01:32 Inverse probability (aka Statistics) 03:28 James-Stein estimation 04:00 a "wave-particle duality" in Statistics ... 04:16 frequentists vs. Bayesians 04:32 Are you a Bayesian or a frequentist (Jordan video th-cam.com/video/HUAE26lNDuE/w-d-xo.html) 07:30 Empirical Bayes 08:27 False discovery rate
Loving that you ask for basic definitions to top tier guests. As Elon Musk says, it is important to understand the most basic/fundamental concepts first and this is a great way to do so. Thanks Lex. Please don't stop doing it.
I've seen Sal Khan say in a TV interview that, while he personally loves Calculus and has recorded many videos about it, he thinks that given their relative usefulness in life and society, it may be overemphasised in the high school curriculum compared to Statistics. He suggested that it might be more important for high school students to acquire a basic literacy in Statistics over Calculus, and I would agree. Many public debates in politics, economics, and science are severely hampered by the fact that the public is utterly uneducated on even basic Statistics. It's hard to have evidence-based policy discussions and election campaigns if the electorate largely cannot follow the evidence.
Statstics is a topic of math that can be defined broadly having two subtopic : descriptive statitics and inferential statistics. Descriptive stats can describe characteritics and relationships of entire data sets (populations) and samples of datasets. Examples measurements from descriptive statistics are : means, median, variance, and standard deviation. Business use inferential statitics for doing research on potential products for customers by doing surveys and doing calculations on data theyve gathered. Inferential statitics can create measurements of likelihood and mathematical confidence that some outcome or characteristic will turn out to occur/be. Science uses inferential statitics when, for example, they want to explore new medicines that they think will/will not be successful given past data they have gathered on experiments. AI research relys heavily on statistical math because we are programming the fomputer to make predictions.
Also, I think I may have found the paper that he mentions on the subject. Stephen Stigler, 1990, A Galtonian Perspective on Shrinkage Estimators projecteuclid.org/euclid.ss/1177012274
I am not are you aware of it but brother you are doing great things for society by sharing theses insights and knowledge to wider auidance. thank you I wish to meet you one day
Great great great discussion!! Risk, decision making, market forces, and externalities. Wow! There is such a major disconnect between university studies, expectations and real world demand for skills right now
Love you're work lex, ever thought of interviewing just random citizens, via Skype to see what the everyday person struggles with, and our views of things?
As an App Developer I'd lean towards the frequency statistic approach over the Bayesian data variations would seem to reflect more true or truthy results to me. At least if there were statistical models or something similar in my application. I've only had limited exposure to the pure math of statistics and when it got advanced I'd be reminded of why I chose coding programs over finance or move on to an MBA. From an AI perspective which of these options will prove more reliable when done by machine brains?
oh, I didn't know I want to be like this Mike. I mean, come on - Michael Jordan; did Lex ask him in the full podcast about his name and students/peers making fun of it? really, MJ
@@rlee4516 I prolly know more popper than you the product of reasoning is deductive. you're right but the premises upon which we reason. or the ground facts we a apply reason to are always inductively gotten (prior to reasoning)
![[#2024MAMA] ROSÉ (로제), Bruno Mars - APT. | Mnet 241122 방송](http://i.ytimg.com/vi/dgGqD28J6aQ/mqdefault.jpg)
01:32 Inverse probability (aka Statistics)
03:28 James-Stein estimation
04:00 a "wave-particle duality" in Statistics ...
04:16 frequentists vs. Bayesians
04:32 Are you a Bayesian or a frequentist (Jordan video th-cam.com/video/HUAE26lNDuE/w-d-xo.html)
07:30 Empirical Bayes
08:27 False discovery rate
This guy is Michael Jordan of machine learning.
I was gonna say underrated comment, but you're actually the leading comment now. I guess it's early.
Lol “Michael (A.I.) Jordan”
Yeah his name is Michael Inteligence Jordan
And Andrew Ng is...Lebron James?
Lex, great question! This kinds of "childlike" questions are the best, which people are too often afraid to ask. Great video, good guest!
fundamental questions is what you mean :() yeah most of them lead to philosophy
exactly
Loving that you ask for basic definitions to top tier guests. As Elon Musk says, it is important to understand the most basic/fundamental concepts first and this is a great way to do so. Thanks Lex. Please don't stop doing it.
Statistics, as a discipline, gets so little attention across all media. It's really refreshing for something like this to pop on my feed.
I've seen Sal Khan say in a TV interview that, while he personally loves Calculus and has recorded many videos about it, he thinks that given their relative usefulness in life and society, it may be overemphasised in the high school curriculum compared to Statistics. He suggested that it might be more important for high school students to acquire a basic literacy in Statistics over Calculus, and I would agree. Many public debates in politics, economics, and science are severely hampered by the fact that the public is utterly uneducated on even basic Statistics. It's hard to have evidence-based policy discussions and election campaigns if the electorate largely cannot follow the evidence.
Inductive statistics shouldn't get attention since it's entirely fallacious.
Statstics is a topic of math that can be defined broadly having two subtopic : descriptive statitics and inferential statistics. Descriptive stats can describe characteritics and relationships of entire data sets (populations) and samples of datasets. Examples measurements from descriptive statistics are : means, median, variance, and standard deviation. Business use inferential statitics for doing research on potential products for customers by doing surveys and doing calculations on data theyve gathered. Inferential statitics can create measurements of likelihood and mathematical confidence that some outcome or characteristic will turn out to occur/be. Science uses inferential statitics when, for example, they want to explore new medicines that they think will/will not be successful given past data they have gathered on experiments. AI research relys heavily on statistical math because we are programming the fomputer to make predictions.
pebre79 isn’t inferential statistics basically the basis for AI?
smartest person I've ever attempted to keep up with
appreciate the wikipedia popup for James-Stein estimator, nice editing work
Also, I think I may have found the paper that he mentions on the subject. Stephen Stigler, 1990, A Galtonian Perspective on Shrinkage Estimators projecteuclid.org/euclid.ss/1177012274
Lex, great interview!
I am not are you aware of it but brother you are doing great things for society by sharing theses insights and knowledge to wider auidance.
thank you
I wish to meet you one day
Great great great discussion!! Risk, decision making, market forces, and externalities. Wow! There is such a major disconnect between university studies, expectations and real world demand for skills right now
Love you're work lex, ever thought of interviewing just random citizens, via Skype to see what the everyday person struggles with, and our views of things?
This guy has a great Anthony Bourdain style voice
Interesting ! Thank you Lex ! You are an interesting person. With a fascinating drive. Really amazing to follow your stuff. Thanksssss
In 9:38 should be "the classical bayesian"
Superb!
Many AI young turks should imbibe into the mind of Jordan a true intellectual to train your GPT
did he answer the question? my brain is not advanced enough to know
Lex Fridman is basically a better JRE
As an App Developer I'd lean towards the frequency statistic approach over the Bayesian data variations would seem to reflect more true or truthy results to me. At least if there were statistical models or something similar in my application. I've only had limited exposure to the pure math of statistics and when it got advanced I'd be reminded of why I chose coding programs over finance or move on to an MBA. From an AI perspective which of these options will prove more reliable when done by machine brains?
Mark Twain said "There's lies, Damn Lies, and statistics"
oh, I didn't know I want to be like this Mike. I mean, come on - Michael Jordan; did Lex ask him in the full podcast about his name and students/peers making fun of it? really, MJ
I liked the thumbnail
I would pull the Amazon App and say Alexa: ... and ask you to repeat the question. C'mon. Ask questions that someone can't Google.
I feel like I am a rock.
What is the image from the thumbnail
3:39 Can someone please help me find the paper?
He may have been referring to this: projecteuclid.org/download/pdf_1/euclid.ss/1177012274
I find that math isn't interesting unless you know what problem it's trying to solve. Here it's explained masterfully.
For me, math is beautiful and interesting in and of itself. It need not solve any practical problem for me to appreciate it's beauty.
Divine Masculine energy
They should retire Michael Jordan's name so it doesn't confuse people
All inductive reasoning is fallacious.
All reasoning is Inductive
@@goyonman9655 Nope. Reasoning is never inductive and it's always deductive. You should read Karl Popper who shows that induction is a myth.
@@rlee4516
I prolly know more popper than you
the product of reasoning is deductive. you're right
but the premises upon which we reason. or the ground facts we a apply reason to are always inductively gotten (prior to reasoning)
My autism makes me mad when i hear you normies freak out over stuff like this that ive been talking about for years.
Shut up bro
This guy is Michael Jordan of machine learning.