I'm skeptical about any of the real-time movements or corrections. But only skeptical, not denying the possibility. Their access to computing power is almost infinite compared to the competition now that the two companies are working together.
The delay in speech interaction is still evident, so there's time taken to query a gpt response based on the camera image and generate a text response. They configured the assistant instruction (or personality as in gpt-4 robot skill) to have mmm's in the response.
One would assume the movements are scripted. And the gpt chooses an appropriate action based on the conversation. The depth data of the 3d image could have been provided with the query to generate the path planning, like how I demonstrated in my gpt robot arm example post.
The thing about any realtime corrections after the initial pause is interesting because the initial request and response pause is there, so how is it able to perform realtime corrections after it takes time to interpret the command. I'm still leaning toward the actions being scripted, but you never know with these guys. Again, their access to computing power is insane - so it's not out the window to imagine they could be computing realtime.
I'm sort of flip flopping on it because I think inside, I want it to be real. But there's details not provided to us on purpose, which is why I'm leaning toward the actions being scripted. But, I can see how the action path can be computed because i did it.
@Jeremie the thing about pressure picking up could be spring tensions on the fingers - so there's give. And also it could have a table of "how much force per object". Or it could have a "how much does an object weight" query that references a multiplier for the finger force.
They should have to provide some type of disclosure but after Bard fake video who knows what is real anymore.
the apple in the start position is a bit of a give away (that’s how we would all do it) and the large soft objects in a defined space with a unique pattern for the garbage is also a very mostly pre programmed response (I wonder how many takes that took)
I do like the hmmm um aghh concept for the robot while computing response though, I may steal that idea with a list of stall sounds, words or phrases. Give it more of a ummm human touch.
this all reminds me of the old brain control interfaces that measure micro expressions I mean read minds. How many Venture capitalist fell for that one.
Boston Dynamics have joined the Humanoid robot using BLDC motors space. This one is interesting although I get the feeling any more rotations on the legs and the wires will break.
Boston Dynamics
I agree it’s cool - I’m curious about the motor sounds. I’d they are super high pitched. Because they’re doing the same sound dampening as they did with atlas.
I like these advancements - because it’ll result in more interest in robotics. We’ve seen a significant increase in users since gpt. I’m gathering these new robots is why we’re seeing continued daily increase.
On a side note it’s interesting that we have more users than ever, by double. But the forum is a thing of a the past. It’s a great resource for searching, but it doesn’t work if ppl don’t share questions and challenges for new searches haha. It’s tough cause ppl don’t talk on the internet anymore. Too many scammers and arguments elsewhere - they’re exhausted. I feel the same lol
Im guessing the ARCx would also inspire a lot of new interest - for those who don’t want to be locked into windows!
I find most people have migrated to discord. There are a number of active robot channels where people collaborate on various topics. Forums are great for boomers and genx but millennials and gen z are all hanging out on discord (btw there are a lot of synthiam users who collaborate on discord as well). Granted you lose control of your content but you can do things like hook Athena into discord as well as be the moderator so you can delete kick ban etc. to be honest it is very much like IRC we used in the early days of the internet.
Athena on Discord would cost a fortune. It seems even the boomers don't want anything to do with forums.
I use Discord for our retro NABU computing group. Most of the people there are boomers and ignore the forums. I like Discord for chatting, but the content gets lost quickly and isn't very searchable. It's too linear for me, and I find it similar to MSN Messenger or IRC.
With the new Atlas I feel kind of betrayed by Boston Dynamics. For years and years, they touted hydraulics as "The" technology to use for humanoid robots. They put so much effort into Atlas's hydraulics over the years and were a leader in this technology. For them to switch to BLDC motors seems like a necessary pivot but it means that they misled the Humanoid robotics community as they tried to make hydraulics work for so long. I feel betrayed because many said that hydraulics were too energy hungry but Boston Dynamics kept on pushing the technology. They even had me convinced over time that maybe there was a chance to make it happen. I put some time and ambition into learning about hydraulics as a result. Luckily, I didn't invest much at all into the technology but rather devoted myself to 3D printing. I'm so thankful that I didn't dive headlong into hydraulic systems, but I'm nonetheless a little aggravated that Boston Dynamics made me believe for so long.
I'm skeptical about any of the real-time movements or corrections. But only skeptical, not denying the possibility. Their access to computing power is almost infinite compared to the competition now that the two companies are working together.
The delay in speech interaction is still evident, so there's time taken to query a gpt response based on the camera image and generate a text response. They configured the assistant instruction (or personality as in gpt-4 robot skill) to have mmm's in the response.
One would assume the movements are scripted. And the gpt chooses an appropriate action based on the conversation. The depth data of the 3d image could have been provided with the query to generate the path planning, like how I demonstrated in my gpt robot arm example post.
The thing about any realtime corrections after the initial pause is interesting because the initial request and response pause is there, so how is it able to perform realtime corrections after it takes time to interpret the command. I'm still leaning toward the actions being scripted, but you never know with these guys. Again, their access to computing power is insane - so it's not out the window to imagine they could be computing realtime.
I'm sort of flip flopping on it because I think inside, I want it to be real. But there's details not provided to us on purpose, which is why I'm leaning toward the actions being scripted. But, I can see how the action path can be computed because i did it.
@Jeremie the thing about pressure picking up could be spring tensions on the fingers - so there's give. And also it could have a table of "how much force per object". Or it could have a "how much does an object weight" query that references a multiplier for the finger force.
Darn it, wish they provided more info for us haha
They should have to provide some type of disclosure but after Bard fake video who knows what is real anymore.
the apple in the start position is a bit of a give away (that’s how we would all do it) and the large soft objects in a defined space with a unique pattern for the garbage is also a very mostly pre programmed response (I wonder how many takes that took)
I do like the hmmm um aghh concept for the robot while computing response though, I may steal that idea with a list of stall sounds, words or phrases. Give it more of a ummm human touch.
this all reminds me of the old brain control interfaces that measure micro expressions I mean read minds. How many Venture capitalist fell for that one.
Boston Dynamics have joined the Humanoid robot using BLDC motors space. This one is interesting although I get the feeling any more rotations on the legs and the wires will break. Boston Dynamics
Boston Dynamics is a company to watch for sure.
I agree it’s cool - I’m curious about the motor sounds. I’d they are super high pitched. Because they’re doing the same sound dampening as they did with atlas.
I like these advancements - because it’ll result in more interest in robotics. We’ve seen a significant increase in users since gpt. I’m gathering these new robots is why we’re seeing continued daily increase.
On a side note it’s interesting that we have more users than ever, by double. But the forum is a thing of a the past. It’s a great resource for searching, but it doesn’t work if ppl don’t share questions and challenges for new searches haha. It’s tough cause ppl don’t talk on the internet anymore. Too many scammers and arguments elsewhere - they’re exhausted. I feel the same lol
Im guessing the ARCx would also inspire a lot of new interest - for those who don’t want to be locked into windows!
I find most people have migrated to discord. There are a number of active robot channels where people collaborate on various topics. Forums are great for boomers and genx but millennials and gen z are all hanging out on discord (btw there are a lot of synthiam users who collaborate on discord as well). Granted you lose control of your content but you can do things like hook Athena into discord as well as be the moderator so you can delete kick ban etc. to be honest it is very much like IRC we used in the early days of the internet.
Athena on Discord would cost a fortune. It seems even the boomers don't want anything to do with forums.
I use Discord for our retro NABU computing group. Most of the people there are boomers and ignore the forums. I like Discord for chatting, but the content gets lost quickly and isn't very searchable. It's too linear for me, and I find it similar to MSN Messenger or IRC.
With the new Atlas I feel kind of betrayed by Boston Dynamics. For years and years, they touted hydraulics as "The" technology to use for humanoid robots. They put so much effort into Atlas's hydraulics over the years and were a leader in this technology. For them to switch to BLDC motors seems like a necessary pivot but it means that they misled the Humanoid robotics community as they tried to make hydraulics work for so long. I feel betrayed because many said that hydraulics were too energy hungry but Boston Dynamics kept on pushing the technology. They even had me convinced over time that maybe there was a chance to make it happen. I put some time and ambition into learning about hydraulics as a result. Luckily, I didn't invest much at all into the technology but rather devoted myself to 3D printing. I'm so thankful that I didn't dive headlong into hydraulic systems, but I'm nonetheless a little aggravated that Boston Dynamics made me believe for so long.