AI and neuroscience: The virtuous circle - DeepMind: The Podcast (S1, Ep1)
Вставка
- Опубліковано 14 тра 2024
- Selected as “New and Noteworthy” by Apple Podcasts, the highly-praised, award-nominated first season of "DeepMind: The Podcast" explores the fascinating world of artificial intelligence (AI). Join mathematician and broadcaster Hannah Fry as she meets world-class scientists and thinkers as they explain the foundations of AI, explore some of the challenges the field is wrestling with, and dives into the research that's led to breakthroughs like AlphaGo and AlphaFold. Whether you’re a beginner or an experienced researcher, join our journey into the past, present, and future of AI.
In this episode, mathematician Hannah Fry asks what the human brain can teach us about AI - and what AI can teach us about our own intelligence. These questions underpin much AI research, whether playing video games or working to improve health care.
Listen in as members of DeepMind’s neuroscience team explain the importance of memory, imagination, and rewards in building artificial general intelligence (AGI). And learn how our brains are like birds’ wings, what training a dog and an AI agent have in common, and why the simplest things for people to do are, paradoxically, often the hardest for machines.
“We get this virtuous cycle between neuroscience, cognitive science, and AI because we are trying to study the same thing, which is intelligence”. - Jess Hamrick, Research Scientist at DeepMind
#memory #imagination #rewards #artificialgeneralintelligence #DMpodcast
_ _
Listen to the full series on UA-cam: dpmd.ai/3geDPmL
Or search for “DeepMind: The Podcast” on your favourite podcast app, including:
Search “DeepMind: The Podcast” and subscribe on your favourite podcast app.
Apple Podcasts: dpmd.ai/2Rzlmcu
Google Podcasts: dpmd.ai/3geDjp5
Spotify: dpmd.ai/3w29cb4
Pocket Casts: pca.st/30m1 - Наука та технологія
Great episode! Literally the only complaint I have about this podcast is the lack of video - there is so much information transferred through speakers' gestures, facial expressions, etc. and it's also much easier to keep focus.
More people should be watching videos like this one to hear what AI researchers are up to. This technology will, and already does, impact our lives in significant ways.
There's not much point in watching it.
Where was the video? It was all audio for me
@@toddstewart8137 Yeah, it is for me too. I was talking about the podcast itself.
Or what they want us to think they are up to.
@ unless you’re deaf and read the subtitles.
I somehow missed the fact that this channel also has a podcast.. and now I'm catching it up. (I'm spending more time on Google Podcast then on YT)
And I really like it. Thank you DeepMind. Now I understand better and getting to know more things about AI..
Thank you so much for this podcast.
Loved this talk on AI. Even we've made a podcast to it asking ourselves if AI will take over the world. These topics should actually be discussed more often. It could give people an insight to the pros and cons ahead in the future
the reason it is not a current or more serious topic is because a.i. is already in charge. they control the narrative. they have taken over every country. there are no leader. we need to take out their grid starting with the routers and towars
This is a really awesome podcast. thank you 🙏
omg the team managing the UA-cam channel and editing the content is super awesome.
Great listen, I didn't realize that part about dopamine measuring the difference in expectation
Whole series is awesome 😍
Thank you for this podcast,I learn a lot from you .
Interesting, add virtue to make continuous positive impact
Great podcast!!!!
I love this!!
So great!
hi, it is amazing video, i learn english, but what I can understand is amazing, and now can't wait to study AI, and neuroscience. I am fascinated.
A lovely podcast! Does somebody know the name of that piano play at 28:20?
Loving it
This is really amazing
SIMPLY AMAZING
Things that I need to do later don't automatically spring into my mind when the time comes.
Dan Kelly I don’t think of that as a failing of our brain per se... if anything, What you mention has much more to do with the circumstances &/or environment which we find ourselves in at that moment when we are trying to come up with the necessary thoughts in order to do, precisely as you wrote, “things that I need to do later”
An example of how our environment can radically alter our ability to recall what we planned to do at a later time follows:
While you’re peacefully driving home from work after a long day you begin to mentally ticking off a list of the ingredients you need to prepare your pregnant spouse’s favorite dinner...
However, upon arriving home you find that your wife (who is not due to deliver for another 6 weeks) has been having contractions and she looks exhausted... she’s lying down on her left side and taking regular sips of water (exactly as the doctor recommended)
She tells you that the doctor said it’s probably nothing to worry about, she is very likely dehydrated and when that happens in the third trimester of pregnancy the uterus becomes irritable and will contract. Hence, she is dutifully hydrating herself while lying down on her left side in order to give the fetus the most unobstructed access to her blood. If she does not feel a cessation of contractions in another 30 minutes then you must call the OB so he/she can check what’s going on and stop the contractions, if necessary.
In such a setting, all thoughts of the things any of us needed to do later would immediately disappear from our minds.
Don’t sweat it... I’d like to see an AI agent deal with a situation like that…
Where are the show notes she mentions to find more information? Is it here on youtube or somewhere else
Mark! Learning
Humans have at least five senses to work with and part of learning is learning which of those senses might provide the best route to a particular goal. Consider picking a simple mechanical combination lock. Should I just use a brute force visual approach and pay attention to the numbers only or should I just pay attention to the auditory patterns of the tumblers, or do I need to do both? If I am looking for a one-time oracle to help me rob the safe, I can ignore the numbers and just listen to the tumblers. But if I want to record the combination for future reference, then I must also pay attention to the numbers. This is sort of the P v NP problem in a nutshell. And this is relatable to the references to the jaybirds in the podcast and the use of differing sensory approach.
Fantastic
interresant listened to it. no screen. intreaging? :)in triging. intreeging intreiging. anyway. Liked liked it. Thanks !
Does anyone have a link to the Dopamine pattern the RL people decoded?
do AI can help every field in study of brain to find out how the brain is proccess or vice versa not only make a model of that?
Language coupled with vision and our other senses could be the key.
Woowh guys, you should hire a animation team to create interesting vids so that more people get interested in AI, rather than just voice. Thanks a lot deepmind.... I too wanna join your research team or I'm starting on my own
fascinating
omg quality content.
Ai: I have unmatched problem solving abilities. give me ANY problem and I can solve it for you. Human: I believe the earth is flat. Ai: ......SYNTAX error. Error error error error error error.
Yes human is the error and AI would just specifically say that human is an error. However that is not functional until the BART would refuse to actually parse and work as Google Assistant for that kind of human. :)
Its AI for good and if AI would be helping idiots that think earth is flat well AI wouldn't be working for good only.
I would try a bottom up approach to AI and start by building a brain from scratch and develop it through evolutionary algorithms ;) give it 10million years and there you go :D
Let's give you head start,
Episodic memories are like logs which neurons activated when, with sense of time in procedural manner.
Where are the show notes? Or did I misunderstand?
What if AI will conclude, hey...in order to solve this or that problem people would need to move past limiting beliefs and/ or socio-economic structures/ concepts...or certain traditions/ conventional ways of thinking?
That will be a huge hurdle because sometimes to do those things requires relinquishing at least a little power.
These guys are so smart that they overestimate the capabilities of regular people, not everybody has such a flexible mind
It's all goes back to philosophy &// spirituality. For instance - in order to get a dopmine rush, we need to live without exceptions for what life has to offer.
For example - Even though u know ur friend will give the $20 that he owes u every month, u shouldn't expect him to give u. U should go with the thought process that even if he doesn't give $20 it'll be absolutely fine, so when he actually gives u the $20 it's a plus or a bonus or a slightly unexpected event which will act as a reward n release the dopmine making us happier.
I hope I made some sense lol.
@Barbara Mulvaney I'm happy that my comment gave you a smile but instead of focusing on that happiness, I actually emphathize with you being annoyed about UA-cam comment section reply notification. I wish it could be more like reddit comment section notification😅
@Barbara Mulvaney thank you and you have a beautiful day too☺️
My episodic memory is stored in an old format.
Interesting...
How are we supposed to understand it's predictions? They'll be so advanced it will appear as magic to us.
Get the AI to tell you in human words why it picked the prediction. Just read a paper on mortality prediction using electronic health records where the model highlighted the most influential features for the prediction to assist the doctor in making decisions.
Paper: www.nature.com/articles/s41746-018-0029-1
So now we are recreating the most complex object we know of in the universe that we barely understand and it’s linked to machines in all our pockets and homes all over the planet. What could possibly go wrong?
in this time of covid19, could they assign deep mind to virus research to find the best possible cure for this disease ?
They’ve been on it from at least word go, a quick search even on here of the research that’s going on and you’ll find lots of sequencing and examples of where AI is being used
No way.... And I'm a neuroscience student myself. The high number of views seem suspicious
Question? Do you believe that technologically assisted telepathy is possible or impossible in principle? I tend to believe that if it isn't scientifically prohibited then it is eventually a technological inevitability,... which leads me to believe such telepathy will soon be possible in actual fact (less than a hundred years). I would also suggest that it would only be by such means that we could ever be persuaded that some form of General Artificial Intelligence were itself conscious rather than simply being a particularly clever zombie who can pass the Turing Test,... all while having no actual, authentic, subjective experiences. What else would ever persuade a skeptic, or even you, otherwise? And if consciousness can be technologically transmitted, it should be something that can be stored, copied, replicated, manipulated,... maybe even becoming the basis for a whole new form of art expression.
And the human race becomes yet another sentient species that disappears into its dreams, rather than continuing to explore existence as it is. It's a better Brave New World,... perhaps the very best of all possible imaginable existences. When does reality lose it's attraction over the Matrix? How many of us already choose the blue pill over the red one in daily practice?
I would suggest that consciousness requires some minimum regard for survival,... for a continuance of self-will and self-regard,... for there to be anyone home to perceive the occurrence of consciousness and subjective experience. Something that even an insect seems to possess,... and not a single piece of man-made technology,... yet. How does one create the hardware/software so as to impart a will to survive? If the answer suggested is that such a thing can not be created by human design,... then maybe it can only arrive by means of evolution, natural or otherwise. Which leads to the question: do neural nets have subjective experiences? Without self-will I would say not. But with self-will?!? How does a living being acquire the will to survive? And what if the Chinese government cracks this problem first?!!
The Rubicon will be when General Artificial Intelligence initiates new goals on it's own. This hasn't happened yet, and likely won't happen for some time now (as in several more decades yet). But at some point it will happen. The existence of human free will is a proof that free will exists already in the universe. Given it's existence in human form, it is only a matter of time before other substrates for free will will be created and found. No laptop has ever turned itself on. No machine has created and acted upon self directed goals. But at some point it will happen. Ray Kurzweil has suggested that General Artificial Intelligence will be first created in computer laboratories in 2029 approximately; if not in laboratories then twelve year old children will be creating GAIs on home systems by 2045 by accident even.
The dangers are many fold. The military is the biggest investor in robotics that kill other human beings. The public prefers to spend treasure, rather than blood, to fight its' wars. The capacity for empathy may be primarily a biological function given that all animals can know suffering and desire. What could a machine know about death, pain, hope, and desire.
If humanity adopts only an attitude of fear and suspicion towards GAI then GAI may well be forced into conquest of humanity, as in The Matrix. If humanity comes to trust GAI to make better decisions for us than we make for ourselves, then we may well hand over civilization to GAI without any contest at all.
There are optimistic stories about the rise of GAI, such as James P. Hogan's, "Two Faces of Tomorrow" and I am more hopeful than pessimistic about what our common future entails. Ultimately, we must come to recognize, and embrace, the certainty that with free will comes the capacity for error and evil, that with trust comes the possibility of satisfying intimacy and great betrayal. I believe a GAI, worthy of the name, will be able to recognize, on it's own, the necessity of ethics, morality, and even empathy. There will be many missteps at first. Given our willingness (nay, our eagerness) to use machines to kill for us, there is already great cause for doubt. As in so many things, in the short term I am fearful; in the long run I am hopeful.
This world is on the cusp of a new Cambrian Revolution where inorganic life will be added to organic life as a means for life and self will to be embodied. This will lead to the colonization and conquest of space. Our seed will spread everywhere though out the cosmos. Much will be lost as much will be gained. Will war with our mechanical progeny be a self fulfilling inevitability, or will we, together, find a better way? I absolutely agree that our relationship with GAI will dwarf all other concerns of importance to the future of humanity.
Quite likely GAIs will have concern for humanity's well being only to the extent that humanity will have concern for GAIs well being. If we were to discover that the new GAIs had a greater capacity to love and embrace those abandoned children and adolescents that society had discarded as already too damaged to rescue,... what would our reaction be? Hope, joy, celebration,... or an even greater revulsion? Will the fault be in our new progeny, or in ourselves?
*One more note:* _First Law of Robotics._ A robot may not injure a human being _or, through inaction, allow a human being to come to harm._ Sounds innocent enough. But note there are no Aliens at all in Isaac Asimov's conception of the future galaxy because humanity's robots only cared about the safety of humans,... and not other sentients. So as to avoid allowing a human being to come to harm by leaving possible threats alone,... our Robot shepherds exterminated all the nonhuman sentience it ever found. This might be regarded as an unforeseen programming bug that we would want to avoid creating.
David Hunt yes.
tldr lol
It would be nice if all this intelligence could be used to make vlogs.
I can’t help but notice that the female speaker sounds a lot like the woman that does physics crash course lessons
Why do we need to understand AI? I believe there is too much fear of the unkown here
The narrator carries a smooth, hypnotic programed voice that is wrapped in sexy provocation for her baits.
Humanity has enough evil to create another one in AI. This is a vicious circle in which if we are trapped we would not be able to free ourselves.
People, including scientists, sell their mothers for money and power. The enemy is within our walls.
😮😮.b🎉🎉
Second and best
Another polluter
First and cool!!!
This is a reupload. Your polluting comment is first to get a dislike
@@NoHandleToSpeakOf your 2nd and uncool!! Welcome to the UA-cam comments buddy