Learning algorithms manage to bypass the information overload problems faced by "traditional" programming thanks to the fact that they allow computers to acquire knowledge intuitively, more similar to human learning. This is how the machine, after being repeatedly exposed to facial images or speaking voices, succeeds in recognizing the recurring - deep - patterns that underlie the information.
Dr. Ohad Shamir. Complex skills. Photo: Weizmann Institute
Anyone can nowadays buy a camera that knows how to focus the image on the subject's face; Sophisticated computer systems use voice recognition to convert spoken words into written text, and even to identify the speaker. These are two examples of the advantages of machine learning. Learning algorithms manage to bypass the information overload problems faced by "traditional" programming thanks to the fact that they allow computers to acquire knowledge intuitively, more similar to human learning. This is how the machine, after being repeatedly exposed to facial images or speaking voices, succeeds in recognizing the recurring - deep - patterns that underlie the information.
Learning algorithms of this type help computers acquire skills that humans adapt to themselves while they are babies. But what about more advanced learning skills, ones that are a challenge even for mature humans? Dr. Ohad Shamir, who recently joined the Department of Computer Science and Applied Mathematics at the Weizmann Institute of Science, focuses on computational learning in cases where various constraints make it a complex challenge. An example of this is the integrated and complex knowledge required of a doctor, even for his most basic activity: in order to correctly diagnose the patient he It relies on the results of several tests, on the patient's reports of various symptoms, and on his personal experience. For practical reasons, the number of tests conducted is limited to the patient, and the amount of resources available to the doctor for the purpose of the decision is limited. Assuming that a computer gets access to the same partial information, can it learn to diagnose a disease like an expert doctor?
An even more complex example is the human brain. One of Dr. Shamir's goals is to study computer systems called "artificial neural networks," which know how to process information using methods that mimic brain activity. "The computer systems are hierarchical," says Dr. Shamir, "similar to the layers of nerve cells in the brain, which transfer the absorbed information from layer to layer, while processing our perception and our thoughts." Studying the artificial neural networks with the means available today requires a lot of effort and adjustment of many fields. Dr. Shamir hopes to develop software that will perform the learning tasks more automatically, and without the need for expertise in this field. Many computer scientists in the world hope that in the future such systems can form the basis of advanced artificial intelligence.
Dr. Shamir's other areas of interest include researching solutions for computational learning in cases of large, distributed and "noisy" networks, such as search engines. "Huge networks, such as Google, which are very large and lack organization, tend to suffer from learning limitations," he says. "For example, algorithms that require access to large amounts of information are problematic when the information is distributed among many computers around the world - because in this case they will have to invest effort And a long time actually communicating with all sources of information."
personal
Dr. Ohad Shamir's joining the Weizmann Institute of Science is, in many ways, a homecoming: he is now in the same department where his father, Prof. Adi Shamir, works. Ohad grew up in Rehovot, and attended classes for science-seeking youth at the institute. He did all his higher studies at the Hebrew University in Jerusalem, and within this framework he began researching in the field of computational learning.
He did his post-doctoral research at the Microsoft research center in Cambridge, Massachusetts. "Microsoft gave me a lot of freedom to pursue my interests. Means and infrastructure are not a problem there. Being there also gave me the opportunity to make connections with several people who work on various practical applications in the company, and these connections led to the registration of several patents." While living in Boston, Shamir married Michal, who is currently finishing her doctoral thesis in Indian studies at Harvard University. His wife's studies served as a successful excuse for Shamir to fulfill his favorite hobby - traveling the world, during which, among other things, he climbs mountains. Although these days he is busy starting his research at the institute, he hopes to find time in the future to continue this hobby.
10 תגובות
No. I just forgot and was lazy for a moment to jump to the website.
She deserves all the credit regardless of being a woman.
I realized that she invented encryption algorithms with a random motif that jumps from point to point. Sounds very difficult to crack. I don't understand it.
Joseph
Her name is Prof. Shafraira Goldwasser - she won the Turing Award in 2012 - you don't mention her name because she is a woman? 🙂 🙂
You're right, I got the name confused. The amount of publications by Adi Shamir is about 50 times greater, that is 300. I did not measure exactly.
The Faculty of Mathematics and Computer Science at Weizmann includes about 30 faculty members and some of them are very well known. Professor David Harel who was accepted this year together with Shechtman to the American Academy of Sciences, Adi Shamir who is a member of the American Academy of Sciences, and one female professor who won the 2013 or 2014 Turing Prize for encryption systems.
Joseph
You wrote "But the amount of Adi Shamir's publications shows that he is talented by virtue and not by grace." … Now I understand you meant a fan.
The list of fan's publications:
http://www.wisdom.weizmann.ac.il/~shamiro/Publications.html
Again: it seems that he has a bright future, and not from the ordinary. That is, it exceeds the average even of a candidate for a faculty member.
The academic way requires him to publish about 3 considered articles per year (which will be accepted) and I have seen fugitives of this requirement. On the other hand, I have seen stars who have no difficulty in meeting the task, even when they do not have doctoral students. That means the system will filter out those who are not worthy. By the way, I saw a Nobel Prize winner who did not meet the requirement (Green - the developer of the renormalization in liquid vortices), and when they forced him to write an article, so as not to leak, the article won him the Physics Nobel. And we saw Andrew Wales who secretly worked for 7 years on Fermat's proof, and met the academic requirements at the same time. The result was an important contribution to the theory of bundles (Tanima Shimura's conjecture and Kollivgin Flach theorems, and application in the proof of the Poincaré hypothesis and from there the quantum theory). All of this leads us to the conclusion that if someone is talented, he will find a way to succeed, and some attempts fail, because this is the way of natural selection.
On the path of the Heitol: the question of whether, in my opinion, we should reject the Bernoulli family. The answer is no, but it is possible to formulate a problem for matriculation five units in Bernoulli distribution with a decision tree:
If he follows Bernoulli on the Tel Aviv staff, then Yitzhak Bernoulli will apply to Hebrew, Weizmann, Bar-Ilan, BS, Haifa. And David Bernoulli will apply for what is left and also abroad.
But with so many Bernoulli, in the end more than one family member will get a position in mathematics at the same university.
RSA together with two other inventors - correcting my lack of emphasis.
I noted that I checked the list of publications and found it very impressive. The guy published about 20 articles in newspapers, many in conferences, and 3 pre-prints, which means they are awaiting publication, as far as I understand. To combine a post doctorate is excellent. There are professors who are in this amount in their professorship. I think he deserves it!!! It is said about this: "Good morning" is also eaten here for breakfast, and it actually adds pepper to the site." Secondly, Professor Shamir, he is not just a professor but
Invents the RSA algorithm. If he receives a penny from each RSA application (license), he is likely a billionaire. That is, the benefit to the industry.
Second: yes. I am an associate researcher at one of the universities in Israel, and a doctoral student, and from a closer look, in my subjective opinion, there is a preference for relatives. It's a human trait. It is clear that if he is not qualified he will not stand up to the academic competition, but with the number of candidates today it is sometimes not enough to be qualified to be admitted to the faculty (I am not at this stage) and protection is needed to put the candidate's CV at the top of the list, from some dean to a professor on the faculty.
And in our country - I saw it all the way, and it is probably a human trait and a global phenomenon. As if we brought it from exile. Look at the immigration from Russia, which I cherish as talented. But when they meet each other, it's a bit of a closed club, and they start talking in Russian over your head, and you have to politely ask them to switch to Hebrew, because otherwise you're out of the game.
I would prefer without preference, but it is also possible with it, it is only required that you have a fast rate of writing articles, and that the circumstances of livelihood and family do not get in the way. In other words, the Torah can still be issued from poor people (Ada Yonat). Most of us compete with no relatives.
Joseph
"Two more inventors" - they have names: Ron Rivest and Leonard Edelman.
Beyond that - how exactly do you propose that a Shamir fan avoid being his father's son? Are you suggesting that he be discriminated against because he is the son of? Are you implying that he got where he got because of his father?
Unlike the civil servants of Israel, who are far from keeping the barriers to family closeness (you don't need examples, do you?) - I don't think any "son of" in the academy would have gotten far just by grace. There are, for example, 7 members of the Bernoulli family - are you proposing to disqualify their discoveries? Likewise, Charles Darwin - shall we disqualify him because of his grandfather?
Indeed, we are rapidly approaching self-aware artificial intelligence, and it is possible that the network is such at the beginning of its journey.
But there is a feeling that this will lead to the transformation of the human race into a kind of slaves, for the masters of the network in the first instance, like the slaves that the cows use for us. The movie Elysium (Latin for heaven) with Matt Damon depicts this kind of reality.
In the second stage, the matrix, which was a movie from DAB, is getting closer.
Adi Shamir is the inventor of the RSA cipher together with two other inventors.
I would say that for the appearance of requiring civil servants to avoid family ties, this should also be enforced
in academia, and there is some truth in that. But the amount of publications by Adi Shamir shows that he is talented by virtue and not by grace.