Lisa Feldman Barrett: How the Brain Creates Emotions | MIT Artificial General Intelligence (AGI)
HTML-код
- Опубликовано: 7 июн 2024
- This is a guest talk part of MIT course 6.S099: Artificial General Intelligence. This class is free and open to everyone. Our goal is to take an engineering approach to exploring possible paths toward building human-level intelligence for a better world.
INFO:
Course website: agi.mit.edu
AI podcast: lexfridman.com/ai
CONNECT:
- If you enjoyed this video, please subscribe to this channel.
- AI Podcast: lexfridman.com/ai/
- Show your support: / lexfridman
- LinkedIn: / lexfridman
- Twitter: / lexfridman
- Facebook: / lexfridman
- Instagram: / lexfridman
- Slack: deep-mit-slack.herokuapp.com 
Наука
I would love to see Lisa Fieldman Barrett and Paul Ekman discuss their respective theories with one another
Yes!
I wonder what on earth Ekman would have to say
This is an evocative interview. Not only for the base engineering problem of how to go about designing and executing AGI, but also for how to structure an AGI's relationship to the rest of us. Lisa speaks with a brave authority. I'd be hard pressed to call her on anything she said, but it is rare to see someone state facts on the brain and neuro-chemical systems so directly . The lecture is definitely worth your time.
Yes.. and such behaviour flagged anything she said with such confidence as "might not be true", at least in my opinion. We can not see what is the source of the info.
Hard pressed to cal her on anything? Really? hagioptasia.wordpress.com/2020/10/12/the-theory-of-constructed-emotion-is-bollocks/
@@davidtracey4457 quite shouty blog but I'm not sure the writer has read LFB's book, How Emotions Are Made.
@@charlieetal1 Why not?
@@davidtracey4457 the first argument already tells me you do not interpret her the same way I do. I might read the rest later but now actually i shouldnt be doing this :) Anyway, my point: "But how do we all have sorrow, joy etc?" Well, we dont, japanese dont have sorrow the way you do and even though we can translate the world Kanashimi it doesn't have the same implications"
you know what, LEX - you inspire me. I'm a phukcing addict, usually wasting my time (except watching Lex). I just have imagined what a great amount of a hard work you had had put to become a Lecturer at MIT. What a true values you have!!! I'm excited to get off the bad habits and go hard af into the best self possible.
Lisa has demonstrated good way of listening and be present with real thinking. That is why when she disagrees, it is so logic and getting to the point. 👏
Listening to Lisa Feldman was absolutely mesmerising. What an incredible talk!
A startlingly good talk. It shows that building the engine that could let an AI learn these things will be the vitally important first step, and that trying to hard-wire emotions will only lead to emulation, not true emotional states. Food for thought indeed.
So happy to witness dispelling of so many myths about how brains work.
Dr. Barrett: I'm buying your book due to your appearance at this event. Thank you for attending. It was fascinating to listen to your expertise.
Superb choice of experts, Lex. Thanks.
Quite an interesting talk, thank you for sharing this.
lol "Hi, thanks for talking here. You're saying cool stuff." Lex - "That's not a question."
Emotions are not a basic element, they are emergent. Brilliant!
Actually pretty obvious.
This is wonderful inputs to AI “nerds” group.
Definite difference between carbon based systems versus silicon based system which do not die!
OMG the preambles on these questions
Wow..we all need to read this book.
To set our thinking straight on many important fundamental trajectories.
I will watch this bid agaim.
Great person, subject, and interview! Thanks!
Enlightening. Especially, I feel the need to adjust to the cultural value after moving to a new culture. Also, it's funny to see the clash between two different perspectives and personalities.
this woman is just amazin .. thaaanks
@21:23 begins a misunderstanding and great analogy for how ideas and labels around emotions are added by the brain, in each culture, rather than being intrinsic or hard-wired. It is the same as how a culture imbues dollar bills, or gold, or numbers (bitcoin) with monetary value. What people believe matters deeply, and varies widely.
Very good talk. Learned a lot.
I enjoyed this conversation.
This was amazing!
Brilliant!
This talk is worth taking notes over to ponder for later. Is there any place I can get a transcript of the talk?
Outstandingly Informative!---Expanded my Awareness!---Real!
She would be a great teacher for me. Very intelligent and she will challenge you. Thank you Lex and Lisa for the excellent discussing.
Wow ! I'm back ref from a podcast vid. what an insight !!
there is such a gap between the two subjects that i believe the discussion and her forwardness was warranted. she says interesting things, but not so applicable to ML at the moment. maybe a couple of years down the line psychology will be important to look at in relation to AGI. its just too abstract for current approaches
23:59 distinguish physical reality and social reality
44:27 social animal - regulate each other's nerve system
Incredibly powerful information on what makes us social animals
Great Talk! :)
I learned more about how the mind works here than anywhere else.
Brilliant lecture. Thank you, Lisa and Lex! Some of the students have an overly computer oriented viewpoint. I feel like folks here have been encouraged to be assertive, but when in the presence of a true expert, it is better to be less so.
there is nothing more true or less about something where we have no deterministic benchmarks. Both of them are modelling the brain in different ways.
At 39:49 minutes into it she hits the nail square on the head when it comes to unlocking what I also believe is the secret to accelerating AI development without having to sit in front of a computer display all day. If you want machines to be like people it has to be taught in the form of a person it needs the most human body possible regardless of what it's being programmed to do. A robotic AI AlphaGo that looked and acted like Marion Michael Morrison would have been far more insulting to China when it defeated Ke Jie back in 2016, especially if it showed up to the match in a self-driving 1959 convertible Cadillac Coupe Deville with bullhorns on the front hood.
First of all there are some certainly interesting concepts/ideas in this talk. Like the one where we are one big brain and we influence each other in a myriad of ways, or that there is no unique mapping from the emotional state to facial expression (though babies seem to always make laughing expression when they are happy and similar for other affects/emotions and they are still not following any social norms or am I wrong here?). Also do dogs have somewhat higher EQ (emotional intelligence) and is that a consequence of human/artificial selection? Did brain in animals in general evolve, to a point where it is today, as a consequence of trying to manage all the "subsystems" or was that just a side thing where the goal was actually to survive, this one seems obvious at least to me.
It's hard though because I am unsure of the meaning she attributes to certain words like the distinction she makes between affects and emotions, same thing as with math, we first have to agree on what words mean before we starting sharing ideas otherwise our conversation may turn into argue/futile, which we saw a couple of times in this talk haha.
But many of these topics don't seem to be provable at all (at least for now) and people end up talking about opinions in a scientific manner rather than talking science/talking facts. Too much philosophy that leads nowhere and too little mathematical/analytic/engineering mindset for my taste. How can we harness knowledge from this talk, rather than talking knowledge to sound knowledgeable and smart? I think we can predict emotions from facial expressions by also using past facial expressions (that's why tech companies invest money IMO). Yes, sure, if you only focus on a single facial expression it's just a facial muscles movement and you are unsure of the underlying emotion but if you track history it's another story. Also it's possible to fool this AI system if you posses somewhat higher EQ, but you can also fool humans using that so I don't think it's of practical concern for now.
i didn't really agree on anything she said as her opinions contradicted all video evidence i've seen about how babies react and learn. and I've also seen the videos of tribes that have never seen a white man before so emotion is universal. not learned.
One person asking question definetly sounds like a robot from future
Amazing talk! Thoughts as constraints over our sensory arrays which work towards maintaining allostasis. Our environment and other people affect this array and produce real effects on the body
Curiosity is the most important "emotion" for an AI, and the first one I added into my AI. It is also relatively easy to quantifying as well.
ok, but it is important not to be interested in everything, equally. How are you going to define what matters for your AI?
Fantastic talk and I didn't feel the rudeness that others point to. However am I wrong in thinking that she is confusing allostasis with homeostasis 41:38 Allostasis is a tool to achieve homeostasis, but she refers to allostasis as an end in itself. Or am I missing something?
@ 21:52, it sounds like emotions are "Transcribed".
Which makes sense, I lose a lot of connections but understand metaphor really well. When connecting my nausea with hunger, I have to look at the shakes I have at the moment or my sloppy writing. Actually, my writing looks fantastic when I have a well-balanced, consistent, diet; which is never.
Awsm discussion..! just downloaded her book and can't wait to read.. Thanks Lex..! Pls continue the good work.!
She is suchh a Brilliant Staar!!! 😊
So she's saying that emotions are not an innate property of the brain but rather a function of outside perturbation and are effectively learnt through social structures. Furthermore that emotions may function as a mechanism to return our bodies back to allostasis.~? That's what I came out with.
According to what she talked, will robots have emotions in future ?
People will never break out of dogmatic thinking. We will always stand behind and defend our views as if it’s a matter of life and death.
That's a loooong tie :)
Extremely interesting talk by the way
I think it's supposed to be a return to homeostasis. According to the dictionary allostasis is the process by which the body returns to homeostasis.
This was a brilliant lecture
"Allostatic regulation reflects, at least partly, cephalic involvement in primary regulatory events, in that it is anticipatory to systemic physiological regulation (Sterling & Eyer 1988; Schulkin 2003)."-
en.wikipedia.org/wiki/Allostasis
We share the work of allostatic regulation to survive.
This lecture is amazing! It truly requires an individual to think about how human brains operate.
I read the book and I found it very provocative, but I thought in the talk, she seemed to give credence to the representative theory of language and facial recognition, when there are dedicated innate systems for both of these that are acutated in development, and it is not a simple building up of elements, whether in faces or language. I don't think this is what she is saying, having read the book, but it seemed to come across that way to me.
Focussing solely on the organic brain when thinking about brain functionality is a big mistake imho. I haven't seen all interviews from the series, but so far none have touched the intersection with what I like to call the outer brain (which is the culture we live in and what our brains need to adjust to just like to every other aspect of the physical reality to succeed). I make this point because many comments here and questions there seem to not account for the difference between affects and emotions Lisa explained so well. Affects (feelings) are internal while emotions (cognitions) are an affect caused by match or mismatch with external conventions. Or put differently, the brain is hardware running culture as an os with apps installed via learning, processing reality as data, all together allowing for experience. This way, I like to think of the brain more as an antenna, an agent or a client, not a real entity. Like your computer does nothing without you doing things to it.
She knows her stuff.
Not at all. She is so confused. People who know their stuff are able to explain it with simple language, not socially coercive psychological posturing. She is obviously a deluded postmodernist. Almost nothing she said is useful scientifically.
That was profound.
I am currently reading the Molecule of More by Dr Barrett and very much enjoying it. That said - I am less impressed when she talks. I think this is because she sticks to the facts in her book while she voices her informed opinion when she speaks.
@ 40:18 Each muscle stops another, the brain too.
"The world's a paradox, it's there inside of all of us." - Iain McGilChrist.
I think meditation has that key element: To stop space and time.
Our muscles, our brain, are all made to calm a limbic creature in a world of snakes, strings, mem, and nun.
I like to think of empty space as the most chaotic. Each particle is so distraught with itself, and most completely to the point they've imploded on themselves; space, as I say, is like a giant ball pit, waiting to be waded.
In this concept, even space, empty space, has a fundamental property of movement, even if placid, the lake can move, and objects move in space. The space itself is like § overlapping imploded particles, but each one vibrates at such chaotic speeds, they become calm too. Just the other way, you know?
~ I think it would be easier to study strings if we could isolate my, imaginary, hypothetical imploded particle. Huh, I wonder if this chaotic struct is what builds anti-matter; if these hypothetical chaotic spheres are rubbing through each other?
Funny how I got all this from that.
Anyway, yeah, muscles, brain, stops spacetime chaos. Mem Nun.
40:11 Epic!
Brilliant - learned alot.
> @ 19:43 That was a great breakdown.
Eyes and motion make emotion
My art disrupts the Child As Scientist
I think it needs eyes before a body
@ 9:31, it is my understanding that these connections were foretold in astrology.
For instance,
I love the ancestors of chickens. Brittany Spears is a great example of how they extrapolated this information. Not that every chicken has to conform to their ancestors' method of chicken, or what some people may do, especially living on chicken farms, is a "pecking order"; not joking.
Anyway, I think they were attempting this type of inference with the only Prime Time Television they had: Jorgunmandr, or celestial water, Tiamat, I think dragons are common when talking about ancient ideologies of large bodies of water.
"Computational theories of mind are often said to require mental representation because 'input' into a computation comes in the form of symbols or representations of other objects. A computer cannot compute an actual object, but must interpret and represent the object in some form and then compute the representation." There has to be a computational representation in order to experience emotion, therefore Computational-representational understanding of mind is how emotion is understood scientifically. It's all relative to representation.
I agree with you.
Marvin Minsky's theory Society of Mind, is most likely based on a mental representation of actual Agent Based Systems.
1:14:33 ... Coco just got very sad!
When she said European American, I got goosebumps....
🤲🏼
what is the term she says @11:58 ??
Aafact..?
Affect
❤❤❤
Ok, so if emotions are learned, then why do new born infants cry? They cry when born, when hungry, when stressed/overstimulated, etc.
even adult emotions are affected by things like physiological changes caused by hunger. Some people aren’t even aware that their behavior to certain stimulus changes when you are hungry, neutral, or satisfied.
Also consider the innate sensation associated with confusion, or exposure to novelty, or learn something.
Feeling seems to be an innate perception. The brain senses the input from neurons in our gut, and other organs in the body, and in turn produces physiological changes in brain chemistry which also seems to be perceived by the brain. It’s like a feedback loop. In fact, the act of conscious thought can provide input which affects brain chemistry and which the brain further perceives.
Lisa is insanely intelligent
Editing problem
In this woman's defense, It can be really annoying when CS people look at eveything through that same computational lense. "yup you heard me correctly, the brain is a digital computer and the Universe is a simulation"
That is what Cs people think but you make it overtly stupid sounding by giving no reasoning. The argument of computers just being a reflection of brain is not a philosophical one, but its a design feature, starting from Turing's definition of universal machines. The ability to read and write from a tape based on a deterministic program is 'thinking'. You can definitely challenge this premise itself, but most soft sciences haven't done much of that yet.
So, this was a very interesting interview... but it begs a number of questions:
*What about the gut-brain axis, emotional link? #NutritionalPsychiatry and #BrainEnergy are a thing; why haven't we begun to discuss protocols for predicting (& strategically planning, for social sciences & statistical developments of) emotions using nutritional neuropsychiatry?
We can do a lot of emotional prediction (*including such things as, crowd control, preventing mob violence, and preventing sexual aggression) based on simple statistical surveys & analysis of nutritional&diet quality of a location or population.
Not all emotions are created in the brain. Think about it: where does seretonin develop? How does one's emotions change after a good meal, or after a bad one? What about when they have IBS (irritable bowel syndrome) or some such similar physical ailment where the waste in the body is literally leaking into the blood stream through holes in the gut/colon microbiome, and how does this affect emotions?
*Nutritional Delivery Matters. When nutrients are more bio-available the body functions better, and the whole system improves. Why isn't this holistic ideal being included in the conversation?
I mean, she did include "metabolic costs" as a variable in emotional function... but I feel she was incredibly lacking in depth. Maybe given the audience.
I just think this superficial facial expression and visual analysis is so cheap and limited. It is appallingly, shockingly inaccurate.
*Also: What about the emotional-vibrational scale?
We have tools such as heat maps and musical scales / acoustic listening / radar / sonar / etc. for visually mapping using heat and sound (*or other environmental chemical, even a stress hormone or sweat measured produces a distinct scent and chemical production in the environment, which can be measured objectively and interpreted using data analysis).
Why aren't we using these tools to measure emotional intelligence? And coordinate social harmony like you would a musical ensemble, or band.
*Memories are held in cellular memory; i.e., the body keeps the score.
We have to look at: cellular garbage collection and removal, techniques for emotional therapy such as EFT (tapping) and musical therapy (*look at Dr. Emoto's water droplet experiments and the fact the human body is >70% water, can be easily manipulated. What crappy, shitty, hostile music is playing on repeat on the radios? It is acoustic bombardment, bringing down cellular function through crappy modern rap music filled with gun violence and hookers, breakups and alcohol lyrics, etc.).
*Why aren't we simply measuring the "color score" of people's auras? This is easily captured with various photography or sonar/etc technologies.
Just some food for thought. I hope to stimulate an interesting discussion. Thanks.
*Sadder thought:
If computer scientists are that bad at comprehending and thus predicting emotional states, how can they be trusted to handle sentiment analysis and neurolinguistic processing, and the semantics of language / social media? I am done with being atrociously mishandled on social media because of these stupidly-produced machine learning algorithms.
It is time to speak up. I am creating change in this sphere. Thank you.
This talk brings home the realization that a deep danger from AGI will come not from something intrinsic to artificiality but rather to the naive assumptions of the designers not yet being aware that emotions are social constructs; that a wookie that can looks like it feels X, inspiring empathy by design, equates to a wookie which deserves that human empathy.
Emotions are not social constructs.
Triggers of emotions can be social constructs.
@ < @ 23: 52 >23:57
The signal is the signal,
The character is the character,
a cigar...
Anyway, true enough, and explaining "Ignoratio Elenchi" is difficult for this reason.
~ Ignoratio Elenchi, the act of refuting an opponent by inserting (interpolating) something that may be perceived as disprovable.
The act itself, is the action, the definition, the crime, etc.
The Ne, Fi, Te, Si, is a great "character" to explore this role and I'm sure it will bring great research; Ti at deceiving.
Right?
@ 5:37, Emotional Expressions tell us a lot, but body language tells a story. Displacement and historical values can create emotional disturbances and force agitation. Unfortunately, we are all here to study life and at any base level, emotional, or chemical expression is fundamental. In a way, we are doing just that, attempting our best to read a person as if they are words on a page. However, their words are living, and reading with the correct language is the issue. Most researchers are attempting to find acute features and not patterns for expressions. The difference is this, working for the principle, or registering the patterns of language. Are you reading the word "hello", as a digital copy? and what is the analog copy? Hey Lamed Lamed? Or Eta Eepsilon Lamda Lamda Omicron? Looking at a face is useless if you can't see how their feet are positioned, elbows poised, or other indicators and history to triangulate the immediate emotional book you call human.
*Note, lamed, lambda is a construction of its own using, yod, vav, caph, which makes lamed; י ו כ ל
A bad fiction in Emotion!
Your emotion theory does not include the wild emotion changes that happen with certain drugs. For example, I was prescribed Zoloft but it made my
emotions feel much more angry. Some other legal drugs change my feelings in other ways. So much I suspect natural emotions are more chemistry based than reality based.
Love this woman. She is cool...
49:00
I wanna see a conversation between LFB and JBP tbh.
Wow, so few actually understand what we are.
"Some cultures don't have emotions" I don't understand how that possible, can you explain further. I feel like saying she's just wrong, but I want to hear further what she means by that.
She's very knowledgeable
Constructed emotion theory is only too easy to disprove; @t
can you elaborate?
@ 26:33, it was triangulated.
This was really an odd interview to me. On the one hand she seems brilliant and shared some really interesting and important details about the brain; for example when talking about intentionality & volition or the function of neurotransmitters in the brain related to emotions.
On the other hand she sometimes seemed to be making some quite obvious mistakes. For example stating that people in some cultures don't have emotions. What really struck me is when she said that every emotion is learned when growing up, implying that there is nothing hard-wired in the brain related to emotions. If this is true, where do innate fears of certain types of animals (spiders etc.) come from, or certain types of instinct related to emotions that many mammals (and probably humans also) poses. I agree that many complex human emotions are probably learned, but I can't imagine the really primal emotions don't have a fundemental (always present) role in the brain.
Innate fears do not exist. I've never seen a baby afraid of snakes, lizards, spiders...it's only after growing a bit that those "fears" emerge hence learnt. The proof is that some people manage to get rid of their phobias and this means that those "fears" were picked up along the way and by definition can be disposed of.
Legoist so who's teaching them spiders are dangerous? I'd say mostly parents do the opposite. I think the fear comes from deeper down and we actually learn not to be afraid. Most still have an strange feeling when watching a close up of a spider, but their rational part of the brain suppresses this. I think she underestimates the amount of instinctive emotions (and reflexes) that develop in the brain which are not learned
I heard her say that some cultures did not have a concept for emotion, not that some cultures did not have them. But later when she did say that some cultures have none, it was to contrast emotion with affect. I took her point to be that it is the affect (eight of them, I think) which is biological and emotion which is socially constructed from it and context (acausal reasons for the affect).
Christopher Inman than she has a different view of emotions than most people do. Most people mean emotions as feelings of happiness, sadness, remorse, fear etc. There are many non social mammals, and I'm pretty sure all of them have these primal emotions. My whole point being that the basic emotions are not socially constructed, but are made more complex by social interactions
Just don't believe her. She is not brilliant at all. At 11:03 she said in some cultures people don't experience sadness. Game over.
Shindlers List and Shawshank Redemption and Paul Verhoven's Turkish Delight made cry. Emotions are a double edged sword, for example love from oxytocin creates parent-child bonding, love bonding, it also makes you nicer to people who you are already predisposed to be nice to and Oxytocin makes you meaner arsehole to people in your out-group. When you look closely, Oxytocin reinforced wonderful 'us' vs. horrible, demonized, dehumanized, 'thems.'
the thing is who counts as an 'us' or a 'them' is easily manipulated. Just give everyone the same colored hat *bam!* instant in-group member. Politicians know this all too well.
Aren't Emotions neurotransmitters and hormones? There's several hundred of types of neurotransmitters which double as hormones. whether a neurotransmitter is an emotion/hormone or a thought or a biological behavior depends on where in the brain or body it binds.
I cry from the oviedo Beaches but I also cry over seeing people work together. Lol
She talks a lot, and is a bit curt, but I like her. Cuts through the shit and gets down to it no sugar coat or side step included.
She was actually lecturing on a subject where most of the "experts" are from the engineering background or at least something to do with CS. It's perfectly normal and acceptable to some extent that she's actually intolerant of people who seemingly make the same mistake of having false assumptions on cognitive science not based on research but their intuitions. Of course, we would all expect that a cognitive scientist would possess a much better set of communication skills but hey, she's one of the most knowledgeable people on this field so why would she have to be humble about it? Also, most of the people thinking she's rude they would've stick with the adjective "confident" if this was a man who was giving the lecture.
Totally agree with ur last sentence.
I will skip your last sentence. Anyway, I think she was awesome, best presentation on human emotion. She just summarized everything to this mathematical geeks who think they can quantify, or numerically summarize our emotions in some arrays and than upload the entire human emotion some future robot brain. Daaamn. I agree with her on her narrative that the human brain is super complicated and our intelligence is a product of what we learn from the environment.
@Deniz Kalaslıoğlu I call bullshit on your last sentence. Please link an example where a man talks over someone answering a question like that and people don't say, 'What a douchebag.' 'What a dick.' - There was a recent video of Steven Pinker: 'Steven Pinker Defends James Damore Against Dishonest Slanderer" where he strongly disagrees with a statement made by a panel member, but does not interrupt him, waits for the speaker to finish, and then strongly objects that 'x did not say etc etc...' That was extremely well done of him and is what both men and women should strive for in debate/discourse.
Please interview more women in science Lex! This is very inspiring and exciting.
Conversation 🙏🇺🇦😎
I guess this is what happens when we must believe there is no soul or spirit for that matter
I can't stop noticing how she is waving her fingers...
the problem was she had a different definition/view of the term emotion than is used in everyday conversation. that was cause for some misunderstanding. and yes, she was a bit rude (didn‘t bother me too much tho).
Dr Paul Ekman
I barely comment but I absolutely have to at this point. It, ultimately, emphasizes on her ability to interpret and subsequently understand (needlessly to say, important) questions.
There is absolutely no doubt that Lisa Feldman Barrett is extremely knowledgeable in her field. However, I have watched quite a few of her videos and inititally was not impressed. This did not turn me off because there is potential for massive conversions, where she otherwise - interrupted, misunderstood, preassumed, or thought she was above the question, either because she feels like the book was her ultimate argument can cannot replicate the answer, or she is getting way too much credit that she unjustly deserves.
As a psyhologist, she fails in every aspect and I lose more respect for her as I watch more videos. I love how one of the audiences absolutely dismissed the rest of her advice and inevitably refrained from further information from her, as he couldn't finish his question with her constantly interrupting, with, seemingly, an arrogant appoach of knowing everything.
its painfull to watch how narrow and non plastic everyone see biology here, and how broad she explain it is. their words really defies stereotypes about everything. (use of chemicals, the emotions as a part of us, etc). her words are trully gold for an Aİ developer, but he/she has to be flexible enough to strip away any preconception about how our biology work, in order to understand her fully.
I think that if AI becomes a monster, it will be a result of foolishly instilling emotions into the logic systems. Emotions can be useful and helpful, but they also easily breach logical restraint. Most crime is emotion-driven.
She's one allo-contradicting lady.
Would be nice if she actually gave a definition of what emotion actually is rather than what it isn't