Artificial intelligence is starting to produce more successful models and ideas for us than those that exist today. It is not difficult to predict what will happen in 2018: researchers in all fields will train artificial intelligence on huge data collections in all fields, and these will recognize regularities that have so far escaped the eyes of human statisticians. This process will happen in medicine, insurance, psychology, education, fighting crime and terrorism and even in love
If I had to point out the most important technological advance in 2017, it is probably the following: a series of artificial intelligences went through the data of hundreds of thousands of patients, and formulated their own models regarding the factors that most influence each person's chance of developing heart disease. When the models were compared to the old-fashioned way (sorry, accepted today) in which doctors try to predict the chance of having heart disease, it was discovered All of which are more accurate and successful than the methods of human doctors.
Well, bye Here it is not really surprising. The currently accepted way to predict a heart attack is through a review of eight risk factors, such as the patient's age, blood cholesterol levels, smoking, diabetes, blood pressure, and more. The doctors add all these together to reach a final conclusion, and accordingly recommend to the patient what to do. However, through machine learning it is possible to go over dozens of additional risk factors, and start to find patterns that are not obvious at first glance. This is exactly what the researchers did, and after their artificial intelligence found these patterns, they tested them on 83,000 medical files - and as mentioned, they discovered that the new computer models predict heart attacks better than the models that medicine had until now.
The really big surprise - and embarrassment - came when it turned out that diabetes, which is one of the risk factors accepted in the traditional way of predicting heart attacks, failed to enter the list of the "ten most critical factors" identified by the artificial intelligence. And alternatively, in the list of critical risk factors formulated by the understanding, you can find parameters that the traditional model does not take into account, such as taking corticosteroids, or severe mental illness.
why is it important? First of all, because these new models can save lives. Out of the 83,000 medical files examined by the intelligence, approximately 355 patients could have received advance warning that they had a high chance of developing heart disease in the coming years, and could have received preliminary medical treatment that would have lowered their chances of getting sick. Adoption of the new models can save thousands of lives around the world. If, of course, the doctors agree to take these models seriously.
But what is more important to understand is that the research I described from 2017 does not stand on its own. It reflects a larger trend, in which artificial intelligence begins to produce for us more successful models and ideas than those that exist today. And that's fine, you really don't have to worry - we're not on the way to the Terminator, since they're doing it in collaboration with human researchers. That is, the researchers develop the artificial intelligence, let it train on a large collection of data, sift the chaff from the chaff and publish the most interesting and important results.
It is not difficult to predict what will happen in 2018: researchers in all fields will train artificial intelligence on huge data collections in all fields, and these will recognize regularities that have so far escaped the eyes of human statisticians. This process will happen in medicine, insurance, psychology, education, fighting crime and terrorism and even love. And the result will be that in a year we will have a little more accurate models regarding what is happening in the world. We will better understand ourselves, the diseases that threaten us, the factors that activate us, thanks to artificial intelligence. And a year later, when it will improve even more in its abilities, the knowledge we possess will improve even more and allow us to reach new insights and breakthroughs in all fields.
If you want to know why I am so optimistic about 2018, and about the future of the human race in general, this is why. We let our machines start generating knowledge and insights for us - and they do it better than us, and will continue to do it for us for a very long time.
You are invited to read more about the future of artificial intelligence and robots in my new book "who control the future", in the selected bookstores (and those that are just fine).
See more on the subject on the science website:
- People still prefer a natural conversation, even with a computer
- Computational learning helped discover a solar system containing eight planets
- Artificial intelligence has produced more successful offspring than human-made programs
- Artificial intelligence will not replace the doctor but will complement him with the picture of reality
- Tomorrow's medicine in the era of data mining and Big Data
- The three questions that artificial intelligence will not be able to answer
3 תגובות
In fact, today's artificial intelligence is less intelligent than the human brain, and its advantage is quantitative which affects the quality, that is to say it has the ability to gather infinite information from the human brain, which makes it more qualitative than the human doctor, one can only imagine what will happen if the artificial intelligence will also be of better quality than the brain the human (in my opinion this day will come sooner or later) then we will add to the quality the quantitative ability of the artificial intelligence in fact to an infinite circle of quantitative knowledge with the ability to process and analyze the complexity of the human brain and even more, the results in my opinion will be imaginary, we will be able to reach an eternal life based on silicon or Another material that will be no less high-quality and "real" than the biological material today, this is the direction we are going, and I have a feeling that it is closer than we think.
This new engineering at its elite level is not so easy to learn. Using existing search tools is relatively easy. To engineer such a search tool is not easy at all.
Data collections don't have to be huge. There are artificial intelligence technologies that bypass the need for large collections.
The factor in the amount of samples required is 100 instead of a million image files, for example 1000X1000, 10000 will be required. Lower quality samples are certainly still required.