Meta presents advances in AI robot-human interaction

The Manila Times

Meta presents advances in AI robot-human interaction  A robot capable of finding and putting away objects that are lying around: this is the prototype presented by Meta at its Paris premises. 'Thanks to more powerful artificial intelligence systems, we are able to interact with robots in a much more natural way,' explains researcher Akshara Rai, while Meta's scientific director of AI, Yann LeCun, advocates open source development in the field.   AFP VIDEO  Subscribe to The Manila Times Channel - https://tmt.ph/YTSubscribe  Visit our website at https://www.manilatimes.net    Follow us:  Facebook - https://tmt.ph/facebook  Instagram - https://tmt.ph/instagram  Twitter - https://tmt.ph/twitter  DailyMotion - https://tmt.ph/dailymotion    Subscribe to our Digital Edition - https://tmt.ph/digital  Sign up to our newsletters: https://tmt.ph/newsletters   Check out our Podcasts:  Spotify - https://tmt.ph/spotify  Apple Podcasts - https://tmt.ph/applepodcasts  Amazon Music - https://tmt.ph/amazonmusic  Deezer: https://tmt.ph/deezer  Tune In: https://tmt.ph/tunein   #themanilatimes #worldnews #meta #robotics

Transcript

00:00This demo, we have built a mixed reality experience with.

00:12So in this mixed reality interface, what we get to see

00:15is what the robot is thinking.

00:17I'm going to say, let's clean up the living room.

00:23So now it's thinking what it should

00:25do about the instruction I just gave it.

00:28And the first action it decides to do

00:31is explore the living room.

00:40So what the robot is doing right now

00:42is that it's trying to look for objects

00:45on this side of the house.

00:46To look for objects, it's using the RGB and depth

00:49camera in its gripper.

00:52It turns over, starts looking at the coffee table,

00:55and now does the same thing.

00:57Now, around the coffee table and this bench,

00:59it's looking if there are other plush toys.

01:22So it should not hit the chair when it's trying

01:24to place the object, ideally.

01:27It should find a location on the dining table

01:29that is the correct height and correct width

01:32for this pineapple plush toy.

01:49It's using its cameras on its hand

01:52to find a location that is good for placing the object.

01:56And then it's going to move its arm and base

02:00so that the donut plush toy is nicely placed on the table.

02:07OK, all right.

02:09So this is essentially a glimpse into a future

02:14where we can give very high level instructions

02:16to our robots that are able to perceive their environment,

02:22adapt to changes in this environment,

02:25and then do really long horizon tasks robustly

02:28and improve our everyday lives.

02:31Our machine learning systems or robots

02:34were able to do simple tasks, but they were not

02:37able to semantically generalize to more complex tasks.

02:42And now with more powerful AI systems, what we are able to do

02:45is a much more natural way of interacting with robots.

02:49How I would interact with you, I can also

02:50interact with this robot now.

02:53You and I can't live without our phones.

02:55I think AI and robots are going to be

02:58this for the next generation.

02:59That their everyday life is entwined with AI models

03:06of one type or another.

Category

Transcript

Recommended