video_id
stringlengths 11
11
| text
stringlengths 361
490
| start_second
int64 0
11.3k
| end_second
int64 18
11.3k
| url
stringlengths 48
52
| title
stringlengths 0
100
| thumbnail
stringlengths 0
52
|
|---|---|---|---|---|---|---|
P0yVuoATjzs | there's a hypothesis that comes from the computational work that says well we should be able to decode you see friends copy signals in you know potentially even in v1 and it turns out that that is exactly what we can do so so these blue bars are in complete darkness and visual cortex so there's no visual stimulus they're not getting any visual optic flow but we can decode a little bit of | 2,042 | 2,062 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2042s | Predictive Coding Models of Perception | |
P0yVuoATjzs | you know sort of six degree of freedom information about which way the animals nose is pointing and then interestingly if you inject mute samal into area M two which is the putative place where these motor signals you friends copies would be coming from you know this is all still new so you know don't get futzed if this ends up not holding up because we still only get more animals at least | 2,062 | 2,083 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2062s | Predictive Coding Models of Perception | |
P0yVuoATjzs | in the first animal we did we can actually abolish that signal so if we take away motor cortex signals we actually can't any longer decode the position of the animal's head in 3-space so this is a case where we actually built them up we built a model that was useful in its own right it helped explain some things about neuroscience more or less along the way without | 2,083 | 2,103 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2083s | Predictive Coding Models of Perception | |
P0yVuoATjzs | having to fit anything and then by looking at it we can make predictions which then led to experiments that we could do to maybe learn something new about how the brain is organized corollary discharged we can't tell and in an electrophysiology we just have some Tet Rhodes recording from some places and we're not exactly sure where we can't reconstruct the image to fully | 2,103 | 2,153 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2103s | Predictive Coding Models of Perception | |
P0yVuoATjzs | disambiguate that we also have two photon imaging going on in my lab so there is an idea that we could go and be getting you know hundreds of cells that are thousands of cells at a time you know and have a prayer of actually doing those those decoding experiments and that's work that we have that's the song going but but you know right now we're just in the stage where we said can we | 2,153 | 2,175 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2153s | Predictive Coding Models of Perception | |
P0yVuoATjzs | decode any information and the answer seems to be yes I agree there's a lot of questions that come downstream of that like well you know predictive coding implies certain kinds of correlational structures that we can start to look at we also have the ability to image synapses in the two-photon image of the synapses differently separately from the cell body so we can actually look at | 2,175 | 2,196 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2175s | Predictive Coding Models of Perception | |
P0yVuoATjzs | feedback synapses specifically and ask how the information that they contained is different than the the cell bodies in layer 2 3 so these are all experiments that are ongoing but I don't have a good answer for you but that's the right kind of question and that's the right kind of question we can ask you know when we have these kinds of models they can guide our experiments ok so this is my | 2,196 | 2,214 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2196s | Predictive Coding Models of Perception | |
P0yVuoATjzs | pitch so apparently a deep mind it's a it's a circle for me it's just a it's just a ping-pong I don't know I like the circle better maybe I'll change it top away oh yeah that's right so people are picking this up Gooding neuroscientists and we love that if you want to pick it up there it is the codes all free I'm told it's not hard for other people to get working so that's | 2,214 | 2,238 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2214s | Predictive Coding Models of Perception | |
P0yVuoATjzs | great that's a testament to Bill's hard work to share things so I just want to try and get us back on schedule so this is my lab just went to acknowledge everyone particularly Bill water and then all the people who funded this stuff and then just one shameless plug you know so I've just started as the director of this institute if you're interested in sort of AI and | 2,238 | 2,259 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2238s | Predictive Coding Models of Perception | |
P0yVuoATjzs | neuroscience Nexus the kinds of stuff that I'm showing you here and you reached in doing that in industry you know find me during one of the breaks I'd be happy to talk right thanks [Applause] yes oh the most of the mud so the motor modulation it's there's a whole cottage industry of like my son trackballs when they run versus when they don't run and and basically there's more information | 2,259 | 2,320 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2259s | Predictive Coding Models of Perception | |
P0yVuoATjzs | about whether the running or not then there seems to be about the visual system that's sort of like a first approximation but but a lot of the there's been a sort of a dichotomy of people who are either either think you know maybe this is a predictive process and kind of thing versus people who think it's more of like an arousal sort of like when the animals running they're | 2,320 | 2,335 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2320s | Predictive Coding Models of Perception | |
P0yVuoATjzs | on and when they're not running they're off or something and you can probably imagine which one I think is the more likely scenario and so so so the idea that we can actually get information about the actual direction the nose is pointing I think is interesting and a little bit surprising and and feels like at odds with the idea that it's just sort of an on versus an off kind of | 2,335 | 2,355 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2335s | Predictive Coding Models of Perception | |
P0yVuoATjzs | transformation now in terms of vestibular signals yeah I mean we're we've talked about you know the you know the experiments aren't pretty but you can imagine what they're like you know it's like I'm just moving I'm moving the rat in the dark Here I am with my rat moving in the dark so you know like sometimes this thought you have to do the science and and it's not it's not | 2,355 | 2,375 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2355s | Predictive Coding Models of Perception | |
P0yVuoATjzs | glamorous so we say so so we aren't in those experiments I mean it doesn't sort of go against the general theme here like wherever you get the signals from you'd be crazy not to use them I think the fact that we see some reduction in them when we knock out em to suggest a little bit that maybe it's more of a motor you friends copy but it's entirely possible that this tabular system the | 2,375 | 2,395 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2375s | Predictive Coding Models of Perception | |
P0yVuoATjzs | vestibular inputs are getting through to v1 through so m2 is I mean it's like it's sort of like frontal orienting fields in a monkey it's like an orienting area yeah so as near as we can tell it doesn't seem to so we've done let's go to all the marginal comparisons so you'd expect it to have some effect it turns out you can actually just scoop out all of them fun | 2,395 | 2,423 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2395s | Predictive Coding Models of Perception | |
P0yVuoATjzs | all the motor cortex by my neighbor at Harvard Ben selves Eskie did this experiment are you basically just carpet bomb all of motor cortex and the animals can still do all kinds of complicated motor tasks so that that's alarming if you study motor cortex but but from our perspective it's it's actually a good things I mean so it's near as we can tell we're not changing with the | 2,423 | 2,444 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2423s | Predictive Coding Models of Perception | |
P0yVuoATjzs | statistics of the movement one of things were moving towards mister is trying to get more quantitative predictions about what would the actual predictive coatings that are subtractive having a model at least gives you a prayer of being able to sort of ask those kinds of questions I agree it's complicated it's not at the end not the end-all be-all and and also in monkey and primate IT | 2,444 | 2,465 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2444s | Predictive Coding Models of Perception | |
P0yVuoATjzs | there was there some beautiful studies that were they were kind of lesser-known where they showed that in complete darkness they were psychotic eye movement signals presence he could decode when cicadas happened so so these ideas aren't aren't new it's just for their sort of driving us in new directions [Music] but when you're comparing that to the neural state law of attraction the first | 2,465 | 2,501 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2465s | Predictive Coding Models of Perception | |
P0yVuoATjzs | example if you're showing if you're comparing that to the N stopping signals or the transient dynamics yeah it turns out if you look at them they almost all look the same which which is disheartening for people who want to disambiguate different populations I mean at least when you marginalize them in this way the the e neurons look like pop and then off response but so did the | 2,501 | 2,531 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2501s | Predictive Coding Models of Perception | |
P0yVuoATjzs | activation neurons and so there are recurrent neurons that are sort of sitting there and like like in Matt's talk like they're ones that are doing the things you know they have to be doing but there are ones in the in the recurrent layer that also have the sort of transient dynamics as well so so the long story is like if you were sticking electrodes into a monkey brain like I | 2,531 | 2,549 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2531s | Predictive Coding Models of Perception | |
P0yVuoATjzs | did for five years you know you wouldn't know which ones of these you were getting and they all look surprisingly similar I mean so in the in the cases where we're doing the cutting we're actually decoding from the ours so but I mean you you can decode lots of things from lots of places the arse we thought the arse made a lot of sense because they're the ones who should be | 2,549 | 2,575 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2549s | Predictive Coding Models of Perception | |
P0yVuoATjzs | holding information about context and representation but I mean there's obviously a lot going on yeah as you get further along though this is a CNN in the static case on the first time step because you basically there's nothing to cancel out this is all zeroes so you just go up through like a CNN so you would expect there to be sort of different levels of representation of | 2,575 | 2,594 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2575s | Predictive Coding Models of Perception | |
P0yVuoATjzs | higher level and lower level features so it's only long story short it's complicated the the simple fact of training it in this predictive mode is gonna affect all of these weights so it's going to induce representations even in these A's so the bottom line is like you can pretty much decode from anywhere this is bad news for Neuroscience right you can decode from | 2,594 | 2,614 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2594s | Predictive Coding Models of Perception | |
P0yVuoATjzs | anywhere and all of them qualitatively look very similar in their dynamics so I think about the relationship between what you're doing here and the stuff that I've been working on but honestly really the ideas that were put forth by hoc right originally about supervise and that work would suggest that this system should be able to do something even richer which is to do | 2,614 | 2,636 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2614s | Predictive Coding Models of Perception | |
P0yVuoATjzs | something like visual system identification so in the phenomena that you've you've talked about you look at the first frame and you can make predictions about what's going to happen because you live in this world but there are situations where I can't predict this observing the dynamics helps me predict the dynamic so if I like let me say it's seeing or someone with a fuse | 2,636 | 2,660 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2636s | Predictive Coding Models of Perception | |
P0yVuoATjzs | joint or something should be able to do that kind of system identification on the I don't know but I agree and we're I think we're all kind of using the same market fuel right it's like Ellis TMS can do a lot and these are sort of two separate instantiations of like recurrent Nets can can actually do do a lot yeah exactly exactly exactly and and there is a spot that you know they even | 2,660 | 2,688 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2660s | Predictive Coding Models of Perception | |
P0yVuoATjzs | after you fix things and it's running you know it has state that's that it's accumulating so it can you can do I think system identification so I mean this isn't this isn't fit exactly into your what you're talking about but we also had it doing things like predicting balls bouncing around and things like that and you know and it works you know like the balls bounce off the walls yeah | 2,688 | 2,717 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2688s | Predictive Coding Models of Perception | |
P0yVuoATjzs | and if we have these like have like electrostatic repulsion or something you could imagine them learning dynamically like well you know this is this is the new rules so yeah and this is the kind of setting I would probably prefer to do it in rather than the other one cuz you can't control the natural world as easily but but I think it's really interesting I mean that there are | 2,717 | 2,735 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2717s | Predictive Coding Models of Perception | |
P0yVuoATjzs | limits of the capacity of this thing to learn so it you can't you know if you have very complicated movies and things it doesn't do a great job of predicting the next frame so we've far from crack the nut but it just feels like we're kind of moving maybe a little bit in right direction so so much like the project I spent 20 years of my life working I've had so | 2,735 | 2,758 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2735s | Predictive Coding Models of Perception | |
P0yVuoATjzs | many people tell me that it ultimately got published at nips as unsupervised pixel prediction and it was just trying to predict a moving dot no recognition of anything it was only doing the motion different continuity in time I wrestled with it it was awful it worked kind of okay yeah yeah while I was working as a coder and an algorithm guy and stuff in Silicon Valley what some C++ and a job I | 2,758 | 2,803 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2758s | Predictive Coding Models of Perception | |
P0yVuoATjzs | worked at Redwood Neuroscience but ultimately what happened was I rely I alternate Lee gave a presentation here that it worked with just using matrices and just the algorithms forget the neurons it was that hard then I throw away even the matrices and finally got a zero parameter model where I just assumed volumetric 3d computational medium intracellular space so if you | 2,803 | 2,833 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2803s | Predictive Coding Models of Perception | |
P0yVuoATjzs | assume brains just do 3d computations I so I understand I think I think we understand continuity in time and it's just what he said yeah but trying to build it on top of deep learning which was built around essentially n dimensional hyperspace all we have to do with 3d continuous it's not easier oh this is a soft key in fits paper is that the one I think I think I | 2,833 | 2,865 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2833s | Predictive Coding Models of Perception | |
P0yVuoATjzs | think we I mean I think we I think we saw I think we cited you but the I think I think I think the you know the tools I mean the thing that's nice about deep learning and maybe others have this sort of intuition about it it sort of nice to be able to say here's our like I'm saying like presto magico deep-learning what I'm saying like I think we've gotten to the stage where but that's all | 2,865 | 2,886 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2865s | Predictive Coding Models of Perception | |
P0yVuoATjzs | abstraction now what we can say is if we just optimize future frame prediction what else obtains and I think that's the right framing for using deep learning to sort of it's not that we're like doing neurons or that they are neurons like this has back prop like clearly you know like there's gonna be a couple suggestions on how we can do it better it's more like if you just optimize this | 2,886 | 2,903 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2886s | Predictive Coding Models of Perception | |
P0yVuoATjzs | one thing and deep learning lets you very effectively optimize that one thing this is what comes out of it right so that that's kind of the mode we're thinking about it and you know could it be reduced to some other simpler thing sure but we just wanted to have the tools that take us to there to that to that point and I think we're I think we're out of time unfortunate this or | 2,903 | 2,936 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2903s | Predictive Coding Models of Perception | |
P0yVuoATjzs | that feature of object this is the kind of work you did in your thesis and your adviser certainly you know should not fire at all if it's correctly representing this thing and so I'm trying to picture how the same room can be both representing something and they're not representing something and and the question is maybe the answer is in that your last picture when you said | 2,936 | 2,961 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2936s | Predictive Coding Models of Perception | |
P0yVuoATjzs | so on the first pass to reader model basically the representational kind of thing is getting passed forward through these error units which are the only things passing anything for which is a pretty strong constraint in their model and so but then later those same units must be passing forward something it means something different which is you know the the non predictive part only | 2,961 | 2,984 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2961s | Predictive Coding Models of Perception | |
P0yVuoATjzs | those are two very different things at very different at different times and so are you is that the correct correct interpretation which is the sort of there's a time multiplexing where maybe a first pass through the ventral Strait is the representational pass and could lead to successful recognition of an object and then later that thing is somehow shut off because everything is | 2,984 | 3,004 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=2984s | Predictive Coding Models of Perception | |
P0yVuoATjzs | predictable or what have suggested time multiplexing in the responses before so there's a paper from from yesterday's group I could be wrong about that where they're basically claiming that the initial pop of activity had different information than the latter part I would say when when when Jim and I have said in the past that this neuron represents that thing we're taking time like cat | 3,004 | 3,033 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=3004s | Predictive Coding Models of Perception | |
P0yVuoATjzs | spike counts within 100 millisecond window after you know 100 milliseconds after onset of stimulus that that same neuron is categorically not doing that 100 milliseconds after that right like you know there's this weird phenomena that a lot of the neurons are shut down so I think this idea that these neurons the firing of these neurons signal that you know without any further additional | 3,033 | 3,054 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=3033s | Predictive Coding Models of Perception | |
P0yVuoATjzs | context I think that that kind of has to be wrong that's on the level I mean the fact that these things are so dynamic I think the end of the day all the system cares about is that makes the right call and initiates the right behavior I don't think is any requirement that there be labeled line neurons that signal features but so this is an example where you have neurons that will look that way | 3,054 | 3,075 | https://www.youtube.com/watch?v=P0yVuoATjzs&t=3054s | Predictive Coding Models of Perception | |
EkzZSaeIikI | I kept trying to get this secret to work for me for such a long time and it just wouldn't work until I changed one thing and I just want to show you what that one thing is are you ready let's go when it did work what happened is I manifested this no I'm only joking watch this because actually what I manifest was much more than just a five thousand dollar massage chair | 0 | 32 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=0s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | aside from a $20,000 watch I don't know if you can see out here and then this all became my lifestyle alongside that beautiful little boat there now your question is why am i showing all of this to you and what a show-off who is this guy anyway I just want to break it all down for you because it's not about me and I showed you all of this stuff it's because I want you to know that it is | 32 | 66 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=32s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | possible one of the biggest things is that we need to believe before we can manifest and that's the biggest problem that a lot of people don't know how to get so I want to break it down for you into three simple steps in this video how to actually manifest and attract what you like what you desire in life using the law of attraction or the principles from the secret but before we | 66 | 88 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=66s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | get into it so real hit it Bob today isn't that they're struggling it's not they can't get out of this it's because the that they're in isn't big enough and why I say that is because if that was big enough right now if the people had a girl acquaintance or that head they would actually get that ass moving always understand that emotions is what these emotion is energy in motion so if your | 88 | 111 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=88s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | emotion or drive is not stronger you're not gonna do anything to get out of it what's up guys this era coherence national speaker entrepreneur and best-selling author and in this video I'll break down for you the three keys to make the secret work for you finally while I relax on this chair love it so much so we're gonna get straight into it the first thing using the law of | 111 | 134 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=111s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | attraction is we must visualize and this what the secret talks about but you know just visualizing alone is really not going to be able to allow the manifestation to work for you the biggest problem is is because there's a lot of that goes on through the day so let's say for example you spend I don't know three minutes five minutes visualizing in the morning the rest of | 134 | 157 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=134s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | the day you're visualizing something else so what that means is it kind of like puts us in a position where we're manifesting everything that we don't want and the small time and energy that we put into what we do want isn't manifesting first so what is the solution the solution is to understand a step number two and step number two is the principle of the law of attraction | 157 | 178 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=157s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | we have Canada here say hi the principle of the law of attraction is to understand that it's all based around emotions the more you are able to magnify your emotions what happens is a speeds the whole law of attraction of but aside from that it makes it much much more powerful because vibrations if you think about it when you are very very emotionally charged up your | 178 | 204 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=178s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | vibration is kind of raised to a crazy crazy crazy vibrational frequency and the more crazy it is the more it sets it in there so even if you just do it for three minutes five minutes in the morning hold stronger vibrations than your whole entire day so it's all about the second part raising those emotions and really getting into the group so step number one is to actually visualize that number | 204 | 231 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=204s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | two is to really amplify the emotions as if you're living it right now and finally step number three Aires with that step number three is the most important part at all which is to let go now you're thinking what do you mean like let go I think a lot of people recently have been getting really really confused when I say set it and forget it that's what I write about in my book | 231 | 253 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=231s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | they say Eric if I said it when do I said it how long do I sell it for when do I let it go what's the balance between setting it and forgetting it well actually you said it early on in the morning when you're most at peace the rest of the time you've got to have trust I was doing this call yesterday for the whole superphone thing I know a lot of you a lot of you have been | 253 | 273 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=253s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | texting me over the last couple of days and I am receiving your messages by the way it's just like there's thousands and thousands of thousands of requests coming in so it's very hard to actually reply to you all however one of the biggest questions that I keep getting from a lot of you is when you said it when do you forget it and also how do you actually trust the universe what if | 273 | 293 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=273s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | you have a lot of doubt happening within you because if there's doubt it's very hard to make the Law of Attraction work for you so the whole idea of it what I was teaching yesterday during the live call was that you need to get this into your head and I want you to type this in the comments below those of you who are new to this channel we always type our learnings below to reaffirm the learning | 293 | 316 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=293s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | Who am I not to trust I've run my whole life based on this very very simple concept Who am I not to trust if you understand we were all created by something that's greater than us much powerful force a much magical force you can call it God you can name it the universe you can name it whatever you want to name it but we can't deny that thing exists now if that thing exists | 316 | 342 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=316s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | who are we not to trust it has the decision to put you into this world when it wants to and it also has the decision to take you away and sometimes think about it how many times throughout your life so far have you felt totally out of control you don't know how to handle it you don't know how to deal with the situation and you felt so so challenged and you were | 342 | 361 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=342s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | just stuck but somehow you pulled through and that magic happened so if it's ever happened to you before whether in relationships and business and finances whatever it is then if nothing actually happened before who are you not to trust because it always always the universe always always has your back and if you understand that and have that peace of mind then when you have that | 361 | 388 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=361s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | peace of mind everything starts shifting and Wow we start shifting we don't need to play your song I'm gonna play your song it's beautiful so I'm ready I made it for too long have you played [Music] a prize for whoever guesses what the song is called you can comment below anyway guys I'm gonna finish off for today because we have to go up to Newcastle with ah okay you know where | 388 | 431 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=388s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
EkzZSaeIikI | are you because today is dad's birthday so I'm a blog along the way I don't know it depends if we have enough time to do so but anyway guys if this video has been of any use to you whatsoever you know what to do hit that thumbs up hit the like button also comment below let us know where you guys have tuned in from and finally if you're new to this channel and have been any used to you | 431 | 451 | https://www.youtube.com/watch?v=EkzZSaeIikI&t=431s | Why 'The Secret' Won’t Work For You Until You Do This.. [Law of Attraction] | |
pWAc9B2zJS4 | hi-yah thanks a lot for joining us today thank you for inviting me actually I was glad to be here yeah one of the world's most visible deep learning researchers life asked me to share a bit about your personal story so how do you end up doing this work that you now do yeah that sounds great I guess I first became interested in machine learning right before I met you actually I've been | 0 | 24 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=0s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | working on neuroscience and my undergraduate adviser Jerry Kane a Stanford encouraged me to take your internet AI class I didn't know that okay so I had always thought that AI was a good idea but that in practice the main thing I knew that was happening was like game AI where people have a lot of hard-coded rules for non player characters and games to say different | 24 | 49 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=24s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | scripted lines at different points in time and then when I took you were injured AI class and you covered topics like linear regression and the bias and various variants decomposition of the error linear regression I started to realize that this is a real science and I could actually have a scientific career in AI rather than nursing great and then what happe well I came back an | 49 | 74 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=49s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | IT a two-year course I'm late I see great education so the really big turning point for me was while I was teeing that course one of the students my friend Ethan Dreyfus got interested in geoff hinton deep belief that paper and the two of us ended up building one of the first GPU CUDA based machines at Stanford in order to run Boulton machines in our spare time over winter | 74 | 101 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=74s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | break and at that point I started to have a very strong intuition that deep learning was the way to go in the future that a lot of the other algorithms I was working with like support vector machines didn't seem to have the right asymptotics that you add more training data and it gets lower or for the same amount of training data it's hard to make them perform a lot better by | 101 | 125 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=101s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | changing other settings and at that point I I started to focus on deep learning as much as possible indeed and remember Magette rain has very old GPU paper race acknowledges you for having done a lot already work yeah yeah as I was written using some of the missions that we built the first machine I built was just something that Ethan and I built that Ethan's mom's house | 125 | 152 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=125s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | that I would wait with our own money and then later we have money to build the first to agree for the Stanford lab wow that's great I never knew that story of Jason oh and then today one of the you know things that's really taken the deepening world by storm is you invention of Gans so how did you come up with that I just studying generative models for a long time sort of a Gans | 152 | 177 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=152s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | our way of doing generative modeling where you have a lot of training data and you'd like to learn to produce more examples that resemble the training data but but they're imaginary they've never been seen exactly in that form before there were several other ways of doing generative models that had been popular for several years before I had the idea for again and after I've been working on | 177 | 202 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=177s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | all those other methods throughout most of my PhD I knew a lot about the advantages and disadvantages of all the other frameworks like Bolton machines and sparse coding and all the other approaches that had been really popular for years I was looking for something how to avoid all of those disadvantages at the same time and then finally when I was arguing about 200 models with my | 202 | 223 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=202s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | friends in a bar something clicked into place and I started telling them you need to do this this and this and I swear it'll work and my friends didn't believe me that it would work I was supposed to be writing the deep learning text book at the time but I believed strongly enough that it would work that I went home and coded it up the same night that it worked | 223 | 239 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=223s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | so thank you one evening to implement the first version as I if I implemented it around midnight after going home from the bar where my friend's house is going-away party and the first version of it worked which is very very fortunate I didn't have to search for hyper parameters or anything it was just for me I read it somewhere where you had a deaf experience and that reaffirm your | 239 | 261 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=239s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | commitment to AI tell me that yeah I was I wasn't actually near deaf but I briefly thought that I was I had a very bad headache and some of the doctors thought I might have a brain hemorrhage and during the time that I was waiting for my MRI results to find out whether I had a brain hemorrhage or not I realized that most of the fact I was having worry about making sure that other people | 261 | 288 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=261s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | would eventually try out the research ideas that I had at the time in retrospect they're all pretty silly research ideas but at that point I realized that this was actually one of my highest priorities in life was carrying out my machine learning research work yeah that's great that when you thought you might be dying soon you're just thinking how we get the | 288 | 311 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=288s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | research done yeah yeah that that that that's commitment yes yeah so today you're still at the center of allow the activities with scans of generative atmosphere networks so tell me how you see the future of Gans right now Jen's are used for a lot of different things like so my supervised learning generating training data for other models and even simulating scientific | 311 | 337 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=311s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | experiments in principle all of these things could be done but other kinds of generative models so I think that games are at an important crossroads right now right now they work well some of the time but it can be more of an art than a science to really bring that performance out of them it was more or less how people felt about deep learning in general 10 years ago and back then we | 337 | 360 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=337s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | were using deep belief networks with bolts and machines as a building blocks they were very very finicky over time we switched to think like rectified linear units and Bachelor realization and deep learning became a lot more reliable if we can make games become as reliable as deep learning has become but I think we'll keep seeing games used in all the places they're used | 360 | 381 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=360s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | with much greater success if we aren't able to figure out how to stabilize games but I think their main contribution to the history of deep learning is that they will have shown people how to do all these tasks that involve generative modeling and eventually we will replace them with other forms of generative models so I spend maybe about 40% of my time right | 381 | 404 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=381s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | now working on stabilizing games he's cool oh and so just as a lot of people they join deep learning about ten years ago such as itself ended up being pioneers maybe the people they joined Gans today if it works out could end up the early pioneers yeah a lot of people already are early pioneers of games and I think if you wanted to give any kind of history of again so far you'd really | 404 | 429 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=404s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | need to mention other groups like indigo and Facebook and Berkeley for all the different things that they've done so in addition to all your research you also co-authored a book on these learnings oh that guy that's right with Joshua Benji oh and Aaron kohrville who were my PhD coat visors we wrote the first textbook on the modern version of deep learning and that has been very popular both in | 429 | 458 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=429s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | the English edition and the Chinese Edition we sold about I think around 70,000 copies total between those two languages and I've had a lot of feedback from students who said that they've learned a lot from it one thing that we did a little bit differently than some other books is we start with a very focused introduction to the kind of math that you need to do deep learning I | 458 | 483 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=458s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | think one thing that I got from your courses at Stanford is the linear algebra and probability are very important that people get excited about the machine learning algorithms but if you want to be a really excellent practitioner you've got to master the basic math that underlies the whole approach in the first place so we make sure to give a very focused presentation | 483 | 508 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=483s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | of the basics that's a start book that way you don't need to go ahead and learn all of linear algebra but you can get a very quick crash course in the piece of the linear algebra that are the most useful for deep learning so even someone whose math you know is real shaky you've ever seen in math for a few years we're going to start from the beginning of your book | 508 | 527 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=508s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | and get that background and get into deep learning all of the facts that you would need to know are there it would definitely take some focused efforts and practice that making use of them great if someone's really afraid of method it might be a bit of a painful experience but but if you're ready for the learning experience and you believe you can master it I think all the all the tools | 527 | 550 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=527s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | that you need are there as someone does work in designing for a long time I'd be curious if you look back over the years tell me about how about how you're thinking of AI and a deep learning has evolved over the years ten years ago I felt like as a community the biggest challenge in machine learning was just how to get it working for AI related tasks at all we have really good tools | 550 | 576 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=550s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | that we can use for simpler tasks where we wanted to recognize patterns in hand extracted features where a human designer could do a lot of the work by creating those features and then hand it off to the computer and that was really good for different things like predicting which adds the user would click on or different kinds of basic scientific analysis but we really | 576 | 602 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=576s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | struggled to do anything involving millions of pixels in an image or a raw audio waveform where the system hasn't built all of its understanding from scratch we finally got over the hurdle really thoroughly maybe five years ago and now we're at a point where there are so many different paths opens that someone who wants to get involved in AI may be the hardest problem they face is | 602 | 628 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=602s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | choosing which path they want to go down do you want to make reinforcement learning work as well as supervised learning works do you want to make unsupervised learning work as well as supervised works do you want to make sure that machine learning algorithms are fair and reflect biases that we prefer to avoid do you want to make sure that the societal issues surrounding AI work out | 628 | 654 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=628s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | well that we are able to make sure that a AI benefits everyone rather than causing social of people and trouble with lots of jobs I think right now this is really an amazing amount of different things it can be done both to prevent downsides from AI but also to make sure that we leverage all of the upsides that it offers us and so today there are a lot of people wanting to get into AI so | 654 | 680 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=654s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | what advice would you have for someone like that I think a lot of people that want to get into a I start thinking that they absolutely need to get a PhD or some other kind of credential like that I don't think that's actually a requirement anymore one way that you could get a lot of attention is to write good code and put it on github if you have an interesting project that solves | 680 | 703 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=680s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | the problem that someone working at the top lab one is itself once they find your github repositories they'll come find you and ask you to come work there a lot of the people that I've hired or recruited at opening I last year or at Google this year I first became interested in working with them because it's something that I saw that they released in open-source form on the | 703 | 725 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=703s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | internet writing papers and putting them in archives can also be good a lot of the time it's harder to reach the point where you have something polished enough to really be a new academic contribution to the scientific literature but you can often get to the point of having a useful software product much earlier so sort of you know Nietzsche book practices and materials and post like | 725 | 752 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=725s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | github and maybe on archive I think if you if you learn by reading the book it's really important to also work on a project at the same time to either choose some way of applying machine learning to an area that you're already interested in like if you're a field biologist and you want to deep-learning maybe you could use it to identify birds or if you don't have an | 752 | 774 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=752s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | idea for how you'd like to use machine learning in your own life you could pick something like making a tree view house numbers classifier where all the data sets are set up to make it very straightforward for you and that way you get to exercise all of the basic skills while you read the book or while you watch Coursera videos that explains the concepts to you so over the last couple | 774 | 795 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=774s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | years have also seen you do one will work on adversarial examples and tell us a bit about that yeah I think every searle examples are the beginning of new fields that I called machine learning security in the past we've seen computer security issues where attackers could fool a computer into running the run code that's called application level security and there's been attacks where | 795 | 822 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=795s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | people can fool a computer into believing that messages on a network come from somebody that is not actually who they says they say they are and that's called Network level security now we're starting to see that you can also fool machine learning algorithms into doing things they shouldn't even if the program running the machine learning algorithm is running the correct code | 822 | 845 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=822s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | even if the program running the machine learning algorithm knows who all the messages on the network really came from and I think it's important to build security into a new technology near the start of its development we found that it's very hard to build a working system first and then add security later so I am really excited about the idea that if we dive in and start anticipating | 845 | 872 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=845s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
pWAc9B2zJS4 | security problems with machine learning now we can make sure that these algorithms are secure from the start instead of trying to patch it in richer actively years later thank you fellas great there's a lot about your story that I thought was fascinating and that despite having known you for years I didn't actually know so thank you for sharing all that oh very welcome thank | 872 | 890 | https://www.youtube.com/watch?v=pWAc9B2zJS4&t=872s | Heroes of Deep Learning: Andrew Ng interviews Ian Goodfellow | |
JgvyzIkgxF0 | from the amazing results and vintage Atari games deep Minds victory with alphago stunning breakthroughs in robotic arm manipulation and even beating professional players at 1v1 dota the field of reinforcement learning has literally exploded in recent years ever since the impressive breakthrough on the imagenet classification challenge in 2012 the successes of supervised deep | 0 | 21 | https://www.youtube.com/watch?v=JgvyzIkgxF0&t=0s | An introduction to Reinforcement Learning | |
JgvyzIkgxF0 | learning have continued to pile up and people from many different backgrounds have started using deep neural nets to solve a wide range of new tasks including how to learn intelligent behavior in complex dynamic environments so in this episode I will give a general introduction into the field of reinforcement learning as well as an overview of the most challenging | 21 | 40 | https://www.youtube.com/watch?v=JgvyzIkgxF0&t=21s | An introduction to Reinforcement Learning | |
JgvyzIkgxF0 | problems that we're facing today if you're looking for a solid introduction into the field of deep reinforcement learning then this episode is exactly what you're looking for my name is Xander and welcome to archive insights [Laughter] [Music] [Music] so in its 2017 Peter Emil gave a very inspiring demo in front of a large audience of some of the brightest minds | 40 | 69 | https://www.youtube.com/watch?v=JgvyzIkgxF0&t=40s | An introduction to Reinforcement Learning | |
JgvyzIkgxF0 | in AI and machine learning so you showed this video where a robot is cleaning a living room bringing somebody a bottle of beer and basically doing a whole range of mundane tasks that robots in sci-fi movies can do without question and then at the end of the video peter revealed that the robots actions were actually entirely remote-controlled by a human operator and the takeaway from | 69 | 90 | https://www.youtube.com/watch?v=JgvyzIkgxF0&t=69s | An introduction to Reinforcement Learning |