Integrated Information Theory - Part 1
Vložit
- čas přidán 27. 07. 2024
- Ever wonder why dogs and rocks are different? Math can help.
The Math of Conscious Systems. Integrated Information Theory (IIT) is a philophical, mathematical and scientific theory of consciousness.
This video was created as part of #SoME1 (www.3blue1brown.com/blog/some1 ).
More information about IIT:
- iep.utm.edu/int-info/
- integratedinformationtheory.org/
Main sources:
- Tononi, Giulio. 2004. “An Information Integration Theory of Consciousness.” BMC Neuroscience 5 (November): 42-42. doi.org/10.1186/1471-2202-5-42.
- Oizumi, Masafumi, Larissa Albantakis, and Giulio Tononi. 2014. “From the Phenomenology to the Mechanisms of Consciousness: Integrated Information Theory 3.0.” PLOS Computational Biology 10 (5): e1003588. doi.org/10.1371/journal.pcbi....
- Tononi, Giulio, and Christof Koch. 2015. “Consciousness: Here, There and Everywhere?” Philosophical Transactions of the Royal Society B: Biological Sciences 370 (1668): 20140167. doi.org/10.1098/rstb.2014.0167
---------------
Attributions
Dog in thumbnail: picture by Fernanda Nuso (unsplash.com/@fernandanuso ).
Icons from www.flaticon.com: "dog" made by Good Ware (www.flaticon.com/authors/good... ); "people" and "rock" made by Freepik (www.freepik.com ); "brain" by Smashicons (www.flaticon.com/authors/smas... ).
Music: Plain Loafer by Kevin MacLeod
Link: filmmusic.io/song/4223-plain-...
License: filmmusic.io/standard-license
Pictures of Giulio Tononi and Christof Koch are from Wikipedia.
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
![The moment we stopped understanding AI [AlexNet]](/img/tr.png)
CLARIFICATION: at 7:40 I say that the probability of current A being 1 is 6/8 because, of 8 possible states of the system, A is 1 in 6 of them. I should've phrased it differently to be more clear: among the 8 possible PAST states, 6 of them make CURRENT A equals 1. That's because A is the logic gate OR, so only the past states 000 and 100 would make current A = 0.
Finally! I was looking for an explanation for months! I already thought the mathematics of IIT may be a scam or something (okay maybe there are too many publications for it being a scam) Thank you!!!!
Bravo! Looking forward to parts 2 and 3.
What a great video. I remember looking for something like this when I first found out about IIT early in university. Looking forward to future parts!
Oh I've looked into this theory quite a while, I'm surprised someone put up a nice visualized explanation of it. Good work man.
Fantastic video, can't wait for part 2!
What an amazing video! It gives me the idea of what IIT is looking at.
It makes sense, but now I am scared that it can easily happen in silicon, and I am worried about turning off conscious computers!
Very well done! Hoping for more parts soon!
Please upload the rest of the videos. ❤️
Great video !
Great stuff! Thanks for simplifying such a potentially deep topic
(1 xor 1 xor 1) = 1 (not 0, as in the chart at 4:32).
Look at it as: (1 xor 1) xor 1.
(1 xor 1) = 0, and 0 xor 1 = 1.
For any number of inputs, xor'ing them all gives 1 iff there is an odd number of ones.
Is this a well established way of approaching this? I was wondering about it when it got mentioned, and it seems like both approaches have the potential to be valid -- your version, where we group into binary (in the 2-argument sense) operations, and apply the result binary-wise to the next argument, or doing... [does some research; discovers some terminology] the (exactly-)"one-hot" method, as is shown here. It seems both exist in the world, though yours is more common, at least within the realm of electronic logic gates. On the other had, it occurs to me that there's still a third possibility: "some but not all". I get to this from noting that A XOR B is sometimes thought of as:
(NOT (A AND B)) AND (A OR B)
... or, since I've got a bunch of parentheses anyway, please allow me to write that as n s-expression:
(and (not (and a b)) (or a b))
In this construction, we could add additional parameters easily:
(and (not (and a b c d)) (or a b c d))
But this would be 0 only if all inputs were 0 or all were 1, and true otherwise.
If there's a mathematical argument for using one of these over the others, I'd be keen to hear it. It currently seems to me that all 3 are valid, but chaining of 2-input gates is the most commonly used approach in practice.
Thanks for pointing that out! The truth is that I first wrote the script for 2 inputs, then I thought it would be more helpful to use 3 inputs instead, but I didn't notice the issue with that particular case. I guess I'm just used to think of it as "only one input as true". I agree that the way you explained is the most common way of thinking about XOR, even though, as David points out, it seems that both have been used. I would also be interested to know if there's a logical/mathematical argument as to why the common one is preferable.
Also, fortunately, either way it doesn't change anything for the examples, as the XOR gate C has only 2 inputs. :)
@@ideastranslated1119 indeed, I too "would also be interested to know if there's a logical/mathematical argument as to why the common one is preferable". I'm _guessing_ that comes from the EE and/or software worlds, rather than the mathematicians... but that's just a guess. We'll have to ask some mathematicians!
You're Brazilian right? I can spot the english brazilian accent in a second! I'm Portuguese BTW, and loved the video. I'm trying to get into IIT and this is much helpfull. When are you going to post a follow up? Um abraço do outro lado do Atlântico!
Yes, I am brazilian! I'm glad you enjoyed the video. Life is a bit hectic right now, so I'm not sure when I'm gonna post the follow up, but it's in the queue :) Abraço!
@@ideastranslated1119 Great! Looking forward to the follow up! Abraço!
Why is the standard definition of information via entropy -log2(P) not used?
Also, i imagine this theory's actual application to consciousness is fairly speculative (do our brains actually follow what is expected?) but ofc this is more of a question for science than mathematics.
I need to ask you this. Your accent makes me believe you are Brazilian. Is that the case?
haha yes, I am. I take it you are too, from your name. What specifically gives it away?
@@ideastranslated1119
Holy smokes this comment got very large very quickly, so here's the short part:
Some sounds are close, but not quite, and small things that sound not natural to a native. Making you sound (for a native) just a bit off, but it sounds exactly how Brazilians speaking English sounds to a Brazilian. (Very good, but with a noticeable accent)
Now, here's the full commentary:
It was mostly you not using the exact vowels and not quite the correct consonants.
Let me exemplify:
The word "apple" is pronounced like
/æpəw/.
The "æ" is a vowel that doesn't exist in Portuguese BUT it's very similar to the "é" as in "café". Again, it's very similar, but it's not the same.
And the "ə" is difficult to explain but it's called the "schwa" vowel. [For more information I suggest watching this video by Tom Scott: czcams.com/video/qu4zyRqILYM/video.html].
So you speak with the closest sounds that exist in Portuguese, which sounds similar, but aren't really exactly the most adequate sound.
I also notice that you are aspirating more than necessary. Like the k sound in "car" is not like the c in "carro". In "car" (in English) the c is a more aspirated k sound, usually abbreviated (in Phonetics) as /kʰ/. So you might have noticed that and decided to aspirate more, but it just sounds like you are exhausted and signing on every word.
However, the rhythm and proper stress of every word is mostly correct, so _parabéns pelo esforço_!
Wow, thank you for the in depth analysis of my accent! hehehe It's something I would like to improve, but I'm never really sure of how to do it. Recording the narration took me a lot longer than I'd like, as I had to keep repeating things. But at least it's good practice.
And thanks for the taking the time to watch and for the nice comment! :)
@@ideastranslated1119 Have a great day sir.
Sincerely, a Linguistics Nerd.
Hahaha Bingo!
E parabéns pela análise, @Bruno ! :)
1 xor 1 xor 1 = 1