Потребителски вход

Запомни ме | Регистрация
Постинг
21.11.2021 16:16 - Изучаване в дълбочина Deep Learning Essentials
Автор: panazea Категория: Видео   
Прочетен: 181 Коментари: 0 Гласове:
2

Последна промяна: 21.11.2021 17:42


 Изучаване в дълбочина Deep Learning Essentials 
Universite de Montreal 

Deep Learning Essentials
Копиране на мозъка, моделиране на мозъка, ИИ да достигне човешкото ниво. Възможността на компютрите да получат някакъв вид интуиция в областта ,
в която са тренирани, да разпознават езици , да извличат всякаква информация от база данни.

Хората имат интуиция как да разрешават проблемите.
Всеки инструмент може да се използва за добро или за зло.
Хора , които искат да имат повече сила. Правителствата трябва да са внимателни къде е червената линия. Кое е приемливо.
2020 започна с риск от война. Може ли ИИ да промени начина на водене на войни? Да , една от употребите на ИИ, която за мен е проблематична , е да разпознава лица и за целите на сигурността могат да се създадат особен вид  поверителност, наблюдение и следене.... The reason that I"ve been excited about deep learning for so many decades is because it"s really at the intersection of two incredible quests. There"s the question of intelligence, what is intelligence? Humans are intelligent, how does it come about and then how can we use that knowledge in order to transform the world in a positive way and build intelligent machines. Is deep learning something of the future or is it already around us now? Well actually it"s been in products that people use since 2012 in your phones for speech recognition and then it"s you know moved out to many areas in vision and language translation so it"s already in many products that people use today but companies are building all kinds of new products and services based on deep learning today that people will use in the next few years so it"s gonna be is also something of the future. So why is such a MOOC needed? Because the world is changing thanks to AI and deep learning is going to be used across society in a transformative way and there are not enough people with those skills. The organizations which don"t start right now to build new products and services based on deep learning are probably going to be left behind. So tell me as the instigator of the lvado-Mila deep learning school, from which this MOOC is derived can you tell me how did you get the idea for this deep learning school and why is it necessary? For many years we"ve realized that there"s a huge demand a big need for skills around AI, machine learning and especially deep learning. This is the reason why we created Mila in the first place because there"s all of these companies students researchers who want to learn more who want to build on top of what we"ve already done in terms of research and applying that research in society and so there are schools and universities but something that really we don"t have enough are those kinds of courses for people entering the field and this is what the MOOC was really about. So what exactly is deep learning? Well deep learning is the continuation of research in neural networks, artificial neural networks you know, done by AI researchers not for the purpose of modeling the brain or copying what the brain does but solving problems for society approaching eventually human level artificial intelligence inspired by the brain and it"s called deep learning because on top of the existing ideas that we had in the 90s with neural networks it brought this idea of learning more abstract representations through multiple levels of representation. What kind of specific questions or problems can deep learning answer that couldn"t be answered before? Well in one word I would say the big breakthrough brought by deep learning is the ability of computers to acquire some kind of intuition about a particular domain area in which they"re going to be trained. So it could be in computer vision where they start to have visual intuition about objects so that they can generalize to scenes and categories they have never seen before or it could be in natural language understanding, it could be in the ability to extract information from time series, from all kinds of records and databases. So companies are starting to use this in places where humans can form judgments that are complex where it"s not enough to have a few simple rules but you need to consider a lot of data, a lot of variables and usually people have an intuition of how to solve these problems after a lot of practice. But do you feel that deep learning may be detrimental to society for example? As a powerful tool that we"re bringing to society, there"s always the potential for abuse because any tool, I mean humans are tool building animals, and any tool could be used for good, to help each other, or could be abused usually by people who want to have more power and so I think governments in particular have to be careful: where"s the red line, what is acceptable and they have to play that role so that we together move society in the right direction. So the year 2020 began with the risk of a war. Can deep learning actually change the way that wars are fought? Yes, unfortunately this is one of the uses of high-tech and AI which is in my mind problematic. It can be used maybe the most obvious case is for recognizing faces and that can be used in many ways in security that can create its own issues of  privacy and Big Brother control but in war also in order to build these killer drones self-driven drones that can target a person by using a camera to find the target person and assassinate them basically. And that clearly poses all kinds of moral and security challenges that the UN and the international community needs to address with an international treaty so yeah and then there are many other uses that military organizations around the world are studying for deep learning because it"s not just faces: you can recognize all kinds of things from the air, you can probably use these things to help strategists and analysts in the military so I think it"s important that societies think carefully about what is it we consider ethically acceptable and what isn"t. So, Dr. Bengio, can you tell me a little bit some concrete applications of deep learning for example in healthcare or maybe around climate change? Yeah I mean deep learning is being deployed and applied in many many sectors of society but one thing I find really exciting is the energy right now from students and researchers and companies looking for what we call AI for social good applications. So these include things like health care, climate change, fighting climate change, humanitarian application, so let me give you some more concrete examples. In healthcare it"s now a lot of applications involving the analysis of medical images. So example: helping doctors to spot cancer cells in images when you know, it might be a very small detail in an image that the doctor could miss and those systems can detect very reliably. In the case of climate change there"s a lot of research going on: we have written a very long paper about 10 different areas of application. It goes from things like more efficient uses of energy that we currently have to designing new materials for batteries or carbon capture to things like improving how we understand the future like climate models and how we can visualize them. And then on the humanitarian side, this is something that I care a lot about, how do we democratize these technologies to make them more impactful especially in places that are the poorest on this planet. So we"ve been working for example with the Red Cross on using satellite images in order to better characterize what is going on on the ground when there are some catastrophic events. We"ve been using deep learning methods to increase the resolution of the images from multiple passes of different satellites so that they can have better information. So there"s really a lot of applications and many people around the world are now thinking how do we join forces in order to collaborate on these kinds of AI for social good projects. And I imagine this MOOC can also be watched by people around the world so that they can also learn and perhaps participate.Yes. Yes, this idea of democratizing knowledge about AI and deep learning is super important because as I said earlier these powerful technologies can have a negative effect of giving more power to those who already have a lot and so we need to counter that force by these kinds of services where we make it more accessible we deliver a lot of open-source software we teach in places where there"s not a lot of expertise so I think that all these efforts are super important. As a woman scientist striving to promote other women in science, I"d like to know what do you think about the very low proportion of women in the field. How do we rebalance that? Yeah, it"s a big problem. It"s already a problem in computer science in general and what"s even more kind of concerning is the proportion of women in computer science has been going down over the years so it used to be say in the 30-40 percent a few decades ago and now it"s around 20 percent. I don"t think people have a clear understanding as to why but we see some places where a lot of effort has been made to make it more appealing for girls to get into these fields and I think we have to maybe also consider well why is it that you know so few of them are there? And maybe there is too much pressure, competitiveness it"s not good for people in general so we need to understand better and do experiments also. So I think different research centers, different institutions are trying different ways of recruiting for example of highlighting role models but really we don"t have a recipe yet and I think we need to keep looking. Maybe starting young also, in young girls? Girls coding camps for example absolutely. So do you have any suggestions or something to say to the people so that they watch this MOOC and learn on deep learning, do you have a specific comment? Yeah, so one reason why I"m so excited about this field and why I have been for many decades is because it"s at the intersection of two really appealing questions first who are we as humans what is intelligence how do we build intelligent machines? What makes us able to learn so many things and be creative and so on. And then how can we take advantage of that understanding to build a better society, to deploy these new technologies and help as many people as possible with it. So it"s these two things together that I think make a field like deep learning so appealing.
Deep Learning Crash Course for Beginners 
https://www.youtube.com/watch?v=VyWAvY2CF9c

 








Гласувай:
3



Следващ постинг
Предишен постинг

Няма коментари
Вашето мнение
За да оставите коментар, моля влезте с вашето потребителско име и парола.
Търсене

За този блог
Автор: panazea
Категория: Технологии
Прочетен: 6840834
Постинги: 3990
Коментари: 11585
Гласове: 56402
Календар
«  Март, 2024  
ПВСЧПСН
123
45678910
11121314151617
18192021222324
25262728293031