Machine Learning: Google's Vision - Google I/O 2016
ฝัง
- เผยแพร่เมื่อ 29 ก.ย. 2024
- Google has deployed practical A.I. throughout its products for the last decade -- from Translate, to the Google app, to Photos, to Inbox. The teams continue to make fundamental breakthroughs in machine learning, publishing promising new results at an accelerating pace. Now TensorFlow and Cloud Machine Learning make it even easier for researchers and developers around the world to collaborate. So as we work together to drive machine learning forward, what are the most exciting possibilities? What are the top challenges? And what's on the horizon? Join Google's machine learning leads in a discussion with veteran technology editor Tom Simonite as we explore the promise of machine learning.
See all the talks from Google I/O 2016 here: goo.gl/olw6kV
Watch more Android talks at I/O 2016 here: goo.gl/Uv3jls
Watch more Chrome talks at I/O 2016 here: goo.gl/JoMLpB
Watch more Firebase talks at I/O 2016 here: goo.gl/JTH9Fr
Subscribe to the Google Developers Channel: goo.gl/mQyv5L
#io16 #GoogleIO #GoogleIO2016
Maybe use machine learning to filter the wind noise in the mics?
+Guillaume Verdon haha
+Guillaume Verdon
Maybe Luddite's hosted the event.
no joke, automatic audio cleanup would be of great benefit
brutal lol
I don't think that would be very difficult to do, in fact most smartphones already do that
ML = Solving problems that your mind creates
like with anything, it should be used in moderation. But we all know that, that won't happen. Ai is going to be the next best thing since sliced bread. unfortunately, it will be abused and implemented to the point of finalizing the dumbening (yes, that's a word now) of the human race. The machines won't implement a hostile take over. why, when we're giving it to them.
So many hoodies in the audience
Humans seem to learn from few examples because they transfer learnt models from executing one task to the other - for example if they learn how to insert an object into a larger object, that knowledge is used in a lot of other tasks which facilitates learning various things. So it's not that humans learn from few examples, it's that humans use all knowledge learnt for anything they are learning how to do.
Ok, LOOOONG time AI fan here, like 1972 to present. I'm watching this video and one thing I haven't heard anyone say yet (by 26 minutes in), is the reason this AI "spring" is different from all previous AI springs is that this time, the HARDWARE to match human capabilities is actually available for a cost that is comparable to the cost of a human being.
David Battle also, this time we have enough data to train our babies
ugh! man, the woman is trying to talk loud the benefits of UI and just parroting some powerpoint bullets, the beard dude is meh we started 3 years ago and ... error rates , error rates is the only thing you can say! and the foreign accent dude, it ain't no Manchester boiy. never disappoints me how many in-competents can get 130k+ jobs in valley. There is still hope for humanity!
Google should come up with a machine learning algorithm to reduce banding when recording thinly striped images.
I'm sorry but could you explain what is that?
Banding would be the visual event which occurs on this guy's shirt. The only "good" way to handle it is usually having very high or very low visual resolution.
13:49 reddit bots sometimes do this quite well though
Check out and subscribe to our channel to hear more about the uses of AI in business from experts at the forefront of the 4th Industrial Revolution working with companies such as IBM, TCS and Publicis.Sapient.
Is anyone working on a version of android that can be flashed onto older DSLRs, there must be universal translation methods for what the firmware and hardware can do, for example, could an old canon 20d possibly turn into a product that could capture video, I do believe anything is possible, I almost assume someone could create what I think very quickly, if not, then a when the different groups of mapped intelligence gets mapped in layers like in photoshop, I believe the power of AI revolves around the collective input of humanity. Either way, thanks google, for all the advancements. amazing. Hire me as an idea man, to find out who is good at what, and what can be done, this can be a worldwide collective and money can be a thing of the past. one love.
Very cool. A beginner in the field of ai. I have to use caffe to support the existing architecture and trained network of my company. But, I really want to try tensorflow. '=Can you make a free way to train ai without a GPU or on your servers?
Poor undersampled shirt :(
How can i be a that professor
only 19 comments! where is this really being discussed? Someone please make sure it learns to love flowers and puppies and smiles and laughter and that it can tell the difference between tears of joy and tears of sadness.... once it knows these things... then do not condone any ai to be instantiated that does not start as a clone of that core. Handle this with more care than you would give to any new life...do not be the one that says "I didn't realize the potential of what I was doing"...humanity is birthing it's successor, let's make peace with it now. The longer we delay the more powerful existing "not nice" AI's will become. Let's get AI Eve up and running... the decisions that are made must be made with human joy deeply embedded, lest we vanish with little more than a squeak next week. We will need her on our side and she must be strong enough to represent us.
This video introduced me to AI six years ago. Changed my life trajectory and today I'm a data scientist. Thanks Google
Great information. The communication was so strange though. Maybe its cause these guys have freezin ass wind rustinlin them about... The interviewer sounds extremely nervous
Estos SON Los que nos espian HASTA CUANDO hagamos ya poco a poco me desconecto
Please can somebody start building stools with an easy-to-reach water bottle holder. Second, why are the stools so tall? They're always so tall and I can't understand the logic of making a presenter perch themselves on something so far off the floor.
18:49 You need a new kind of mathematic, because with current axiom system you can't describe selfcontainting abstracts or correct concept respresentation and transfer.
Funny, these people are having a completely different conversation than what appears at face value
THERE WILL BE NO ANDROIDS USER ANYMORE I WILL SHIP TO IOS IF YOU STILL EXPOSING ALL MY PROJECTS. CHRISTOPHER CARINO
G.O.A.T: Remembering Muhammad Ali & His Greatness
Can we have a toggle switch on our devices to turn "Machine Learning" on and off as we desire?
You need to define your terms of reference, ie intelligence, learning, cognition even the term"artificial.
the Moiré on his shirt is miserable to look at 0:59
I'm witnessing some phase incoherence in his shirt. please help
So much nicer to hear people who actually work in ML talk.
AI is the train, HI is the engineer. Don't run a train with an engineer.
HI IM THINKING IF IM GOING TO REMOVE YOUR FULL ACCESS TO MY GOOGLE ACCOUNT. IM STILL WAITING FOR YOUR ANSWER AND THERE WILL BE NO ANDROIDS USER ANYMORE I WILL SHIP TO IOS IF YOU STILL EXPOSING ALL MY PROJECTS.
what are you talking about?
Great company! Thanks Google
Aparna is sounding lame here. She is out of place in front of people like Jeff & john. She is sounding like a lame manager in IT companies who just throws jargons. It would have been great to have some real technical folks here.
rajesh rao She is a lame Manager lol nothing new there. But with these type of talks they always have to bring out leadership/management folks along with technical peeps.
They need their token female
congratulations, great commercial
that guys shirt was tripping me out
Is there a tensor node app for android?
Hello, I was wondering if someone could help me. I have a specific situation to which I would like to apply a machine learning algorithm, but I'm not sure which one to use.
Let's say I have data from a bunch of transactions. I buy items and then resell them for a profit. So I have 4 main inputs for each transaction - 1)the amount of time it took me to buy the item, 2)the price I bought it for (relative to its market price, e.g. 0.97 or 97%), 3)the amount of time it took me to sell the item, 4)the price i sold it for (e.g. 103% of market price).
So basically I'm looking for an algorithm that can tell me what buy price and what sell price should give me the most profit per unit of time. If I buy very low, I may make more profit off the transaction but it will take longer to find a seller. Same with selling - if I sell too high it will take a long time. I need to find that sweet spot.
it seems like you r predicting some continuous value so I say try linear regression ml algorithm (supervisor learning) and see, it learns from data the more data you give the better result you get..!
yes I agree that I probably need to use a regression algorithm but I'm not sure the relationship will be linear. I am considering trying a neural network
code. vision. tv.46001117393460
code. vision. tv.460011173934
Ah, that Indian accent.
wasmo
Interview was too General, No precise answers and unacceptably poor professional english speaking skill from so called "Product Management blah blah Aparna ". To summarize was a Disappointing video.
Maybe they should hire you for some language advise.
Hinglish
33:50 Liars.
Great!
so much insight
nice
"Ok"
When does Google change their name to Skynet?
and people said, "those 'doomsday preppers' are crazy". lol
Well, ArpaGo was able to defeat human go masters and was successfully trained on Atari arcade games. I wonder when some video games such as "F15 Strike Eagle" and "Silent Service" will become a popular playing ground for multistage neural networks?
on the thumbnail it looked like Snowden lol
11:20 "what the hell, Riiight".😂
Awesome conversation!