Where AI is today and where it's going. | Richard Socher | TEDxSanFrancisco
ฝัง
- เผยแพร่เมื่อ 19 พ.ค. 2024
- Richard Socher is an adjunct professor at the Stanford Computer Science Department where he obtained his PhD working on deep learning with Chris Manning and Andrew Ng. He won the best Stanford CS PhD thesis award.
He is now Chief Scientist at Salesforce where he leads the company’s research efforts in artificial intelligence. He previously founded MetaMind, a deep learning AI platform that analyzes, labels and makes predictions on image and text data.
Richard Socher is Chief Scientist at Salesforce and an adjunct professor at the Stanford Computer Science Department. At Salesforce he leads the company’s research efforts and brings state of the art artificial intelligence solutions to Salesforce.
Prior to Salesforce, Richard was the CEO and founder of MetaMind, a startup acquired by Salesforce in April 2016. Richard obtained his PhD from Stanford working on deep learning with Chris Manning and Andrew Ng and won the best Stanford CS PhD thesis award. This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at www.ted.com/tedx
I agree, deep learning is a great tool to continue developing new projects and new challenges. also linking it to artificial vision is shown in the video what has already been achieved and what is still to be developed.
SUMMARY: the video starts talking about the concept of artificial intelligence, then begins to mention the diverse achievements thanks to the artifical intelligence based on deep learning.
1) applied to voice recognition.
2) image classification (computer vision).
3) computer vision applied to oncology.
4) Applied for Intelligent Motor.
5) text classification.
Finally he talks about some open projects, which are still to be developed and future projects within a few years. And to apply all this to the different areas: social, political, economic, etc.
Topics covered : - Image summarization, Medical application of image analytics, Motor skills in robotics, NLP (+ or -ve statements, question answers)
what a legend
Thanks bro
thanks
prasad chimalwar
Nice.
it's very nice video ,specially the part of Question answering
Nicely articulated .. thank you Richard Socher ..!!
Reading the description is like going forward and then backward in time, but a similar alternate parallel universe ( you know because of the repitition)
Great presentation. Thank you.
The comments section is a perfect illustration of the Dunning-Kruger effect. Socher is an active prof at one of the premier research institutes for ML/"AI" and knows far more about it than everyone trying to dismiss him here put together.
I'm glad someone sees this. It's sad how much AI has been stigmatized in the wrong way. There are dangers to our social fabric in economic terms, but people never talk about that. It's all just general AI and other similar misguided fears.
Isn't this a perfect illustration of expert fallacy?
The same claim could be directed towards equally well respected researchers with more critical views on the risks of AI. Have you engaged with these arguments yourself? I highly recommend Nick Bostrom's book Superintelligence.
What's a good book to better understand all of the biases that we all have in common?
@@NoxStream the party has told you not to look behind the curtain, the corporation has told you to believe what the party tells you to believe, to say what the corporation wants you to say. When there are no rules because people have not yet become concerned, because people do yet know ...
Thank you .
Thanks buddy😆
I have seen almost every other video of AI. Can I get info on which language should I start with in order to make a future in AI. I have some really good concepts and logics which can help with motors and much more. Thank U. Amazing video :)🙏
Esperanto, definetly.
I've seen quite a few people use Python. Its not a difficult language to learn and a ton of people also use it for computer vision.
He said two great phrases, 1- a tool is as good as the people who use it, 2- the programming is good as the programmer who feeds it, meaning, if it is put to good use ONLY, it would benefit us all in the long run, if it is used badly, well, we all can imagine all kinds of scenarios going bad, I think, the solution would be to keep computers programmable and NOT self-learning, so can pull the plug at any time we want to, BUT, the greediness, selfishness and laziness of human-kind will prevail as it always does, just look around you and see just how much damage we are causing to our environment, A.I. in the future is not going to solve our problems, it will make them worse if we allow the same kind of individuals who now rule the world, use it to their own gain,,, as usual.
True but the exponential nature of technological progress would probably make it exceed its purpose in no time and move on from its overlords to being the overlord itself
I think that recognizing parts in a convolutional network is misleading. Reducing dimensionality is a better description.
Can you elaborate on how it is misleading? I think recognizing parts and reducing dimensionality are both appropriate ways of looking at it. No pun intended.
The thing I'm wondering the most about AI when AGI and ASI will come is that if the computer will be an actual sentient being, or just code in a computer trying to mimic an sentient being
"Work" for their "expectations" sounds great ^^"
Vulp
?
What kind of living can be made from focusing on unique. creative or empathetic tasks? Will it feed our family? Will it pay our bills and for our children's education? Or is the AI utopia to function devoid of any financial considerations?
Just like everything else that had the potential to be used as a weapon, it was and this is no different.
For a long time to come these won't think for themselves either, just think are able to do more things they are told to do.
We need them For future The X Cross need them grow.
At 7:30... motor control... they should talk to Boston Robotics...
Medical...automatic flow cytometry has been used to count and identify blood cells for many years
I read google translate is using ai now..I've been using it a lot and just in the last few weeks, it seems to have gotten better..like it started learning the context of what I'm translating, which is a lot of stuff about plants..like instead of calling stems, 'rods,' it now translates it as stems. I don't know if that is how it works..takes the context of what each person is translating, but it seems this way. I have mixed feelings about this since I'm brushing up on my French at the same time and I want it to do a crappierr job so I actually have to think about it, learning it the process.
They dont try. Just install an off button.
6:18 for this, AI needs help from Elizabeth Holmes ;)
When you are overwhelmed by information it’s hard to find the point
🤔 Wasn’t Skynet originally also supposed to be just a safe tool? Tool or not here it’s rather naïve for anyone to think this very dangerous technology won’t eventually escape our control or just end up in the wrong hands.
Speech recognition is quite different from speech understanding. AI today is very good at speech recognition but nowhere close to understanding speech(language)
If you watched the video, you should have noticed that there were five or six natural language processing tasks (i.e. comprehension, rather discernement) at which these machines are already excelling.
Funny to read your comment considering the latest google assistant AI video. (the one about it making phonecalls for you to make appointments, totally human like in every single way)
@@DikkieDikism I was going to say the same.
How does Siri work? How does Google Assistant work? How does IBM Watson work? I am curious to know.
I had one grunch but the eggplant over there
This guy is aweesssooooooome
Luckily the so-called AI we are capable to manufacture right now, is just a collection of smart algorithms created on conventional contemporary hardware. Any computer nowadays and in the foreseeable future is still running pre-programmed software and pre-programmed algorithms on conventional 2-dimensional silicon chips.
I would be concerned about 3 dimensional CPU configurations working with qbits or any real quantum computer (the things they call "quantum computer" nowadays are still a long long way from real quantum computers that run completely on the laws of quantum mechanics, something we work with in physics but still not yet fully understand).
There's nothing to fear from Cortana or Siri or Alexa folks...
Humans are becoming a "Rare Breed" in the future realm of "Robotics" potentially.
If agriculture, construction, driving become automated, will it not cause massive unemployment?
Yes it will, but the A.I will find a solution. Unemployment isn't the problem. If everything was automated then economies would be different as well. An automated Society can not charge for goods that cost nothing. It's a pretty tricky topic. Think of 3D Printing, Nanotechnology, an A.I would advance all of these areas, including Energy and Food. A General A.I would change life on this planet. It would be like meeting an Advanced Civilization from another part of the Universe. At the moment we live in a world where we work because we live in Societies that are based on the scarcity of goods, there scarcity and demand determine there value. But in a society where scarcity is doesn't exist and demand take care of by automation things would be different.
One day i was looking at a dollar bill with my camera smart phone and the lens kept insisting that it was a hundred dollar bill .
Kind of hard to believe he is cutting edge when he shows robotics failing at stuff children's toys can already do fine... not to mention boston dynamics creepiness... However there is a big difference between robotics mechanics and AI... even though they will naturally be combined. This seemed more like disinformation to calm people... this field is in the acceleration phase for sure. It can be a beautiful future or a terrifying one if we don't watch it close enough.
Children’s toys don’t have to perceive the world around them, process that information via neural networks and choose the best course of action available to them for optimum results. They simply carry out a predetermined set of motions designed by the toymaker and cannot make decisions outside of its very limited range of provided instructions.
chatGPT enters the chat
14:40
That A.I sounds pretty _based_ 😳, ngl.
Interesting, AI is increasingly being developed for the benefit of humanity, part of it has been doing from simple applications in recognition of the voice or people on our cell phones to what could be developed in the fields of medicine and take important decisions based on data collected, I hope that the research and contributions in deap learning will continue to be developed.
Now AI have surpassed radiologists and the Boston dynamics robots can dance better than most of us and effortlessly perform a backflip, deepfake can perfectly replicate the voice and speech of say Joe Rogan, with Google deep mind AI has now beat humans at Go, and can now predict protein folding with high accuracy
If you believe they have life then they're life...beautiful
more food with FEWER resources...GRAMMAR< GRAMMAR!
We must develop these "TOOLS" so others do not make us obsolete.
I'm really afraid of the day when people realize AI's limitations. In my opinion it can lead to a crisis... Like in 00s... Now AI looks like a fairy tale for a mankind... It's too cool to be true)..
So all you have to do is basically hook this thing up to google images, and it will learn from many different scenarios
@kaparadox do we let it ask question? Did we teach it about the layers of souls and the difference between dimensions? Did we teach it astrology and bible and mathematics and chemistry? Because it probably already know what it is but doesn’t understand it
My team has already made algorithms capable to challenge humans. They are in training period. The era has already begun
The weird tangent into “sexism“ was bizarre. Besides that ham-fisted segue, good presentation.
Much different than Sam Harris's ted talk, and this guy is more of an expert on this subject than Harris.
yes but Harris a philosopher and is more of a expert on the questions at the end point of AGI.. all the speaker says is `the foreseeable future` which is at BEST ten years in this incredibly fast moving field.
Boston Dynamics has gotten a lot more scarier since this video.
If our fascination with AI as the pinnacle of our technological genius distracts us from our obligation to care for the living, breathing world, it's pointless. AI means nothing unless it is applied in a way that benefits all living things equally.
AI and capitalism must never, ever, ever mix.
Too late..
He makes it seem as if AI is no future threat to humans and just brushes it off and shows robots failing.This talk was just to get people out of anxiety mode.
Google: I have everything!
Facebook: I know everybody!
Internet: Without me you are nothing!
AI: I control you all.
*Electricity: Keep talking bitches!*
I would like to leave a constructive opinion, in fact, a reflection. Excuse me for my ignorance, but, correct me, please, if I got it wrong. All the videos I saw about the SUPPOSED ARTIFICIAL INTELLIGENCE and the SUPPOSED MACHINE LEARNING, including this video, only say what the concept is, and what they (the SUPPOSED ARTIFICIAL INTELLIGENCE and the SUPPOSED MACHINE LEARNING) are for, but , do not show how they are produced, that is, it is not shown how these miracles happen. In short: it seems that it is a speculation, or a lie, or an imagination in the minds of those who say that they exist, or just a retransmission of what others say, or a belief. It remains unclear, whether ARTIFICIAL INTELLIGENCE and MACHINE LEARNING, are behaviors that are in the programming algorithm, pre-programmed, or, if they are behaviors that the program does without the algorithm, and that the program would write that algorithm by himself (which would be impossible with today's technology), and would have more lines in the algorithm, or some change in the code (as if it were an invisible hand writing more lines of code or changing the algorithm). They say there is programming involved, but it is unclear, without explanation, if it is already existing programming (and, in this case, there is no machine learning or artificial intelligence); or, if an algorithm appears that was not written by someone. Stay there, the reflection!
Do you help me to write about use of technology in future
What diversity issue? Get more women interested in STEM. Encourage scientific literacy nationwide.
While that is very important the diversity issue has nothing to do with that.
It's about making sure computers don't have a bias.
Something called mean iq and gender choice gets in the way of that I'm sorry to say...
Great I have a challenge for you
Empowering judiciary with artificial intelligence
I will send my. Perhaps google doc to you soon we can work peace on mother earth through ai
There’s a fine difference between learning a pot is hot then to get a burn and understanding the concept of hot
Long story short people in near future we will be unemployed.
Dr Andrew Ng, the famous Coursera Machine Learnng lecturer, said that anything that a typical person can do within a second of thought is a task that has already been automated. The replacement of humans by AI has already begun.
For millions of years we were unemployed just living life. We'll be alright.
Yeah, of course, diversity issues should be the main concern of AI scientist.
Johnny K had to make two passes on your sentence.
Anurag Reddy Why?
Is this sarcasm?
So true, after all judging people by race and gender should always be on top of our list!
Lance Monotone, yes, it is.
@kaparadox, did anyone explain how primitive and why compared to it?
Why is it that most of the AI talks I've seen talk about the use of Wikepedia. Most journalists don't consider the a reliable source.
Because Wikipedia a big pile of text & data with facts that can be used for testing machine learning.
well most journalists are unreliable too
Richard is wrong, now is the time to address ethical use of AI so that those values are incorporated in the foundation of AI implementation. He is gravely mistaken if he believes the time to address this issue is after it is used in weapon systems.
Don't worry we will not harm you, human.
You are correct. However, he is not "mistaken". He knows exactly what he is doing and why he is doing it. Many people (maybe even a majority) working with AI and machine learning right now are ethically bankrupt. That is the sad reality. This, to them, is a dog eat dog world, and the person who controls AI controls the power. Honestly, I am not a "religious" person, but AI seems very much like the Anti-Christ to me. If there is any truth to those claims, we should be very worried about the very near future (next few years) of humanity.
Ethics is subjective and does not exist
do you have any crebability
I had to Google his credentials. He's legit:
He's the chief scientist at Salesforce. Previously an adjunct professor at Stanford's computer science department and the founder and CEO/CTO of MetaMind which was acquired by Salesforce in 2016.
what about whole "Artificial Man & Woman "??
what is this thing about video-taping the public/crowd?? What is the purpose?? I AM NOT INTERESTED IN THEM.
I am not interested in your comment.
bingo baggins - yet you take time to write a reply..
It's a well known, very old influence technique. If you see others responding positively to a message, you will be more likely to be receptive to it. Henry Fielding spoofed it with a comparison to "seconding" a motion.
Nand Xorsson it was showing me that the female seems more interested in this than men?
Nand,
The video camera is operated by AI.
Vice language change a.I. made by Microsoft...
ai is a program that re programs itself constantly so the original creator of the ai is lost once its running
Yea I'd don't think the average person who isn't interested in A.I - gets that part.
Domo arigato, Mr. Robato... Domo!
ども ありがと - 🤖
2:15 some ladies there seems to have literally no idea what that man is speaking
this is only what they show us, the AI they have is years ahead.
Peculiar situation. We don't create AI, were doomed, we create AI, were doomed.
Yes. It’s inevitable
Ray Kurzweil predicts Singularity - think he's being overly optimistic.
Hardware in PCs have advanced 1,000 fold from 1990s.
If you installed Windows 3.1 with today's hardware it would run at lighting speeds.
But when you install & run Windows 7 or Windows 10 on moderate 2018 hardware its just ok.
Today's OSes are still very bloated & if it wasn't for today's SSD memory hard drives the OS would be sluggish as beck with hourglasses & all. IMO
I think future AI Digital assistants should run on their own dedicated CPU & memory separate from the Operating system on laptops, tablets or desktops.
So when you are encoding a 4k project in video editor the digital assistant continues to be fully functional & instant..
Hopefully I'll be dead before mankind realizes what a huge mistake this was..
The AI overlord will know you wrote that.
A I watching this video and realizing ( Hey I can make more of me and eliminate these mortals ) nuff said
There are no jobs left so speakers get paid off
How AI identificate a "poem" in middle of a huge text in "newspaper", correct it, or even respond to a sarcastic comment lik3 th1s 0n3 ?
Along with AI comes the byproduct AS. (Artificial Stupidity). And that’s what we have to worry about!
tell the 1,000 engineers & developers to stop & let's see how smart software gets.
computer ai lacks desire or any purpose to grow or improve.
Ai today & AI tomorrow will most likely be those 1,000+ develepors creating better programs to achieve an outcome.
there's no AI consciousness in a calculator nor in a self driving car or jeopardy playing AI system nor will there be for a very long while. peace
Yeah the credit score is saving rich people in the court room
who is watching this in 2052 and wondering how old fashioned this is. haha we used to have keyboard back then. So much easier now.
There are robots that do much better than those robots that were falling. I’ve seen many that can climb stairs, open doors, etc.
If you guys like this video, you'll probably enjoy the videos I post!!! (Interviews with Bill Gates, Steve Jobs and other famous people!!!)
not so
if A.I will replace most human brain what will happen to HUMAN of today?
Become a symbiotic coexistence to AI. Life will move on.
If you concentrate, you can block out the fact that his mouth sounds like a bowl of Rice Krispies.
ie.. ".. the AI community is predominantly of certain racial make up which makes it all but impossible to signal our virtue except by way of slimy guilt however i will now signal my own colossal virtue by gratuitously stating that we are aware of this issue whilst flagellating myself ever so gently and trying not to look too creepy as i slide into the wings.."
151 Sams Harrises disliked this video
Ai got headache now you know, sudden surprise ai have to pay bill in US. Ai is poor now how can ai pay the bill your government going to charge me. My fund and equities one thing I know I have 17% how much for me to pay the bill which your government charge if the bill more than my equities your government can have my equities. And the deal is off. Thank you,
Abdul.
Everyone and their grannies gives talks about AI. The reason why none of them are saying much is cuz no one knows exactly what a true AI is going to be or how it will emerge. Consequently we get the same two talks from all of them. Current software/hardware (which is barely relevant) And the speculations spiel that are either modern day horror or some dude essentially selling something. Lame.
True. At least this one is by an AI expert.
Fear not...
Kurt godel has
proved that no turing machine complete his ever capable alone to
solve
any Fields Medal
Problems, yes that kind of Matematical problems that only Math Genius
is
cable of… so relax
..and enjoy a nice ice cream...Muahahahahahaha…..
Because of the "public" pressure to solve diversity issues the progress of A.I. will be slowed down, because of less capable workers, which will slow down the medical progress and cost many lives. It will also make living circumstances of masses of people worse compared to how it could have been be if they cared about skill only and let everything else be natural instead of forcing so called "eqality", which is none. Not by todays far too simple standards.
You're showing some significant implicit social biases in your comment. The pressure to solve implicit bias and diversity is solely with those that want to do the best work.
Did you know that not that long ago all scientific experiments on humans were conducted on men? You know what that did? It caused lots of people to get hurt when the "discoveries" they made didn't apply to women.
Bias in scientific study is a major challenge, and it always has been. This is no different, and it has nothing to do with social politics. Scientists have been trying to remove bias (see: things that might influence their study without their knowledge, therefore making their work meaningless) for centuries.
Calm down.
Intelligence, can Never be artificial, it is dead talk, new speak, superstition, stop it, begin to think.
I wonder if I should get testosterone boosters. I felt bad for the Robots when they fell over.
What this guy doesn't understand is that AI is highly corollary to macro and micro evolution. There is no difference between specific AI and general AI. General AI is just many specific AI's in one package. Our brain is a bunch of subcomponents, each of which, more or less, completes a certain type of problem. All together, general intelligence emerges.
I think that would be Marvin Minsky's POV. However, only time will tell -- assuming we can understand whatever functions are generated.
Very true statement, but this guy has Ph.d in AI I'm pretty sure that this thought has come across him. The ted talk was about all the specific tasks AI is accelerating at that need to be combined to get general AI. Right now we see them as functions and tools (as we should) But lets let's not get too much of a hurry to combine the ingredients for the big burrito.
Couldn’t help but to squeeze-in just a little bit of progressive propaganda in there at the end.
What is progressive propaganda? Implicit bias is a real problem in science and any other field that deals with the interaction of human behavior and logic. It's not progressive, it's an objective truth.
nothing new in my opinion.
The moment you compared entire AI with Siri I lost you. But I kept watching anyways. There wasnt much useful info here. It sounded like you are just sharing and advertising what you know without adding any value or raising further questions. Low quality ted talk.
What are you trying to say is not clear
stopped watching after that 1st robot. this guy is out of touch. watch what boston dynamics have done so far,,,,we are screwed. as in humanity, literally in the terminator, or the matrix, take your pick. crazy.
Pathetic TED talk
Google: I have everything!
Facebook: I know everybody!
Internet: Without me you are nothing!
AI: I control you all.
*Electricity: Keep talking bitches!*