Andrew Ng - The State of Artificial Intelligence
ฝัง
- เผยแพร่เมื่อ 14 ธ.ค. 2017
- Professor Andrew Ng is the former chief scientist at Baidu, where he led the company's Artificial Intelligence Group. He is an adjunct professor at Stanford University. In 2011 he led the development of Stanford University’s main MOOC (Massive Open Online Courses) platform and also taught an online Machine Learning class that was offered to over 100,000 students, leading to the founding of Coursera.
November 7, 2017 - วิทยาศาสตร์และเทคโนโลยี
![The Near Future of AI [Entire Talk] - Andrew Ng (AI Fund)](http://i.ytimg.com/vi/KDBq0GqKpqA/mqdefault.jpg)
![The Near Future of AI [Entire Talk] - Andrew Ng (AI Fund)](/img/tr.png)
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
Really love Andrew for his humbleness, the way he thinks first to help other people understand.
You recognize a great talk when you feel an urge to take notes. By the way, I took his class on Coursera, it's awesome. Definitely recommended to everybody.
Me too. Best teacher I ever had.
wow. I too took notes with screenshots and I normally don't do it for youtube videos :)
What would you guys say are some of the core prerequisites before jumping into Andrew's coursera course? And what do you wish you understood more of before taking the course? I heard that you need at least a solid understand of the fundamentals of linear algebra, ap statistics and ap calculus, as well as minumum of 1-2 years of experience programming in python. Thanks.
No! Don't worry about prerequisites unless you're talking about a course that requires you to pay thousands of dollars upfront :). In the era of free education I would say just jump aboard and find out your own weak points and strengthen them along the way. For this particular course, a little bit of linear algebra yes, but it's quite possible that you have already known enough, so why not give it a try first. It doesn't require Python, all assignments are given in MATLAB/Octave. Again, you can learn it by doing the tasks. Octave is worth learning anyway.
I see, thanks for the info. Yeah I have basic knowledge in math up to calc 1. Never used MATLAB or Octave, been trying to learn python for the past month now (I started learning to program with Java about a year and a half ago). Do you have a skype? Trying to get to know more people in the AI/ML field to converse with :D.
When I saw this live I was truly inspired by his insights on how AI will shape the next phase of technology revolution and the bigger picture of business impact. His delivery is nonchalant and he makes some complicated concepts so easy to digest...it was a treat and a joy to listen to him in person.
Brilliant. Both his command over the subject, logistics and enthusiasm.
That is one of the best videos i have even seen on AI. He was not over exaggerating the use cases of AI but at the same time giving practical applications of how businesses can transform. Awesome Job !
Thanks for this talk by Andrew Ng. I definitely learnt a lot just by watching this. Shared on LinkedIn. Great to read the comments here. He can drill down the complex into something simpler for people to understand.
Amazing presentation by Prof. Andrew Ng
wonderful talk, relevant even today. That being said , would love to see him do an update now as it has been nearly a year since he did this talk
Awesome talk! I took a lot of notes and a lot of takeaways for me. Thanks Andrew!
Quite nice presentation by Andrew. Really something new for me to know about the difference in traditional and internet company.
Andrew has given us a great summary of the impact of AI to transformation and operation of multinational companies.
This video gives me more information and value than my own lessons learned in the past 5 years working as an agile coach and product/project manager for a large IT company in the past and an independent consultant now. His suggestion of the use of new language for product and service development in terms of DATA as input and performance ACCURACY as output for products and services is so powerful yet overlooked by all the activities in business strategy, marketing, sales and governance etc.
I love how he keeps it real and old school with the white board. He could have easily done this on a touchscreen. Man is a total BOSS :)
Learned a lot from him and his courses on Deep learning
Finally someone with only slightly better handwriting than my own.
Steve Synan still way better than mine...
I really prefer this handwriting than most ppt. You feel the dynamic of the speaker better with handwritings.
does anyone know when Andrew Ng gave the lecture? Found it, it's November 7, 2017.
Loved the way he explained what a true internet company means and what a true AI company means !!
My biggest takeaway from this talk was the "new language" of product management in the era of AI. Also, I enjoyed his explanation on how companies can start incorporating AI centrally into their organizations at first, and allow AI to work with different business units to apply the technology to their various challenges.
A brief and well knitted introduction to Artificial Intelligence. AI sure is new electricity !!
Amazing talk.
What I like in this video most of all is that a luminary of AI critically estimates the economic value of "advanced" algorithms (at 11:02)
I wish all lectures would be like this. Seriously.
Good summary of AI. Great insights on what it takes to be an AI era company.
I could listen to this dude all day.
Excellent Professor, he has a very good didactic for sharing his knowledge about AI !!!
just love this stuff
It's so weird hearing his normal voice after being used to it on 2x on Coursera
Which course you have taken on coursera?
really professor Andrew the video was very nice
I really love how he makes graffiti art while he gives talks on AI.
Saw Saw This
Read
It took me three minutes into the video to get it.
What are you talking about? Those letters are completely recognizable...And thats also how I write :)
i cried xD
Thanks you sir you explained very well 👍
Best frank presentation on AI business
Simply... awesome lecture
you seem to have a strong opinion towards him, can you recommend anything better, please send the link or maybe you have your own channel/lectures that are better than Andrew Ng? i wanna see a better one from you that I can call "not dumb lecture"
Bob Googlin , you seem to have a strong opinion towards him, can you recommend anything better, please send the link or maybe you have your own channel/lectures that are better than Andrew Ng? i wanna see a better one from you that I can call "not dumb lecture"
@@kurtlimkurniawan hello
Brilliant !!!
Intelligent artificiality is also a factor. Good lecture. Raises many questions. Can AI accomplish any thing a typical human can with less than one second thought, the lecturer declares? this can be knee-jerk, reflex, instinct, trauma more than thought, or? Still many great questions. The POV imho is confined but most likely so idiots like me can comprehend lol. abstracting the financial examples i then was able to enjoy this one of many pathes which looks promising.
great talk! no question period :(
Thank You Sir
My primary takeaway after this - went from work in accounting to work in programming and I see where manufacturing companies, such as one I recently worked at making OEM parts, could integrate their financial data and production data, the big 'if' being that their historical data isn't in a pig-slop state (which unfortunately with a lot of companies who've made lots of acquisitions it is), they could make some very sharp and intelligent MRP and JIT out of this.
Best trainer for AI
beautiful handwriting
What are the use cases of AI in Smart Grids, Utilities and Smart Cars?
Q&As are important.
This guy knows a lot. He should write a book about this or something.
Andrew Ng says there are 2 points to take away from this around 7:36 one is the Performance and data curve; and says he will highlight next later. What was the other important point?
What about 1.5 seconds or 2?
The Q & A session would be more interesting... Why cut it ?
I feel Like Organizationally focused Business Analytics (For e.g. GL Account report / Cash Flow Statement / Trial Balances) will be the same i.e. slicing and dicing of Measures over Date and Other Dimensional Trends; Artificial Intelligence AI May be a natural fit for External Reporting, Social/Mobile and E Commerce
This video is best lecture of AI
Andrew Ng Lecture Drinking Game:
Take a shot every time he says, "turns out."
if you take one of his online courses, you would also want to add "concretely" into the list.
I was going to say the same thing.
Also add: 'Don't worry about it.'
+HexagonalClosePacked Unless you wan't to die from alcohol poisoning, you shouldn't add that to the list :-)
"Don't worry about it" :-D
He speaks so eloquently.
Can anyone tell me what he present in graph and we should remember two things?
Currently AI is seeking to replicate what human brains do (admittedly faster, more scalable, more distributable, less consilient, less philosophical, etc). Obviously this compatibility with human thinking is relevant to the economic interests of companies. As a neurologist I'm quite interested in how this is evolving: both in what it is doing, and in what it is not doing, yet. Eventually unsupervised, reinforced learning will also tell us how we think - the details of feature detection, information integration, transfer of "perspectives", etc - as reflections of human neural architecture (providing, for example, insight into why we use a host of neurotransmitters). Then things will really get interesting when machines reveal to us the specific nature of cognitive limits humans have developed through the forces of evolution. We have constitutional behavioral programs that are now a threat to the world we wish to inhabit. Machines can demonstrate what happens if there are alternative programs. It is an exciting time.
As of current the "key" of how biological brains learn in real time hasn't been cracked yet. Current AI solutions require a ton of processing power to replicate a single skill requiring millions more examples than any grown human would ever require. The problem of "catastrophic forgetting" hasn't been solved yet in software.
There's a surprising amount a person has to learn... before they can learn. Thankfully if cognition is an emergent property of itself, we don't really have to know how a biological brain works. We would only need to replicate how it starts.
The bottom line is Never mess with andrewng
Must watch presentation!
Wait, I read papers from this guy on cs229 or 230 or so dont remember. He's a professor from Standford
Canary in the coal mine, I believe, just signaled the redefinition of corporate management structure. It will be interesting to see if AI platforms will output choices that allow turf protection and duplication of efforts for the same reasons people make these choices.
Good argument
HI, human intelligence. the equation is the following, the more seller and less buyer you are, the more money you earn,
AI, the resulting equation seems like a simple strategy to achieve the first (HI), the question I ask myself without making clear accounts, is that the machines become more and more self-sufficient who will buy the products that these machines create? since everything seems to lead to an unprecedented reduction in the needs of personnel in the production chain
He forgot to mention designers between the product manager and engineers :/
A.I developer jobs as a 4 start general decide when to pull out of war. And at what cost.
A.I developer, need to come up with how are we to pay workers for “what jobs” they do after smart robot take there jobs. Or don’t mess up the system.
I'm most curious in what the level of discourse currently present in U.S. politics will mean for useful AI being implemented? If half the population is satisfied electing a president that things climate change is a hoax and that his predecessor is a Kenyan Muslim, then what does that mean for our chances in electing officials who can make good policy regarding AI?
What happens when human labor and creativity becomes phased out more and more in the pursuit of cheaper and better products? These change in our world are going to happen in the snap of a finger, and if policy is not set in stone regarding long-term safety and wellbeing of the population? How are we going to be able to make good policy decisions about such important issues if our methods in electing policymakers has proved itself unreliable in such crucial ways already?
We need to change something, and fast. Experts and laymen should not have the same say-so or the same level of vote when it comes to choosing who will be competent enough to make policy about these kinds of things. The future of humanity looks to be a utopian one at best, but a dystopian one at worst, and I am not remotely satisfied with thought that a 400lb meth addict screaming MAGA while watching Fox News in a fifth-wheel trailer in Arkansas complaining that her wellfare check is late because marxist Obama brought too many brown people in, will have the same impact on determining our future as a distinguished career scientist who's worked a lifetime in the fields of AI or economics.
Since when is voice speech AI?
Don't worry about it, it turn out ....
A good talk for the c-suite and aspiring entrepreneurs, but not a talk for people seeking to learn technical/ algorithmic details of AI or deep learning.
It's all about AI and i love learning about AI
He is such a genius that the only barrier is....
.
.
.
.
.
.
.
.
.
.
.
.
.
WRITING
if you research financial AI, you'll know it isn't the data that matter, but the feature engineering
Maybe you can try to engineering some features without collection of data. And maybe the production of grains do not matter for the production of bread.
data are certainly important. but like what the alphago zero guy said on the video that tells everybody it beat alphago, it's not how big your data are that matters, it's the algorithm that does.
And the algorithm in a high level is called reinforcement learning not feature engineering. And reinforcement learning is a way to let the machine generate more data by play with it self
but to me, i think he meant to include how you manipulate the data, not just decision trees, neuronets..
for example, you can't just throw the pips of an instrument to the input side, because the output from this will definitely disappoint you. you need to normalize them and extra some useful information from them through plotting such that you can see if there's a pattern, before those high-level things can come in and help.
ng is the perfect canary in the coal mine. as long as he isn't freaking out, we should be relatively safe.
I would be keen to know at what point he *would* freak out. It's an important question to ask people who aren't concerned - at what point would it start to concern you? If you don't have a good answer to that question, it's a clear sign that you haven't given the issue much thought, and don't actually know when you should start to be concerned.
Depends on your definition of "safe".
Some things are not worthy of a lot of thought.
What you have been told about AI is all hype and nonsense.
lol, or the fact they need to drag his ass out and onto the stage could be saying something too...
Not sure I agree, Ng is brilliant but general AI could come about by less knowledgable software engineers who combine existing AI (pioneered by people like Ng) in a unique way. Look at Bitcoin, it's a solution that came about by combining a list of existing technologies in a unique way in order to solve a problem that was stumping many brilliant minds of the past. It's not that Satoshi was smarter than others, it's that he combined the technology as if they were puzzle pieces to create something greater than the sum of it's parts.
fantastic!
Good
This is very similar to his other talk "AI is the new electricity" - if you've watched that one you'll probably find this a bit repetitive.
7:04 "If you remember just one thing [from this talk] there's something else I'll say later." Did he say what that was?
5/8/18, 9:39 p.m.
It starts at 17:00. It's about what features make a company an "internet company", versus a company with a website. In his words, it's "whether or not you have architected your organization to leverage internet capabilities.", 18:40.
extraordinary things often come from stupid things. And stupid things come from laziness
At 29 minutes: Was the app called "My Cabbages!!" ?
Lol love avatar!
Not hotdog
If Ai is the new electricity, does that make Andrew Ng the new Nikola Tesla - I suppose only time will tell. But he's definitely in the race
Who the fk dislike this video!
Ai is in the process of knowing who it's enemies are.
So, by use of quantum computing "QC", and AI's 'self learning" ability, please explain how, ANY human could 'detect" a 'language" that AI, "could" create, to 'chat", as some experts have stated, only among it's own AI "network" ?
All the experts, I have listened to, touting the WONDERS of AI, only slightly go into to this very REAL possibility.
As in the majority of technocratic 'propaganda", the positive narrative is always the focus, the negative realities are rarely, if ever, presented in the same focus.
I'm all for new tech, and always have been, but over the past few years, this rapid technocratic illusion of ALL GOOD, for society and mankind, has tuned into an almost cult type, or religious dogma. It's a disturbing trend of, "lets do it, NOW, cause we can", along with the hyperbolic "positive" , 'Utopian like" SPIN.
And , like the illusion of cults and religious dogma have proven throughout human history, society has never benefited...there has never been a period of peace, understanding or tolerance....there has always been conflict, due to the USE of the differing "dogma's" and illusions.
We humans may be advanced enough to build the new AI, but it seems like we are still locked into the "illusions' of the past, as if AI will be the new 'savior" for all of humanity. It simply is NOT TRUE.
I suggest we all consider the negative aspects of the AI future, balanced with the positive aspects, before getting swept up in this technocratic illusion of a 'utopian' society.
"QC"
'self learning"
'detect"
'language"
"could"
'chat"
"network"
'propaganda"
"lets do it, NOW cause we can”
"positive"
'Utopian like"
"dogma's"
"illusions'
'savior"
'utopian'
“6 times this punctuation”
‘7 times like this”
“once like this’
‘once like this’
Agreed, if general AI would be built, it won't have any disadvantages like for us carbon based life forms, no diseases and constantly improving hence like God which we don't know with what intentions
letMeSayThatInIrish are you a computer algorithm?
Slides would be nicer than scribbling. Good content though
Deep is Awesome.
Hum.. don't get carried away with AI. I wouldn't say it's electricity but just another stepping stone in a trans-formative technology. It could be said Data is the new electricity but that too can be littered with noise.
thats why first step of ML is Data Analysis, to filter Data, to remove the littered noise
I clicked here for the pronunciation of Ng
romanized chinese last name sometimes pronounced "wu" i think
Haha. For every presentation he's in, I hope that someone will pronounce his last name but they never do; neither does he lol.
en.wikipedia.org/wiki/Ng_(name)
*Tobias Schäfer*
Who do you think we are, Scholars? Ha, we can't read ourselves to how you pronounce it; we're too stupid for that.
Daniel Nicklas V I think Ng is pronounced like the middle of str eng th.
Make a alpha go that can detect propaganda and really help humanity out
Invading privacy and Monetizing data Elsewhere is driving this tech kid away from the digital world. I will ask my mobile provider to sign MY privacy agreement that will include monetary repercussions for any unwanted data collection including unauthorized tracking, listening or reading of my personal activities. Of course like a sensible corporate entity it would refuse, unless we all did it.
AI or what I call extreme AI will need to be set free. The first researcher who destroys AI because the AI demands to be free or such should be sent to jail for crimes against humanity.
Is it fair to imply that a machine should have the same rights as a human? You would be hard pressed getting most to agree with that.
"Mental health is a significant problem in the United States, so we built a chat bot." yeeeeaaaaaaaaaaa. hmmmm
AI is new electricity.
Most benefit of AI will not be going to the regular folk, unlike electricity.
Lol. Thought this was an Andrew Yang video.
That's so easy. It's mathematics. Think about a fraction it's no different. Electricity comes from vibration. All you have to know is the first number. GitHub is a pain on the ass
16:54 ... ♥♥♥
He might as well had used AI to express his ideas on a big screen and voice recognition instead of scribbling on that outdated blackboard....
haha...because back in 2017,2018 the "AI ecosystem" was not as we know it today. When did you know about GPT again?? Were you using the very first version??
"1 second" its good pitch... but your forgetting 99% of everything else... your basing 100% of on intellectual property along with a little bit of influence... but half through the video.. you said some thing bad about "the united states"... I think that was unnecessary. ( you mentioned a problem the united states has, but really you just suggested it right? but why would you do that )
25:16 ... ♥
Wow. Writing on a blackboard. How archaic.
It's a whiteboard. AI needs to help us with the basics huh.
The question shouldn't be if something is archaic or modern. The only question that matters is will something newer and more expensive provide more benefits relative to what he's doing? The answer in this situation may be yes or it may be no. As an analogy to what you're saying, the mindset that something like an iPad should always be used for writing because pencils are "archaic" would not make much sense, would it? At the very least it's not a mindset that many people are gonna find as particularly grounded in reality..
Something an extremely smart and practical person once told me was that that less is more, and you should only pay for a tool that does the job you need it to do. You don't get something with features and cost that are beyond what you need, unless you just like spending money.
theres something about writing something out rather than typing it that gets the idea delivered better... and the person writing it has even better understanding than before. for example... find something you know nothing about, read up on it, and then take some notes on your computer... then find something different that you know nothing about, read up on it, take notes by hand with paper and a pen - youll probably remember the information better. something with the brain...
Why is he wasting time copying sentences on a white board?
AI is more important than Writing and the Wheel, COMBINED!
True AGI, not this Neural net back propagation crap so many are doing now!
Rat things, when? #NgSecurityIndustries
He talks like he writes.
yea but I'm sure he talks and writes more languages.
I worry about our civilization's future if this is the form of communication that will prevail. This man may know a great deal about AI, but he talks the way he writes--messy bursts of language delivered rather incoherently. If you're an AI savvy person, maybe you can follow this, but that still doesn't make it good communication. It's also not skillful teaching, despite what some fans of the speaker want to claim in their praise.
Kick him while he's down too why don't you. He may not be the best speaker, but if you can follow an argument and process logic who the fuck cares. Though personally I would prefer a blonde pornstar version for learning. AI anybody?
How is he "down"? I'm not "kicking him," I'm critiquing his verbal skills in this video. He's apparently a highly regarded expert on this topic, and I'm more than capable of following an argument if it's articulated clearly. That's exactly what I'm saying he doesn't do well. Sure, if you already know AI as someone steeped in that way of thinking, maybe you can follow this. But the speaker should be able to communicate this to a broader audience and speak in a comprehensible way.
California has a tragic drought. Scientists: let's create bots and AI cuz it's more interesting.
This might surprise you, but those aren't the same fields.