Even with chefs - you don't need as many chefs as would have needed without chef tools. So instead of 3 chefs, you just need 1. With AutoML, you don't need a team of data scientists, just one is good enough.
Allen Yu At this point I think even Data scientists become less meaningful, it’s so easy that anyone can do it ( why even consider humans anymore ...) The thing that I don’t really like is that this will certainly lead to a centralization of these tools ... I guess it will end up that only few companies will own these platforms, bringing less democracy and of course even less transparency.
There are advantages to using auto ML systems, but we have to truly understand data science if we want to make the most of these advantages. (If we use Siraj's analogy of a chef and his utensils, imagine giving a food processor to a skilled chef and the same food processor to a person who knows little or nothing about cooking. Whose meal would you prefer eating? 😊) Also, these auto ML systems will become more sophisticated and more common. It's one of the things we humans do: we constantly try to automate our tasks. There's no stopping technology. What I recommend is giving each of the systems Siraj mentioned a try. See if any of them can help you with part of your work. See what they are missing. See how you can use their limitations to your advantage. For example, these systems can be easily fooled by slightly altered data, they can be hacked, they can't be use to create competitive advantage (once everyone is using them), etc. It's foolish to think that auto ML means mindlessly dumping data sets into a black box and getting smart insights out, and it won't be long before people realize that. In the meantime, like I said, get good at understanding auto ML. Think of it as of a super cool, super fast car. You can use it for some awesome rides (until a cooler car comes along. 😊). Also, you can still use your regular modes of transportation. Just because cars are fast, that doesn't make them the smarter or better way of getting somewhere. 😉
Bianca A. - There's art to data science Yeah Good point and clever observations. There is no stopping technology and of course it could be quite helpful in expert hands. There is no bad means at all, it all depends on how and who uses it. Unfortunately I know a LOT of foolish people that would literally throw data in and expect good results without any kind of efforts; that’s why I don’t favor these kind of things, not because of the systems themself but because of people.
Hey Siraj, just wanted to let you know that I think that you've made a lot of progress in the quality of your videos and the explanations you're giving. I have not been able to work on AI for a while due to a very busy year, and the difference from a year back and now in terms of the way you speak, the words you say, the timing and clarity of visualizations (and how they focus on what you're talking about better so they add to the explanation and don't distract) have all greatly improved. Thank you very much for all the informative content you put out. Glad to see you still going, keep it up!
Great video! I don't think the problem is that AutoML doesn't work for unsupervised learning, rather just that it hasn't been applied much. You could easily imagine using an NAS technique or DARTS on an auto-encoder or a rotation prediction task. Thanks for mentioning Particle Swarm Optimization, I was looking for a term for that! I'm curious what the differences are between Population-Based Training and Particle Swarm Optimization, or if they refer to the same idea?
Hey Siraj, I agree with everything you said here, and I'm glad you are always bring up the most current topics. However, I haven't seen you make a video about the latest and best technologies for data pooling, and data manipulation. Where and how should I pool data from many data sources so that I can easily run ML models on it (including AutoML)?
2:03-How to frame DataScience Problem 2:48-DARTs 4:00-Grid Search and Random Search 4:30-MetaHeurstics 5:10-Simulating Annealing 7:42-ML Box 8:20-Auto-Sklearn
Data itself is a kind of inference. We are supposed to find more inferences. I.e. Combining basic level inferences using And, Or etc, at multiple levels can help reach manageable set of inferences. Levels of inferences: level 1 inference is data, level 2 inference is what data is pointing to , 3rd level inference is combining inferences using And, Or, then grouping similar inferences. And then continuing the process.
Aha. I like that! I have some ideas on ml implementations but the knoledge required to get started is a steep learningcurve. That energy is put to better use if i spend that time defining the problem and its datasets.
Thank you for this video. My knowledge was limited to a few automl solutions I had heard about. Now I feel like I have a better understanding of what steps the automl ecosystem tries to automate, and some good candidates to benchmark for my problem. Super useful video.
Any suggestion of one of these or another that you can import JavaScript not Python? Also for pattern recognition learning. Any help would be appreciated. Great video.
In the explanation of the Bayesian, I think the thing that you call it activation function is actually called "acquisition" function. But I'm not sure whether people call it also "activation function" or not!
Great video, you should also consider AWS Sagemaker which provides a great implementation of Bayesian Optimization for hyper-parameter tuning among other great tools.
I love h2o because I can use it in R or Python, and it has a similar api. PLUS! you can export the model and import it into one or the other. Pretty slick.
Dear Siraj, I was to ask you about ml or ai with the project of neuralink (at least something or your saying about it). but here you said that the presentation was awkward. please explain what actually going on because I also found that people are not believing on neuralink. Thank you and GOD may keep you safe and prosperous.
If you use ml on something with time you should try using time as an independent object instead of a global factor. Think of it making the data more like sonar instead of a picture.
At robocup, a competition which aims to make robots win a match against the fifa worldcup champions, the teams uses ml to improve their robots/agents. There are a lot of categories, from soccer simulation to humanoids robots
Hi Siraj thanks for the awesome contents you offer very very valuable useful and inspirational. I have a quick question in a GAN can the generator be considered as a creative aspect(left brain ) and discriminators (right brain)
Why? by it’s very essence being a random number generator can be similar to chaotic system or far from equilibrium state causing a sort of self organisation to emerge from within
@@MohanKumar-gj9th By "no" I mean not necessarily. Don't worry about what left brain or right brain does!! You understand what is GAN... that's great and enough.
Singularitynet is Democratised decentralised AI ecosystem of ecosystems... Ai in a mesh network for emergent Intelligence. Partnered with Dominos, Ping An (10th Worlds Largest Company) and many more. Look it up!!! this is the future of AI
No you do not. Software Development is not for everyone though. If you're chasing it for money then consider other options like sales, financials, etc...
Siraj I would like AI to tackle the biggest scientific and theological questions in the world! Q1: Does God exist? Q2: Which religion is the most consistent within all the analyzed data? Q3: Can the answers to these 2 questions be statistically/probabilistically represented? Not sure how to break these logical problems down, so AI can solve them realistically! Do you think this is possible with AI's current state?
It's a very discouraging video. But for the lack of my competence I cannot fully comprehend how it will or may affect data-science positions. Now I'm on my way to making data_science career or even just presenting it as my hobbie while working at McDonald's, because I've found out that I like it and this study field makes me come back again and again every day, regardless that I had long time gaps in my learning journey. I spent quite a bit on learning and now don't want to give up on that. And I don't know how to express that I'm so f*ck*ng concerned with what will happen next
1999 I spent the summer “coding” html making websites with angelfire.com. 2019 Im watching machine learning videos on TH-cam. Realizing I really missed the curve.
Previously you seemed to be a great fan of Elon Musk, but in your recent videos you kind of throwing negative comments on him. Why ? what changed your mind?
Elon has been talking a lot and hasn't been delivering. Neuralink is a bit of a joke when he's(Elon) juggling so many unreachable goals. Ambition is great, but Elon has yet to achieve the promises he set out five/ ten years ago.
to be totally honest i'm a bit salty that he didnt respond to my very public request for an interview. Also, the AI doomsday talk is misplaced. It's not the statistical operations he should be fear mongering around, its the intents of the public corporations using them to exploit people
Sir I need small help how can I build a model to recognise the text written in different font ( ex calligraphy ) and display output as normal text. Can you guide me. Thanks in advance 😊
But most people don't eat from the hands of chefs everyday. They cook for themselves.
That makes them chefs
Even with chefs - you don't need as many chefs as would have needed without chef tools. So instead of 3 chefs, you just need 1. With AutoML, you don't need a team of data scientists, just one is good enough.
Allen Yu At this point I think even Data scientists become less meaningful, it’s so easy that anyone can do it ( why even consider humans anymore ...)
The thing that I don’t really like is that this will certainly lead to a centralization of these tools ... I guess it will end up that only few companies will own these platforms, bringing less democracy and of course even less transparency.
There are advantages to using auto ML systems, but we have to truly understand data science if we want to make the most of these advantages. (If we use Siraj's analogy of a chef and his utensils, imagine giving a food processor to a skilled chef and the same food processor to a person who knows little or nothing about cooking. Whose meal would you prefer eating? 😊)
Also, these auto ML systems will become more sophisticated and more common. It's one of the things we humans do: we constantly try to automate our tasks. There's no stopping technology. What I recommend is giving each of the systems Siraj mentioned a try. See if any of them can help you with part of your work. See what they are missing. See how you can use their limitations to your advantage. For example, these systems can be easily fooled by slightly altered data, they can be hacked, they can't be use to create competitive advantage (once everyone is using them), etc. It's foolish to think that auto ML means mindlessly dumping data sets into a black box and getting smart insights out, and it won't be long before people realize that.
In the meantime, like I said, get good at understanding auto ML. Think of it as of a super cool, super fast car. You can use it for some awesome rides (until a cooler car comes along. 😊). Also, you can still use your regular modes of transportation. Just because cars are fast, that doesn't make them the smarter or better way of getting somewhere. 😉
Bianca A. - There's art to data science
Yeah Good point and clever observations.
There is no stopping technology and of course it could be quite helpful in expert hands. There is no bad means at all, it all depends on how and who uses it.
Unfortunately I know a LOT of foolish people that would literally throw data in and expect good results without any kind of efforts; that’s why I don’t favor these kind of things, not because of the systems themself but because of people.
Hey Siraj, just wanted to let you know that I think that you've made a lot of progress in the quality of your videos and the explanations you're giving. I have not been able to work on AI for a while due to a very busy year, and the difference from a year back and now in terms of the way you speak, the words you say, the timing and clarity of visualizations (and how they focus on what you're talking about better so they add to the explanation and don't distract) have all greatly improved. Thank you very much for all the informative content you put out. Glad to see you still going, keep it up!
thanks for the helpful feedback!
Siraj 2016-2019 he was data scientists
Automl yes
Great video! I don't think the problem is that AutoML doesn't work for unsupervised learning, rather just that it hasn't been applied much. You could easily imagine using an NAS technique or DARTS on an auto-encoder or a rotation prediction task. Thanks for mentioning Particle Swarm Optimization, I was looking for a term for that! I'm curious what the differences are between Population-Based Training and Particle Swarm Optimization, or if they refer to the same idea?
I like the way you think, moar autoML for unsupervised learning plz, and yea they are basically the same
@@SirajRaval Awesome!! Thanks for the reply!
I attend School of AI Program at Chennai and now I love you more than before coz you don't stop in TH-cam and Take offline meetups and conference too
Siraj you make the best of the explanations dude, awesome
Sometimes I press like before watching your videos.
That was very interesting topic, thank you. 👍
He's so good at explaining complex stuff.
thanks so much! I'll keep improving for you
Hey Siraj, I agree with everything you said here, and I'm glad you are always bring up the most current topics. However, I haven't seen you make a video about the latest and best technologies for data pooling, and data manipulation. Where and how should I pool data from many data sources so that I can easily run ML models on it (including AutoML)?
2:03-How to frame DataScience Problem
2:48-DARTs
4:00-Grid Search and Random Search
4:30-MetaHeurstics
5:10-Simulating Annealing
7:42-ML Box
8:20-Auto-Sklearn
Data itself is a kind of inference. We are supposed to find more inferences. I.e. Combining basic level inferences using And, Or etc, at multiple levels can help reach manageable set of inferences. Levels of inferences: level 1 inference is data, level 2 inference is what data is pointing to , 3rd level inference is combining inferences using And, Or, then grouping similar inferences. And then continuing the process.
Wonderful as usual!! I like the way you compress so much information in a short video.
Than you again for the shout-out to Ludwig :)
Aha. I like that! I have some ideas on ml implementations but the knoledge required to get started is a steep learningcurve. That energy is put to better use if i spend that time defining the problem and its datasets.
Thank you for this video. My knowledge was limited to a few automl solutions I had heard about. Now I feel like I have a better understanding of what steps the automl ecosystem tries to automate, and some good candidates to benchmark for my problem.
Super useful video.
Thanks Siraj! Working on a paper in this area and this was a great overview!
The only point to remember -
There's always need for human being and will never be replaced.
Thanks Siraj ❤️👍
Any suggestion of one of these or another that you can import JavaScript not Python? Also for pattern recognition learning. Any help would be appreciated. Great video.
In the explanation of the Bayesian, I think the thing that you call it activation function is actually called "acquisition" function. But I'm not sure whether people call it also "activation function" or not!
Great video, you should also consider AWS Sagemaker which provides a great implementation of Bayesian Optimization for hyper-parameter tuning among other great tools.
This was great. I have a lot of stuff to look into!
I love h2o because I can use it in R or Python, and it has a similar api. PLUS! you can export the model and import it into one or the other. Pretty slick.
Hello, I work in evolutionary AutoML like TPot! Thanks for making this video
Cool video siraj. I also really like Uber's Ludwig because of its simple design and easiness to extend
Dear Siraj, I was to ask you about ml or ai with the project of neuralink (at least something or your saying about it). but here you said that the presentation was awkward. please explain what actually going on because I also found that people are not believing on neuralink.
Thank you and GOD may keep you safe and prosperous.
Nobody : what's the buisness
Automl : your data is the buisness
4:06
Bruh that's like saying "look how easy this machine learning algorithm is, I just have to write >import ML; lol"
Only when you're a java/JavaScript developer XD
@@utkarshupadhyay2359
*python
No colons in python mate!
@@utkarshupadhyay2359
No colons _necessary_
If you use ml on something with time you should try using time as an independent object instead of a global factor. Think of it making the data more like sonar instead of a picture.
Plz Make a vedio on training sports by ai( football,cricket etc)
At robocup, a competition which aims to make robots win a match against the fifa worldcup champions, the teams uses ml to improve their robots/agents. There are a lot of categories, from soccer simulation to humanoids robots
@walber can you show me where. Thanks
What is the difference between H2O AutoML and Driverless? Are any of these free?
Great video Siraj!
I LOVE ALL OF YOUR VIDEOS
Siraj bhai. The presentation of Elon Musk wasn't awkward. Relax! Apart from this for such an informative video once again. Thx
we are working on a auto ml too . called jelly monster .distributed machine learning and designing neural architect with no code .
Hi Siraj thanks for the awesome contents you offer very very valuable useful and inspirational. I have a quick question in a GAN can the generator be considered as a creative aspect(left brain ) and discriminators (right brain)
No.
Why? by it’s very essence being a random number generator can be similar to chaotic system or far from equilibrium state causing a sort of self organisation to emerge from within
@@MohanKumar-gj9th By "no" I mean not necessarily. Don't worry about what left brain or right brain does!! You understand what is GAN... that's great and enough.
I’m obsessed on GANs and how random functions allows order to emerge within
Hey Siraj can u make a video on 'sparse representation and dictionary learning algorithms"
i never thought of that interesting. Thanks for the idea!
PyTorch is Love 💜
PyTorch is Life
Fascinating.
What do you think about KNIME?
Cool video, I learnt a lot from it.. thanks.
First like it then watch it. #siraj
Could you make a video for adaptive instance normalization (AdaIN)? It would be very useful, nobody on TH-cam did this before
Singularitynet is Democratised decentralised AI ecosystem of ecosystems...
Ai in a mesh network for emergent Intelligence.
Partnered with Dominos, Ping An (10th Worlds Largest Company) and many more.
Look it up!!! this is the future of AI
Firefly algorithm is also one of the metaheuristic nature-inspired optimization algorithms.
Hey siraj do i need a degree to do freelance for any field of cs?? for eg. Coding, hacking, networking etc? thanks
No you do not. Software Development is not for everyone though. If you're chasing it for money then consider other options like sales, financials, etc...
no
Thanks a lot Siraj.
came from jabrils channel!
Amazing video!!!
I love Uber's Ludwig! you should do another video about it!!! great video as allways!
The starting words of Siraj were recited by Oppenheimer while testing atom bomb and is an excerpt from Bhagavad-Gita.
Siraj I would like AI to tackle the biggest scientific and theological questions in the world! Q1: Does God exist? Q2: Which religion is the most consistent within all the analyzed data? Q3: Can the answers to these 2 questions be statistically/probabilistically represented? Not sure how to break these logical problems down, so AI can solve them realistically! Do you think this is possible with AI's current state?
you are amazing man , thanks
Very valuable knowledge.
Awesome summarizion of topic.
Stranger things Kinda music has suited the video well .😊
Thank you as always 😊
Siraj. Is there a virtual AI version of a physical maker-space?
Hi siraj,pls make video for time series forecasting and best time series database..
It's a very discouraging video.
But for the lack of my competence I cannot fully comprehend how it will or may affect data-science positions.
Now I'm on my way to making data_science career or even just presenting it as my hobbie while working at McDonald's,
because I've found out that I like it and this study field makes me come back again and again every day, regardless that I had long time gaps in my learning journey. I spent quite a bit on learning and now don't want to give up on that.
And I don't know how to express that I'm so f*ck*ng concerned with what will happen next
1999 I spent the summer “coding” html making websites with angelfire.com.
2019 Im watching machine learning videos on TH-cam.
Realizing
I really missed the curve.
I am getting motion sickness between your galaxy warping and bopping up and down.
Are you watching in VR?
hm noted. i'll reduce that effect big time
Now you mentioned it I felt a tiny bit of it too.
I think that Elon's neuralink needs a Siraj digest of its own.
hope so lol
Hi Siraj! have you tried AI-based website > Midomi? < it recognizes what you sing or hum!
Go to sleep bro, its late
siraj stop playing with us..when is the album dropping? we tired of the teaser snippets. haha stay blessed.
From which region are you?
I realized my IQ was low asf when he said "shoutout to my 'bayes' squad."
Hey Siraj can u make a video on how to become a "bug bounty hunter
Please make video on autoDL too
please make a video on efficient Net ....
Thanks in advance.
Greatness
How can you install auto sklearn in Anaconda?
Previously you seemed to be a great fan of Elon Musk, but in your recent videos you kind of throwing negative comments on him. Why ? what changed your mind?
becoz Elon has said Automation will deprived jobs for people
Elon has been talking a lot and hasn't been delivering. Neuralink is a bit of a joke when he's(Elon) juggling so many unreachable goals. Ambition is great, but Elon has yet to achieve the promises he set out five/ ten years ago.
to be totally honest i'm a bit salty that he didnt respond to my very public request for an interview. Also, the AI doomsday talk is misplaced. It's not the statistical operations he should be fear mongering around, its the intents of the public corporations using them to exploit people
Hey can I join your school of ai as a masters degree
does anybody know how I can install MLBox in Anaconda MacOs
Doesn't all this make learning the background math largely pointless?
No it doesn't, ignore math and you 'll be ignored
How many body movements! And i'm italian lol
ciao
Great! Any week a way to simplyfy AI!
Its happening
Hi Brother can you please share the video about Robotics in python.
We are two physicists who did a podcast on how FaceApp works (beginner friendly). Check it out here: th-cam.com/video/Kx8S3fqPMoE/w-d-xo.html
Nobody:
AutoML: Yes
Sir I need small help how can I build a model to recognise the text written in different font ( ex calligraphy ) and display output as normal text. Can you guide me. Thanks in advance 😊
Siraj is definitely not satisfied with Elon musk's presentation!!
nice
Summary of the video guys = more automation from automation
Data scientists better take creativity classes or face basic income like the rest of us.
This should be called MLML or ML^2
Also the creative and else work of a data scientist will be done by an ai, so even much less to do for data scientistd
Okay
Is it me, or did this channel get smaller?
in what way? number of subscribers?
@@SirajRaval Subs, views. They both seem a little low considering the quality and subject of the videos.
I made a review about siraj on my channel!
Siraj: now i am become death
Me think, why waste time say lot word, when few word do trick?
Menelik Eyasu Kevin
bro please make video on multi label classification, its so important for my work
No love for Microsoft's AutoML too?
Have you seen/tried DataRobot? :)
yes but its a paid offering lol
First
can you create an app which is all voice from start to end !!!! Which includes voice input and voice output.
How can I use autoML in Arabic sentiment analysis
task?
Did anyone notice that the background acts like a gif though?
Rocket science in 1 week
haha "... by rapping"
I liked Elons presentaion