In 9+ months, only 18K views and 0 comments... really?! AI will be more transformational in our lifetime than electricity, computers and the internet combined. Videos such as this one offer us a glimpse into the workings of our shared future. By watching, commenting and understanding the content of these videos you are increasing the probabilities of a better future for us all.
How exactly does passively watching these videos lead to a better future? Open AI dont need youtube for funding and if you were even remotely serious about learning about AI, watching youtube videos is far from the most efficient way to get there.
Yes, it's often the case that what appears to be an "overnight success" is actually the result of years of dedicated effort behind the scenes. OpenAI's journey reflects this pattern. Despite being perceived as an "overnight success," it took the organization approximately 8 years of rigorous research, development, and collaboration to reach its current status. This underscores the importance of perseverance, patience, and continuous effort in achieving significant milestones and breakthroughs .
Okey! The vision network predict the locations of the blocks(and it is propagated earlier). But how you train the imitation network. So confusing to me. Because the demonstration has thousands of images. Just think about it you have 128×128 image, with duration about 5 minute, each has 24fps. So the math would be 128x128x24x60x5 plus locations of the blocks. Eww! That's a minimum but huge input feed to the imitation network. My point is how you train the imitation network. No back propagation I think is possible.
Hey person from the past. This AI doesn't actually work in camera captures alone. They only use it extensively to train a model with recorded data, which uses its own language to describe a given situation in numbers and code. Much like you have specific coordinates in games to determine your location in a map. It doesn't need to see the map to know where, for example x100 x y500, is on the map. Only based on these coordinates and the placement of the map in the tiling, the game knows where walls are and all other triggers. The code to make this possible is a lot smaller than actually needing to see the footage every time. After training, the footage can be deleted because it learned everything already. Works especially good for stationary cameras as the plane where the objects are can be predefined. Otherwise you'd have to scan (photograph) from a lot more positions and angles to train it to account for more variables than just position on a fixed plane.
@@fraizie6815 Interesting, If understand it correctly, Its turn video data into some kind of language model data, which also has block description(like 3D coordinate, and color) then its generate some prompt to drive the actuators?? Correct me if i wrong?
@@saltyyolk9934 Yes, basically. For example: the code can have the hex value of the colour of the object. When the camera sees a yellow object at coordinate e.g. x200 y500, it knows, from its self learned code, that this is the object to grab and where to grab it on the plane. With some more learning it may be able to differentiate between a number of yellow shades or pick one that's close enough to the wanted value to consider it 'true' - or in other terms the right object to pick.
Could you explain any benefit from this system compared to a regular edge detecting vision system in combination with an industrial robot that is able to be handguided, like the Kuka LBR IIWA? Are you able to make a sloppy stack and then let the robot learn to create a perfect stack over time or what is the point here? ☺
am not sure if i am helpful to you the basic point of this is to build system planner to solve tasks by simulating them. any plan basically is define the current situation, your goal and the process to reach the goal. and when robot have the task will define the current status as starting point. stacking blocks will be the final state or goal. and the robot process using the minimum moves ..................... i hope that's what you wanted to know, else i am still student i am not sure if i can be more helpful
You don't program it, you don't design control theory neither a phisical model of movement for the robot, nor driver. Pretty cheap, and saves a lot of time huh? You just plug a computational agent with a goal and train it. It's like it where alive. That's what the heck why.
Yes , after passing 34.7 wormhole to this Era my life is very easy because in this Era humanity achieving poor development about technology..................?
Bro I am from amdromida galaxy's south Central Western corner From 5 billion year ahead from this time I just passing through our super crystal of galaxy and saw you comment that make me happy that 2023 human are so polite
I'm here after invention of ChatGPT 4 ChatGPT 5 very soon, maybe at the end of 2023 or at the beginning of 2024. 9/9/2023 [ 5:20 PM ] + 71th comment ✨✨🔥🔥💀💀
I’m here after invention of DALL·E 3 Still waiting for GPT-5 in 2024. Fri, 13/10/2023 [ 3:15 PM, GMT+8 ] + 74th comment Edit: I have returned here at 10:06 PM on the same day because I just realized that today just so happens to be Friday the 13th... on the Halloween month too... what a coincidence
@@eclassiskandar8190 Nice That's Cool AI is developing very quickly and time passes very quickly, I hope GPT-5 release in the end of the year, Nice it's been 1 month since I wrote this comment Fri, 13/10/2023 [ 1:26 PM ] + 75th comment :) .
I'm here after invention of SORA (text prompt to video generater) Still waiting for GPT-5 Edit: forgot to put date Thursday, 22/02/2024, Time 3:58 pm + 92th comment ^_^
@@SIIIUUUU. lol time so fast I remember I wrote this comment like Yesterday, SORA is massive and stunning AI video and GPT-5 ( AGI ) will arrive or come out after 7 - 9 months + Stable diffusion 3 is better than DALLE-3 so Im waiting for DALLE-4 also Agent AI like ( MultiOne ) and much more has happend and came out and more will come in the future Monday, 26/2/2024, 3:44 PM + sry for this newspaper and thats fine my friend + 95th comment !!
