Yohan J. John
@DrYohanJohn
Followers
7,238
Following
1,004
Media
3,434
Statuses
31,910
🧠 computational neuroscience | emergence | history of science 🤖 Research Assistant Professor at Boston University
Boston, MA
Joined March 2009
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Telegram
• 1480760 Tweets
Hezbollah
• 175749 Tweets
Lebanon
• 101834 Tweets
Mark Tuan Superpower Stage
• 78929 Tweets
#VALORANTChampions
• 75700 Tweets
嫌いボタン
• 73401 Tweets
SB19 RED CARPET PREMIERE
• 48596 Tweets
#SB19PagtatagDocuPremiereNight
• 48202 Tweets
Christoph Daum
• 28141 Tweets
BINI PLEASE PERFORM PANTROPIKO
• 27449 Tweets
新潟2歳
• 24453 Tweets
ビシエド
• 22644 Tweets
アルコス
• 19260 Tweets
キーンランドC
• 17924 Tweets
サトノレーヴ
• 16123 Tweets
モレイラ
• 13574 Tweets
キーンランドカップ
• 12195 Tweets
1Z BALIBAGAN SA RED CARPET
• 11398 Tweets
ナムラクレア
• 10941 Tweets
BEOMGYU FIRST PITCH
• 10754 Tweets
Pinned Tweet
I'm very happy to announce that our paper on "proxy failure" — analogues of Goodhart's Law ranging from molecular biology to brains to business to ecology — has been made available at Behavioral and Brain Sciences.
A thread. 🧵
.
10
55
227
Neuroscientists and psychiatrists really should read this.
31
238
946
Artificial Neural Networks tell us about brains in the same sense that cars tell us about horses.
25
58
490
Time to get over Kuhn and/or Popper: we have Nancy Cartwright, Manuel DeLanda and Hasok Chang doing far more sophisticated work with an eye on the intersection of theory and experiment.
(Which other philosophers of science would you add to this list?)
70
41
433
I just breezed through 'The Knowledge Machine' by Michael Stevens. Improbably, it's a pop science book about the philosophy of science. And it's quite good, even if you disagree with him.
Most radical is his "iron rule of explanation".
🧵
5
60
413
It's been well over a year, but I still haven't fully processed that I am co-author of a theory paper suggesting that "psychedelics are analogous to cognitive laxatives."
And this happened in part via a question I asked during a journal club!
21
35
346
Started reading this last night. DeLanda might just be my favorite contemporary philosopher.
So far it's almost like reading what my own philosophy of perception would be if I bothered to write it all down. Of course he's a much better writer than I am. 😅
16
20
300
The biggest problem with reductionism might be that it paints a wildly misleading picture of scientific history as well as current practice.
It concocts quaint fairy tales about how other sciences relate to physics, and even how macro scale physics relates to micro physics.
16
54
290
Perhaps it was inevitable. What Chomsky did to Skinner has been done to him.
There are serious flaws in this Piantadosi paper, but I suspect it is destined to be a classic — or at the very least, stimulate a lot of conversation.
Some highlights.
🧵
23
53
265
As far as I can tell, every scientific 'theory' of consciousness smuggles in the concept of experience while trying to explain the origin of experience in the first place.
You can catch this in words like 'illusion', 'hallucination', 'information', and 'representation'.
26
27
239
What if every attempt to theorize about the nature of cognition subtly changes it?
25
16
233
"one could go as far as to argue that the nature and function of consciousness is perhaps the single most contentious issue among the different schools of Indian philosophy, a development without parallel in the West, prior to Descartes, Kant, and the British empiricists."
8
31
220
The more one reads about chemistry, the less convincing reductionism becomes.
Chemistry has not been 'reduced' to physics in any sense other than the verbal.
25
26
217
Neuro-twitter! Do you recommend any textbooks on computational neuroscience? I am looking for good reading material that is accessible to an undergrad*. Preferably something involving models of synaptic plasticity.
* a talented undergrad who is capable of implementing ODEs.
22
33
213
Found a nice article on the origins of the normal distribution.
Not the importance of 'top down' idealizations in Gauss's proof — these are essential even for the most empirical and 'bottom-up' of concerns: error in data collection.
4
39
212
Why I am not in the 'Bayesian Brain' camp, exhibit A.
"...Bayesian models are significantly unconstrained, both because they are uninformed by a wide range of process-level data and because their assumptions about the environment are generally not grounded in [...] measurement."
8
33
198
Everyone interested in AI, neuroscience and/or consciousness should read this. I imagine it will infuriate the techno-whigs, just as Hubert Dreyfus did half a century ago.
7
58
194
A spike is an all-or-nothing event... only if you completely ignore the wacky stuff happening in the dendritic tree. :P
2
22
174
The neural representation of relations like this (which vary in subtle ways across languages) is a deep mystery that I want more people to think about.
We have decent models of pattern recognition...but how exactly does the brain mediate relations?
13
20
169
If your theory of consciousness makes no explicit contact with the world of the anesthesiologist, you are probably doing philosophy, not neuroscience.
:P
19
16
168
"Liking" and "wanting" are not the same. And dopamine is more involved in the latter than the former.
5
24
159
It seems very likely that similarities between real brains and artificial neural networks in performance and even feature-extraction have more to do with the shared data set and the presence of many free parameters than any structural or procedural commonality.
12
18
159
This is really cool.
'Heating and cooling are fundamentally asymmetric and evolve along distinct pathways'
2
41
156
For reasons that are unclear to me, people seem to struggle with the idea that artificial intelligence might be orthogonal to biological intelligence.
An AI system can arrive at humanlike performance through entirely unhumanlike means.
25
14
157
The Age of Metaphysical Clickbait.
13
29
154
Ted Chiang interviewed in FT about ChatGPT. Lucid, as usual.
4
39
150
One great thing about the David Graeber &
@davidwengrow
book 'The Dawn of Everything' is that it provides plenty of evidence that Europeans did not invent notions like freedom, democracy, or the public sphere. Versions of all these may well be as old as humanity itself.
7
14
143
Here is an analogy I've been chewing on lately: discrete concepts are like mountain peaks --- they are easy to identify & therefore name, but the surrounding regions that 'belong' to each peak may have vague borders. The unity of the mountain range maps to conceptual coherence.
8
15
138
Wow. They used eye movements to decode stages of mental rotation! Embodied cognition indeed.
4
27
133
Just attended a fantastic talk by Paul Cisek on the neural basis of decision-making. He beautifully linked evolution, behavior, task design and computational modeling.
Not sure if it was recorded but it looks like the content overlaps with this talk:
4
21
137
I am not one of those who thinks GPT and DALL-E are steps on the road to sentient machines, but I still find them amazing (and hilarious).
They are like some sort of statistical core dump of collective humanity. We should treat them like funhouse mirrors.
5
14
132
This is your regular reminder that there is no such thing as a 'canonical' cortical microcircuit.
There are broad similarities of course, but what V1 is up to is quite different from what, say, A32 is up do. Just look at stained tissue (something neuroscientists rarely do!).
7
18
130
Fascinating!
I am a bit embarrassed that I did not fully understand this when I was a physics student!
12
23
121
I have used "the map is not the territory" so many times that I'm tired of it. Not because it's wrong, but because it needs to be 'sublated'.
How do we know what the territory is? This isn't a rhetorical question. It has something to do with the relationship among maps.
19
9
123
Folks I hate to break it to you but the problem of confusing the map with the territory is far worse than any of us thought.
16
11
120
It is perfectly coherent to simultaneously be
1. Stunned and amazed by the capabilities of LLMs,
2. Worried about the social and economic consequences of automation, and
3. Skeptical that such systems can tell us about biological phenomena (beyond summarizing research).
11
16
110
The importance of inference to perception, from al-Haytham's Book of Optics (1011-1021 CE).
5
25
108
It sometimes strikes me that "affordance" is used like a magic word in psychology and neuroscience.
How exactly does an animal recognize that something is an affordance? And at an even more basic level, that it is one thing and not several?
25
6
108
Now is a good time to be a "splitter" rather than a "lumper" when it comes to concepts.
Performance is not understanding.
Models are not theories.
Engineering is not science.
Usefulness is not truth.
10
11
103
What would you put on a literary reading list for the cognitive and neural sciences?
I'll start with something out of left field: 'The Quantity Theory of Insanity' by Will Self. Hilarious, and a great a parody of academia. The whole collection of stories is excellent.
🧵
12
19
108
I get why we do this, but referring to brain regions as if they are singular — "the" amygdala, "the" hippocampus etc — creates a strange neglect of the fact that there are two: one in each hemisphere.
Do they cooperate and/or compete? Divide up labor? The literature is spotty.
7
8
102
I know the pandemic has been a disasteroid, but because of it, a wonderful ongoing collaboration emerged unexpectedly.
Our review of dynamical systems thinking for fMRI, with
@kslays
, Karthik Srinivasan,
@eli_j_muller
,
@DrBMunn
&
@jmacshine
is out now:
9
22
100
Just got access to DALL-E 2. Prepare for nonsense.
“tiger in a fashionable suit playing pool in 1987 vaporwave”
8
4
98
Scientists, how much philosophy of science do you read? Has it ever changed how you think about a phenomenon?
Which thinkers do you like other than Kuhn and Popper (the zombie defaults)?
33
14
96
"Finally, we challenge the classic view that functional modules are contained within specific brain regions; instead, we propose that certain regions prioritize specific types of information over others ...
2
19
97
Close the loop: language is just a byproduct of biology. 😛
6
9
97
"The modeller’s quest to pin the organism down and preserve it in aspic destroys the very thing that makes it living."
Subtle critique of the Free Energy Principle and other forms of "invariantist" modeling.
5
8
96
When I come across the term 'embodied cognition' these days, I wonder: as opposed to what?
If you work out how anything works in sufficient detail, you will have to deal with embodiment eventually. And that includes abstract math and logic.
23
4
93
"If an event results from the joint operation of a number of causative chains, and if these causes “interact” [...], it becomes conceptually impossible to assign quantitative values to the causes of that individual event."
Levins & Lewontin, 'The Dialectical Biologist' Ch 4
7
13
89
Do you think of "information" as something that exists "out there" in the world, independent of the properties of the observer?
60
13
94
The tldr:
"Modern machine learning has subverted and bypassed the entire theoretical framework of Chomsky’s approach, including its core claims to particular insights, principles, structures, and processes."
2/n
2
16
93
Phase transitions are a great example of emergence.
And the existence of such sharps transitions in theoretical models shows that emergence is not necessarily a mark of ignorance about fine-grained details.
Contra the biology maxim, natura facit saltus.
Just as water turns to steam, infinite graphs will suddenly form infinite clusters of connected nodes, undergoing a phase transition. A new proof shows that where such transitions take place can be determined by looking at only part of a graph.
6
84
275
5
15
89
One thing I'm not entirely clear on is why so many philosophers adopt something like the epistemology of science. Why not stick one's neck out beyond the limits of science? Does science need what is effectively a theology wing?
32
5
86
Reasoning that the universe is deterministic because a successful model of it is is analogous to arguing that the world is flat because your ruler is.
13
14
90
"So why do animals keep evolving into crab-like forms? Scientists don't know for sure, but they have lots of ideas."
The crab-form is an attractor in the evo-devo landscape. 🤓
10
17
88
"The trouble is that too many people know more statistics than they understand."
3
10
91
"Is there evidence for the claim that humans have a Bayesian brain in the realist sense? No direct evidence has been presented to date. "
Rahnev (2019).
7
11
90
The more I read the word 'information' in neuroscience, the less clear I am on what the word means.
16
6
88
Googling "brainwaves" is a triggering experience for any neuroscientist. How did all this gibberish become ubiquitous?
10
9
88
Today I was introduced to Dempster-Shafer theory, and the broader idea of generalizing probability beyond single events to sets.
This book on the topic seems quite interesting.
5
11
89
From a preface to Alfred Korzybski's 'Science and Sanity'.
In the 1930s Korzybski coined a saying that has become a mantra for me: "the map is not the territory". Apparently the line was inspired by mathematician Eric Temple Bell, who said: "the map is not the thing mapped".
6
18
85
Manuel DeLanda's deceptively simple definition of emergence:
"a property of a whole is said to be emergent if it is produced by causal interactions among its component parts".
6
15
84
I find this conversation deeply strange, but the fault likes with scientists who keep repeating the nonsense that perception is a 'hallucination'.
A more neutral word would be 'construction'. And most of the time our constructions are pretty accurate.
8
12
83
"Ordinary words convey only what we know already; it is from metaphor that we can best get a hold of something fresh."
- Aristotle
10
16
81
My latest post is about the invariance perspective, and how it offers us a more accurate picture of objectivity than the phrase "mind-independent reality".
I also describe why computational neuroscience is a "second-order quest for invariants".
10
13
86
I have spend the last few weeks diving into "higher math" in a (thus far) quixotic quest for ideas about how higher cognition might work in the brain.
If anything, I now find higher cognition even more baffling than I did before. 😅
🧵
10
5
82
I hate to break it to you, process metaphysicians, but processes are just objects in a more complex representation-space. :P
(Which is one way to think about why it is so easy to noun verbing and verbify nouns.)
30
7
81
The strange thing about probabilistic reasoning is how much it depends on one's imagination: asking "what are the chances of this event/hypothesis?" requires constructing a representation of what *can* happen.
This is called the 'reference class problem' and it's a doozy!
...
4
11
83
Time and again, we find that some neural phenomenon that initially seems to have one function also contributes to the 'opposite' function.
The zero-order response is to be skeptical of attributing any function in the first place. But perhaps there is a dialectical solution...
5
14
82
"What appears to be intelligence in LLMs may in fact be a mirror that reflects the intelligence of the interviewer, a remarkable twist that could be considered a reverse Turing test."
Terrence Sejnowski
2
16
82
"The reason cancer’s not a problem in today’s robotics is because we build with dumb parts.[...] Whereas in biology, it’s a multi-scale architecture where every level has goals, and there’s pros and cons to this."
...
2
16
85
Texture is one of the most fascinating aspects of perception, bu I rarely come across popular treatments or philosophical discussion of it.
Do let me know of excerpts if you are aware!
9
15
82
'Objective' is not the opposite of 'subjective', but a subset of it.
15
12
80
Philosophers interested in neuroscience! Is there a good book or review paper you recommend on 'neuro-adjacent' philosophy? I'm thinking of topics like consciousness, agency, qualia, etc.
An overview of analytic, continental and non-western approaches would be nice.
25
9
81
The great Russian vision scientist Alfred Yarbus on why motion is required for vision.
Seeing is a lot more like touch that you might realize. To see, we must caress the world with our eyes.
2
14
81
A paper by philosopher Alan Hájek called 'The reference class problem is your problem too' helped a lot. I distilled what I learned into this essay:
2
12
81
Last night I read Borges's short story 'The Theologians'. Just brilliant! Of all the writers I read, Borges is the one I'd most like to emulate.
The density of detail and allusion in his prose, combined with what might be called 'theoretical rigor', is something to behold.
12
6
81
At this point I think words like 'entropy' and 'complexity' have jumped the shark — in neuroscience at least.
We are in the era of measures in search of something to measure. :P
7
7
79
"Depending on how you arrange carbon atoms, you will end up with something soft and opaque or hard and transparent. Clearly the properties of the substances are not to be found in the properties of the atoms." (From my post above.)
5
9
79
This has been brewing for a while. I decided to explain why I think the 'anti-representation' position is confused. 😅
This post focuses on why 'placeholders' aren't a bad thing, and also on the history of the concept of neural representation:
15
20
80
If I said to you that thinking is little more than virtual movement, would you find the idea controversial or strange?
33
2
77
"The problem with today’s AI, he says, is that it tries to imitate the results of brain processing instead of probing the mechanisms that give rise to the results."
Yup.
9
8
77
How do you understand 'computation'?
What criteria do you use to decide a particular biological phenomenon is *not* computation?
Do trees compute 'degree-days' etc.?
(If we say the planets 'compute' their orbits, then everything, and therefore nothing, is computation.)
29
10
77
The proofs are over my head, but I still really love this paper. It also provides a nice metaphor for understanding emergence in a system that is completely "transparent" (since it is an abstract mathematical phenomenon).
1/n
4
11
77
Not a bad intro to Grossberg's way of thinking about attention and top-down processing.
And this is from 1995, years before "active inference" and "predictive processing" became popular.
5
8
78
Sometimes I think radical claims about causality arise from confusing a *spatialized* diagram with a *temporal sequence*.
Phenomena that are part of a spatial/structural loop are sometimes described as enabling "circular causality", but I think this is just regular causality.
7
10
77
"Yet the concept of “particle” falls apart on closer inspection."
I think all concepts fall apart on close inspection.
10
14
76
Just finished this book yesterday. Really excellent! And timely.
It dovetails with our proxy failure (Goodhart's Law) paper too: sometimes aiming for a thing via proxies (quantities, scores, metrics) can lead to the opposite. Here the problem is aiming for control of life.
...
5
12
77
My default position for years has been that neuroscience is too immature to ground self-help advice.
Do any of you see value in framing behavioral/psychological recommendations using neuro-talk?
I was discussing this with
@kslays
and other friends recently.
(...)
21
5
74
I confess I've never really understood the appeal, or even the meaning, of the question "why is there someone rather than nothing?"
33
12
76
Very nice article on emergence by physicist Barbara Drossel.
Not perfect, but a handy resource if you find yourself in an argument with someone who thinks physics and reductionism are synonymous.
3
13
75
"Our analysis supports a top–down structure in quantum mechanics according to which higher-order correlations can always determine lower-order ones, but not vice versa."
3
16
78
David Krakauer, in this collection:
"The idea of emergence can be very simply described as the successful factoring out of the fundamental laws of physics without any loss of explanatory power."
Part of a very cool theme issue ‘Making and breaking symmetries in mind and life’ organized by
@adamsafron
,
@adeelrazi
, Zara Sheikhbahaee,
@DaltonSakthi
and Magnus Bein
with our (actually, MidJourney's) cover:
6
29
124
2
13
78
One reason for non-mathematicians to take a look at category theory is to get over the notion that set theory is the only way to view a formal symbolic system. Instead of isolated elements floating in a void, cat theory invites us to foreground relationships, i.e., morphisms.
4
8
76