130. Bayesian Epistemology & Predictive Processing | THUNK

Подписаться 34 тыс.

Просмотров 14 тыс.

50% 1

Bayesian statistics - what are they? Do I want them? What do they have to do with epistemology, rationality, & prediction? Is it possible they’re happening inside me RIGHT NOW?
Links for the Curious
Arbital’s Guide to Bayes’s Rule - arbital.com/p/...
Bayes’s Theorem Illustrated (My Way) (komponisto, 2010) - lesswrong.com/l...
How Bayesian Inference Works (Rohrer, 2016) - • How Bayes Theorem works
Book Review: Surfing Uncertainty (Alexander, 2017) - slatestarcodex....
The Stanford Encyclopedia of Philosophy - Bayesian Epistemology (Talbott, 2008) - plato.stanford...
Bayesian Epistemology Course Notes & Readings (Stefánsson, 2016) - orristefansson....
Bayesian Statistics: Why and How (Dablander, 2015) - blog.efpsa.org/...
The New York Times’s review of “The Theory That Would Not Die,” a book about the history of Bayes’s Theorem (Paulos, 2011) - www.nytimes.com...
Perceiving as Predicting (Clark & his amazing shirt, 2013, for the 9th International Symposium of Cognition, Logic and Communication) - • Andy Clark - Perceivin...
Wikipedia on "Receptive Field," a primitive neural structure that shows exactly the sort of shape/border recognition posited in "Layer 2" (thanks to Elena for this one!) - en.wikipedia.o...
Predictive Processing And Perceptual Control (Alexander, 2017) - slatestarcodex....
Predictive processing simplified: The infotropic machine (Thornton, 2017) - www.ncbi.nlm.n...
Whatever next? Predictive brains, situated agents, and the future of cognitive science (Clark et al, 2017) - www.fil.ion.ucl...

Опубликовано:

2 окт 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 70

@henrymourland8379 3 года назад

Predictive Processing totally turns our understanding pedagogy on its head. We treat children like passive receptacles of information when they are in fact in the perpetual process of predicting and adjusting their subjective map of reality. Great video! Very exciting trend in cognition theory that I’ve been struggling to find accessible info on!

@spectralisation 3 года назад

This stuff should be taught at school everywhere. So much stupidity, especially in the political realm, could be avoided, if only people were less binary and less certain in their beliefs about reality.

@THUNKShow 3 года назад

I definitely think that a dollop of doubt would do us all some good!

@dunbar4 7 лет назад

Fantastic video Thunk! I think this may tip me over to being a patreon supporter. You have a fantastic ability of getting across the broad strokes with a beautifully charming simplicity. It would be awesome if these videos came with some sort of advanced modulus? Maybe in podcast form? May I also request a video on strange loops?

@THUNKShow 7 лет назад

Oh, thank you so much! I'm glad you enjoy the show! :D And "beautifully charming?" Shucks, this is for you: imgur.com/a/WiI59 I have recorded several podcast episodes with interesting & thoughtful people, but I'm a lazy ass & haven't edited them yet. TBH, life-stuff has been sapping my energy reserves for a while, so I'm less productive than I should be. :/ But yes! There will be some more in-depth stuff in the future! You certainly may request such a video! I kinda ditched IaaSL when he seemed to be recapitulating ideas from GEB in a less rigorous form (or so I imagined) - should I give it another shot?

@tomholroyd7519 9 месяцев назад

The excel joke was funny. So I'm like the only one left here now. Wow I left a comment 11 months ago and I'm re-watching it

@THUNKShow 9 месяцев назад

I am also, as always, still here. 😁

@nogusvelohistory8252 5 лет назад

Oh man, thank you so much. You save like 10 manhours of my time, I've been trying to walk over Surfiing Uncertainty book and you really saved the day. Also it's great to see your community and subscribers. Smart guys, finally!

@THUNKShow 5 лет назад

Always happy to boil stuff down as best as I can. :) The commenters are definitely a cut above - they really make the whole thing worth it.

@sandeepvk 7 лет назад

Bayesian theorem is the underpinning basis of Deep learning - AI etc

@thryce82 3 года назад

this is 100% not what I was looking for. ( I just wanted prior predictive distribution explained) Still nice video found it pretty interesting

@THUNKShow 3 года назад

Glad you enjoyed it regardless! :D

@Evan490BC Год назад

Too much handwavey talking for a very simple concept... Prior probability/belief, likelihood (a conditional probability). That's it!

@wardencobb7442 Год назад

I'm just gonna pretend I actually comprehend this until I truly grasp at least a portion of it... And no, it's not the (false) Denning-Kruger effect in action, that's just an artefact of statistics according to Kruger; it's just keeping my mind open to learning with positivity.

@grahamhenry9368 4 года назад

Deductive reasoning is not infallible either. All deductive conclusions are only conditionally true. Deductive conclusions are only true if the premises are true, and guess what? The premises are all derived via induction if you dig deep enough. Its induction all the way down.

@LeeCarlson Год назад

I do not doubt that Bayesian epistemology is a valid way of evaluating our environment. I also believe that the consensual environment in which we all live is a result of everyone developing the same expectations of how reality functions.

@scottboor1390 7 лет назад

Great video. to pick a nit - @ 3:10ish, It seems the more correct term would 'econometrics' rather than economics. It's econometrics which deals with the statistical models by economic relations.

@TheOhpapaNoah 3 года назад

Man you really helped me with my Bachelor thesis. I already had an understanding of PP but your explanation drove it home! Thank you so much for your awesome content.

@THUNKShow 3 года назад

Glad it helped! :D Congrats on your thesis!

@AdriandeSilva-rl3lg 8 месяцев назад

i thunk it, therefore i am .,,,,,,,

@ian9266 7 лет назад

Love the video. I would love to see an episode about General Semantics/The structural differential/orders of abstraction/e-prime/korzbyski/Weinberg

@adamrobison6737 2 года назад

This was awesome

@THUNKShow 2 года назад

Thanks!

@subramaniannk3364 4 года назад

I was relieved when you acknowledged the existence of Bayesian criticism at the end. I have a request, though. I was wondering if you could make a video of Maximum Likelihood method. Ronald Fisher is credited for developing the Maximum Likelihood method and he sure was an anti-Bayesian. Are there any philosophical difference between the two methods? It appears to me that Bayesian is more top-down analytic while ML is bottom-up synthetic.

@THUNKShow 4 года назад

I've added it to the docket, but I'll need to read more before I grok it fully enough to characterize it. It has some connections to another video I did recently on Occam's Razor & how it's actually a stand-in for some other heuristics, including likelihood - ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-HkgNFG-YLqM.html

@TheAdmiralBacon 7 лет назад

EXCEL IS ABSOLUTELY BEAUTIFUL

@nothingnew....9341 7 месяцев назад

Excel is beautiful

@amremam8937 6 лет назад

Can you provide a reference for the multi layer cognition? is it a generalization of Danian Kahneman system 1 & 2

@Arbmosal 6 лет назад

See the slatestarcodex links in the description

@junsangpark5013 3 года назад

amazingly precise information with the easiest explanation. Thank you for this knowledgeable video.

@THUNKShow 3 года назад

Thanks!

@veo_ 7 лет назад

I just realized... is that wooden ball on your set from Riven? I was the first person on the internet (well, in efnet #riven and the newsgroup, whatever it was) to solve the marble puzzle back in 1997 and spent the next 90 sleep depreived hours giving hints without spoiling the solution to a stumped internet. There was a good 12 hours where I was the only one who had finished the game. #nostalgia

@THUNKShow 7 лет назад

Holy crap, dude! Thanks from the bottom of my heart for doing the Lord's work...or should I say the Lord's whark? :D I was doing OK on my lonesome until I got hung up on the frog trap...didn't realize that I had to close the doors.../hangs head in shame

@aBigBadWolf 7 лет назад

Great video! Good sources. This is great. Do you have a patreon? I'd like to support you.

@THUNKShow 7 лет назад

I dooo! It's...I'm working on making it better for Patreon people, I promise. There is a rap about science & cognitive biases in the works. www.patreon.com/THUNKShow

@aBigBadWolf 7 лет назад

Cool, just keep in mind that I don't want to become a Patreon to see extra stuff. I just want to support this video series as I find you choice of topics very interesting and your presentation entertaining!

@THUNKShow 7 лет назад

Duly noted, & thank you. It's very much appreciated.

@rxixcievzncksoduf 7 лет назад

Dude, you are awesome - please keep making these videos!

@THUNKShow 7 лет назад

Thanks! :D

@tomholroyd7519 Год назад

Two thumbs up

@rishabbjain1363 3 года назад

sehr toll!!

@THUNKShow 3 года назад

Dankeschön! :D

@EmileSilvis 3 года назад

At the end of the video I was super convinced that his shirt was red at the beginning of the video. It was not. Well played Bayesian predictive engine, well played.

@GuyMichaely 5 лет назад

2:04 "You can actually find an equation that balances the infinite number of possibilities for the marble question" Do you have a link to where you found that equation/what it's called? Very interested in it, very counter intuitive!

@THUNKShow 5 лет назад

It's a personal spin on something I picked up somewhere - I think I adapted it from this paper: www.lirmm.fr/SIPTASchool08/Material/PRED-INF/PRED-INF.pdf

@waylonbarrett3456 2 года назад

If I'm not mistaken, that is basically LaPlace's rule of succession

@leiterlab 7 лет назад

Amazing video as always

@THUNKShow 7 лет назад

Thanks! :D

@Lenox414 3 года назад

What an amazing Video! Thank you so much :))

@jasonhamm3370 4 года назад

Not a fan of interpreting nulls :p

@THUNKShow 4 года назад

You and everyone else. ;)

@traich 7 лет назад

Didn't Popper eliminate the problem of induction ? PS: another great video! loved it

@THUNKShow 7 лет назад

I'm glad you enjoyed it! AFAIK, Popper's falsifiability criterion is certainly in the mix, but it was only really meant to vindicate scientific inquiry, & it's debatable if it does even that much. Personally, I find Bayesian epistemology more representative of how people learn thru induction - I don't usually go around looking to falsify premises, I just get more or less sure about them!

@traich 7 лет назад

I can appreciate the utility of Bayesian approaches although I am a frequentist myself. As to the problem of induction I was referring to the hypothesis from conjectures and refutations rather than the demarcation criterion however intertwined these are. The idea that given that our knowledge is a model a 'conjecture' about reality based on the evidence that is available, the criteria we have for choosing one hypothesis rather than another is its quality as explanation. A good explanation is one that is hard to vary and has deep reach. Some of our hypothesis make predictions about the future exactly because the explanation requires them to. If an explanation is hard to vary and explains all of the evidence that it is intended to without introducing redundancy, the question " How do you know that it will be true tomorrow? " , which is the problem of induction, becomes meaningless. I don't know that it is true today, as a matter of fact I know it can not be, because it is just an conjecture, a model, but by the virtue that it is the best explanation that we have and no one had yet refuted it I am justified in relying on its predictions about the future.

@THUNKShow 7 лет назад

Sure, but "refutation" is a tricky thing. How do you quantify when a model hasn't worked enough to count as being refuted? How many trials do you need? My main hangup with Popper is that it feels like he's trying to crowbar a floaty & inexact process into this framework of Truth - I sincerely appreciate the attempt, but it seems a little cramped. I could see it being isomorphic with BE with some tweaking, but the latter is more natural for me, & I like it for its built-in skepticism about True Things. ;)

@Fewkulele 7 лет назад

Excel IS beautiful

@tochoXK3 5 лет назад

Is mathematical reasoning the highest layer?

@saavestro2154 7 лет назад

Now you will have to make a video about the frequentist epistemology, is that even a thing? en.wikipedia.org/wiki/Frequentist_probability

@THUNKShow 7 лет назад

I feel like the Frequentist/Bayesian tribalism thing is a very weird philosophical line in the sand, even for nerds like myself who have strong opinions about metaphysics. :P