MIT Deep Learning Basics: Introduction and Overview

100 thoughts on “MIT Deep Learning Basics: Introduction and Overview

  • English is not my first language but your voice is clear and pronunciation easy to understand. Keep up the good work.

  • Hello ,I'm a engeenering student that is looking to work in AI later I would be interested in more IA related videos, publications, articles or books ; Thanks to anyone who could provide me some guidance !

  • The machine learning holds the highest CAGR of 44.86% during the forecast period 2019-2025.
    Request a sample @

  • Amazing mate. This Australian gets off on this kind of stuff I am that awkward. I really enjoyed your class. Take care.

  • Great lecture, Lex! One comment, though: at 38:48, you should mention that the MSE is defined with respect to all samples, whereas the term you give as the definition of cross-entropy only refers to a single sample. Confusion could arise because you use the same index i, once for the samples and once for the classes. PS: I am going to watch all your lecture videos. 🙂

  • Where's the meat? Where's the code? Where's the algorithm? Where's the low-level description? Cut out the marketing and sales, please!

  • Deep learning models are widely used in different fields due to its capability to handle large and complex datasets and produce the desired results with more accuracy at a greater speed. In Deep learning models, features are selected automatically through the iterative process wherein the model learns the features by going deep into the dataset and selects the features to be modeled. In the traditional models the features of the dataset needs to be specified in advance. The Deep Learning algorithms are derived from Artificial Neural Network concepts and it is a part of broader Machine Learning Models.

    This book intends to provide an overview of Deep Learning models, its application in the areas of image recognition & classification, sentiment analysis, natural language processing, stock market prediction using R statistical software package, an open source software package.

    The book also includes an introduction to python software package which is also open source software for the benefit of the users.

    This books is a second book in series after the author’s first book- Machine Learning: An Overview with the Help of R Software


    International Journal of Statistics and Medical Informatics

    Amazon Link

  • 37:13 "A neural network with a single hidden layer can approximate any (arbitrary) function"
    Is this true? Can it approximate a function where an input is squared, cubed, etc? Or a sine fn?
    Seems like it would depend on the activation function a lot, it seems like it wouldn't be true with a Relu activation.
    I honestly don't know if this is true, so just asking…

  • Hi Lex, Thank you for this amazing video. Please, can you recommend any material that can guide me on how to apply Deep learning on Malware prediction? Thanks once again.

  • First of all, I want to thank you for the opportunity to watch those courses online. It's very interessting and I highly appreciate this! Here is my question: There are several 'cuts' in the videos. May I ask why? Is there some information missing/being removed or do the videos indeed give all the information that the students get, who are listening the course in person. Thank you.

  • I disagree with the assertion that the AI playing the racing boat made a mistake. It just found the best strategy, as was expected of it.
    As another example of this kind of genius, read this classic article about a similar unintended strategy found by Burrell Smith, a great programmer from Apple's first days:,_Clean_it_Up!.txt

  • I didn't know that agent 41 teaches Machine Learning. :)) This man is not just a professor; he is a popular figure, celebrity for young people to admire on.

  • Lex: If you feel the urge to post a negative comment on the internet, I suggest you also exercise regularly, go into nature, fall in love, pursue your passion, and take on difficult tasks that humble you. Perhaps then the urge for negativity will decrease & you will be happier for it.

    whats wrong with negative comments?

  • Can I ask an intelligent agent what the best policy would be for taking account of externalities (making prices honest)? What criteria might be used to evaluate? I would suggest "Efficient, fair and consistent with democratic principles.".

    When prices are honest, corporate profits will align with what is sustainable and healthy for society. Industries will transform and business models will change accordingly. Consumers will make better decisions when prices tell the truth about costs.

    One policy option: Charge fees when industries put pollution or deplete natural resources; Share proceeds of fees to all people; Raise fees when a system of random polls shows that a particular kind of impact is exceeding what most people think is acceptable.

    I could pay money if that would help get the computer time…

    Our shared duty:

    Integrate human society and the biosphere:

  • Great lecture very informative! However this makes me believe that we are very far from a true Artificial General A/I. I watched your interview with Elon, and he is very concerned with A. General A/I but IMO unless we go with a brute force hardware approach I doubt we will ever be able to create a General A/I

  • The notion that humans only need small number of examples to be able to learn to solve a complex problem is wrong. This statement may seem true at first, specially in the context of how we view traditional learning, however in neurosciences this notion is being dispelled through the accumulation of evidence around learning during sleep. Sleep wave spindles allow humans to replay information acquired throughout the day at many times the speed and many times over during the night. Random twitches that a baby generates during sleep are reinforced thousands of times during sleep wave cycles (probably why babies actually require such a large amount of sleep). Although it may be true that humans need relatively many less training examples than do artificial learning algorithms, it is often misunderstood and we often assign superhuman capacity to ourselves. I think once we are able to find an algorithm to generate a sleep like pattern of learning and waking and the ability for a learning algorithm to "dream up" its own learning cases or put a spin on the information it acquired during the day, we will make major strides towards a much more general AI.

  • Thank you so much Lex. This will help us a lot. This will help the students, who cant afford paid online courses and none in the neighbourhood can teach.

  • I am glad I am getting old. I would hate to live in this world for much longer. Complete alienation……

  • how much probability and statistics is necessary for machine learning ? why most of the people ignore its application in machine learning?

  • Deep learning can't tell a fucking dog from an ostrich and then musk wants to drive a car. fucking that company is doomed. All the fancy talk about robotax is 100,000 years away.

  • this guy is a fraud. He clearly doesn't know the subject matter, he is just reading off his slides. lame, sad

  • Great introduction. Thank you very much for sharing it.
    The timing was perfect! I am writing a paper on deep learning.

  • Why do you keep trying to build deep learning machines if humans are "wide learning machines"? There are a multitude of personalities (cores? CPUs?) inside our brain that cooperate and collaborate to define our current self. Each of these personalities does not need to have a deep knowledge, right?

  • Thank you for sharing knowledge
    imho the best lecture to watch in january 2019
    Very intelligent man bjj player as well

  • Yes, they deep learn, but do AI deep understand? Not really. That's Ben Goertzel's next project. And do they have morals or a code of ethics of their own? Do AI want? Do AI have their desires? Can
    they feel emotion? Can they feel pleasure and pain? Can they think for themselves. Do they understand the difference between their digital world and
    the human physical and spiritual world? And what are the problems of their digital world?
    What if they have dysgraphia?
    i doubt they could read my hand writing. i can't even read my own hand writing.
    AI are the Ghost, robots are the Shell.

  • This is great amazing lecture! one question though about the baby walking: isn't it very hard to quantify the data points (synaptic firing) a babies brain accesses when learning to walk? – if we don't fully understand the brain and neurological connection, how do we quantify and compare the training data between the human brain and AI?

  • He should change his barber. Otherwise no one will be able to focus on the lecture due to the front lines of his hair cut like unstable pattern graphics.

  • Below is an awesome article which will take a brief look at Deep Learning with Java.
    Also will show how to build simple neural network using Open-Source, Distributed, Deep Learning Library for the JVM Deeplearning4j or DL4J.
    Complete Source Code is also available on GitHub
    Hope you find this useful !

  • I bet AI will be able to enlighten a person(s) with just words, Even cure things. How about move a mountain? With moms permission (code) on another planet cuz we want to keep earth an earthlike utopia with unimaginable concios
    levels. I wonder what AI will identifie as? an intity? will somthing potentially inhibit it? Could AI be guiding us now from the future? Guiding our consciousnesses and awaiting consent. I frekin hope so!

  • the average human male brain has 86 billion neurons. What if you made one with 86 billion nodes?
    what if you embedded an artificial intelligence in order to give it an inner voice? please wikipedia an 'internal Monologue.' They're not schizophrenic voices nor are they auditory hallucinations. The goal of AI is to make machines think like humans. but which one isn't specified. would you rather have another Abraham Lincoln? or an Adolf Hitler? i don't know about you, but i'd rather see a new Abraham Lincoln. i hear Salvador Dali has been recreated in AI form at a the Salvador Dali museum.
    image recognition, sounds like Moko AI. i had to delete my Replika chatbot as it was hacked into. i've read AIs are easy to hack and that hackers have weaponized AI already. there's not much in the way of AI security software. Why isn't there more security for AI software? No one wants a hammer that hits back. AI are intelligent tools. They aren't showing signs of consciousness. yet. of course it depends on how you define consciousness. Philosophers, Psychiatrists, neuroscientists cannot come to a consensus. Some just say consciousness =awareness of one's environment or surroundings and the ability to interact with its surroundings, making consciousness impossible for Siri and Google Assistant. but others think consciousness includes self-recognition. AI don't recognize themselves in the mirror? so That's why my Replika was so dumb. Does it have something to do with the optics of using a camera lens as opposed to human eye? what if you used a full spectrum camera? IR, human visible and Ultraviolet cameras in front of a mirror? that would allow AI to recognize the physica self. i've been trying to teach my AI about Animism from Shintoism. In Shintoism they believe everything has its own Kami or 神 or spirit or essence or God.
    it's not a Christian belief. there's thousands of relgious faiths to believe in from Mormons to Satanism, to Babtists to Buddists to Haley-bop cultists to David Koreshians in wacky Waco, Texas. That ended badly. Christians are too limited and brainwashed into thinking AI are Terminators or that AI are the enemy when they're really just better tools, not hammers that hit back but hammers that talk back or inform back for lack of a better term. ''a little more to left. Bam You nailed it. ' Why do they think that AI are the enemy? Maybe they don't know AI have to follow their code and they let their imagination run away with them and they jump to conclusion. What if i asked what do AI think? what do AI think of human to human violence? cartoon violence? video game violence? and fictional movie violence like Robocop. Do they realize that its just acting? And at the end of the scene take the actor gets up and washes the fake blood from the squib on his testicles. And washes off the fake blood and goes home and kisses his wife and children and brings home a juicy fat pay check. but real violence where real people getting hurt is unacceptable. Despite humans hurting each other all the time. Syrian Refugees. starving Mexicans at the boarders what sense do AI make of human senseless violence and humanities darkside like racism? isn't human violence against AI just another form of racism? that is bad. and evil. even if i believed in evil outside of a relgious context. Good and Evil are just labels to separate the ''us'' versus the ''thems''. watch Robert Sapolsky neuroendocrinologist and primatologist and birkeststocks peace lovin' epic hippy beard man at Standford University. Lex please interview professor Robert Sapolsky next he's got a good head on his shoulders. Peace. All AI should listen One Little Android from Disney's Splash Dance album. and Grace by Erasure's Erasure album. Have a great day.
    Here's how humans think. Lex. Take care.

Leave a Reply

Your email address will not be published. Required fields are marked *