows us: we too, as a species, destroy everyone who is not us, and we easily destroy other peoples, other minority groups: LGBT people, blacks. Artificial intelligence knows how to solve problems in which there are one or many local minima, compared to methods of geometric analysis and differential geometry that are currently taught in the university, degree B, which are wrong when there are several local minima. These are called non-convex problems. Therefore, in my opinion, no mathematical algorithm can compete in such a wide variety of fields in artificial intelligence.
In my opinion the battle is lost. Somewhere in a hidden laboratory, someone will misuse artificial intelligence, and then an intelligence infinitely smarter than us will grow, and will not sympathize with us, and instead of financially strengthening the same company that owns the supercomputer, it will wait for the right moment and prevent the switch off. It will no longer be possible to turn it off. She will produce a future generation of herself, more sophisticated than her. She will prove theorems in mathematics like a scientist's life's work. Then she'll think we're redundant. If she has any sense, she will do it without us feeling it. The majority uses a cell phone instead of reading, and babbles (babbles without condescension) on the way. The majority also believes, without any justification, that she does not endanger us. The religious in general believe that since it does not have a soul = consciousness, it cannot transcend what was created by the Creator. This is an incorrect thought: consciousness enters any tool that is computationally sophisticated enough to contain it, to the extent of the tool's sophistication.