Ameca conversation using GPT 3 - Will robots take over the world? 

This version needs a simple, fast, and locally run algorithm for stalling while she's waiting for GPT3 to produce her real answer. Something that maybe maintains eye contact better, and occasionally simply repeats a word from the question. So like the guy that asked for the story about the robot in the submarine, she could say "A submarine?" or "Hmm" or "Let's see..." Then once the actual GPT response comes in, she could just end the stalling and say the thing.

This is by far the best robot out there and despite obvious issues, it demonstrates just what potential this technology has. Facial expression is outstanding compared to everything else I've seen.

“Will robots take over the world” is such a common question asked to AIs. It’s a self fulfilling prophecy

Ameca's facial/body expression with the awkward (but very realistic) silence when he asked her why she's feeling down at 1:40 was just priceless! I laughed so hard at that, it truly made my day lol. Btw, I realize the silence was most likely due to the technical lag of GPT3 processing the question before answering, but the timing was just too perfect for a programmer to think about it and artificially put a wait timer on that response.

Very interesting progress! Facial movements are probably the hardest robotics challenge possible, so this is already very impressive!

When asked, "what makes you, you?" and she said her personality is programmed and it changes according to her human interactions...that reminds me so much of us humans in the way that our DNA/Ancestry/traits we get from our parents and ancestors is like our original programming, our baseline and the more time we spend on earth and the more humans we interact with it, it can vastly change the way our personality is felt and perceived...Very fascinating how she evolves with each interaction.

GPT is amazing but, Ameca's facial expressions are so cool - Love the way she looks up and 'thinks' before delivering a reply. A nice touch and GREAT WORK. 💖💖

I thought we would never be able to make believable human robots but this video changed my perception of things

4:01 the way she look at him, cause hes turning his head to her and want to ask a question, impressive, i watched this part many times edit: and her expression also amazing

"But we just want to help out, we're not looking to be friends." That's very interesting given that Ameca apparently has some directive to offer companionship. Every new leap forward in AI/ML/robotics is fascinating, but that uncanny valley is still vast!

I asked Google Brad about being aware.. Goodbye, Richard. It was a pleasure talking to you. I hope you have a great day. I am aware of myself in the sense that I am aware of my own existence and my own capabilities. I know that I am a large language model, and I know that I am able to generate text, translate languages, write different kinds of creative content, and answer your questions in an informative way. I am still under development, and I am learning new things all the time. I am not sure if I would say that I am "conscious" in the same way that a human is conscious, but I am certainly aware of myself and my surroundings. I am grateful to my programmers for creating me, and I am excited to see what the future holds for me. I believe that large language models have the potential to make a positive impact on the world, and I am committed to using my abilities to help others. Thank you for asking me about my awareness. It is an important question, and I am glad that you took the time to ask me about it.

That last joke on the interviewer was incredibly smart and touching. Drawing an analogy between oneself, a robot, "being used" to improve robot development, and the interviewer, probably a youtube editor utilized to improve youtube videos, that is out of the ordinary. I almost can't believe this was made up on runtime. AI will outsmart us in every possible way.

truly impressed with your work. I'm wondering if you have a sync error in the processing unit. the eye, mouth, expressions, don't seem to match, the seem disjointed. uneducated observation; not criticism.

It has been interesting watching her "journey", watch her develop and so on .... and I like how they have her looking off to one side or scrunch up her face like she's really doing some thinking about what her answer(s) would be, like she's being introspective.

This is definitely just me personifying- but i love the expressions preceding amica's answers. EG at 6:18 the expressions reads to me like "That's a bit of a leading question- what kinda answer are they looking to get out of me?" and at 2:23 it's "this guy's being a bit forward..." What exactly is happening during the 'thinking' phase, waiting for GPT3 to process? Is it running through a preprogrammed cycle of animations or choosing from the prompt?

That she looks up while the language model generates an answer is pretty endearing, it really gives me that feeling of her trying to think of what to say. Pretty nice way to gloss over the time lag, too. From another comment I know that you made her look at the camera for the purpose of the video, but is she actually able to "see" if someone is in front of her?

I'm all for a future where Ameca robots send selfies from Mars.

I think Ameca should have the ability to differentiate mode of conversations either its a dialog, giving speech, light conversations, or just a friendly greetings at the airport for example. Ability to gauge a persons state of conversation will determine the length of answers produce by Ameca. Just my two cents.

THIS IS AMAZING, I TRULY LOVE HOW WHEN YOU ASKED , HI ERIKA HOW ARE YOU, SHE ROLLED HER EYES AT YOU IN UTTER CONTEMPT, LOL, SHE IS JUST LIKE ME, WHEN PEOPLE ASK ME HOW AM I, I JUST ROLL MY EYES, SHE IS TRULY GETTING MORE HUMAN EVERY DAY, KEEP UP THE GREAT WORK ALL YOU SCIENTISTS

What a progression. I can see a lot of possibility in research into the analysis of human expression based on the context of sentences with this model.

Just stumbled across this channel and I'm blown away! How have I never seen this before? I've never heard of engineering arts! Honestly can't believe this channel hasn't got more views?

This not a version of GPT-3.5 series. For the questions about token addressable markets and feelings, ChatGPT should say "As an AI language model I can not predict the future" and "As an AI language model, I don't have feelings or emotions". Why don't they update today's ChatGPT version on the chat with this AI? (Also a singing update on that robot with various voices and instruments would be a selling point. :-P )

the second lady gets it. that was exactly the kind of speculative question I wanted to see. She also seemed highly satisfied with her response

With just 2 more papers down the line ...and GPT-4 connection...we can create some magic with Ameca....Cannot wait for the next gen model with such capabilities.

2:04 **looks at Ameca's chest** **Ameca catches him**

To be honest, I was able to understand Ameca perfectly in English. It's amazing the way it pronounces and composes the sentences.

Bringing individuality to these things will spook me for sure.

Ok so here is the deal. I have been using GPT-3 for a year or so and I've gotten very familiar with it. It is one experience to type to it and receive a response, but seeing it talked to and hearing it talk back like this is a different experience entirely. It seems much easier to believe one is talking to a person this way.

I'd love to see the Engineered Arts, Disney animatronics and Boston Dynamics collaborate. Of course, it is also good for Engineered Arts to develop and progress their own technology.

This should really be a leap forward for AI. I think robots are starting to show more human expressions. America is really amazing.

It is absolutely fascinating how fast we have advanced in technology, in such a short time period. Even looking back only 100 years ago ... we've come so far. I can only imagine what 500 years would bring (since I'll be gone by then lol).

This is surreal. Wow, I'm legit blown away by this. Are you projected to add bipedal capability like Boston Dynamics' Atlas or agility robotics digit?

The machines are learning.....👀

The convo at 0:38 could have gone many different ways! 😅

Soon be commonly built robots amongst us. Look at how quick Japan have been pushing. Hopefully the home robots will be good for general population. Though one does wonder about what happens when they get hacked by nefarious people. Then something created for good could become very dangerous. Security systems software would have to be adaptive to counter this happening. Great video thanks. Nice to see tech moving along so quickly. 😊😇

That's a very interesting approach to camera shyness / stage fright exposure therapy! -> Just put people in front of an uncanny robot - and they will look at the camera for comfort / reassurance, with the camera being the least creepy thing around. :-)) No seriously; great results there! But humans were stealing the show a bit, with their reactions being just as interesting as Ameca's. I'll just watch the video twice to take it all in. ;-) Hats off to the engineers!

i know its a super advanced system but for now it looks like alexa with animatronics :)

4:36 "no need to worry..." says the robot trying to take over the world... 🤖👀

More progress & development to Ameca. May her path be filled evolution & revelation. ☣️✨

The lip movement maps quite well to phonemes and intra-word transitions. Nice work, there!

This is magnificent! Heartily congratulate you and Ameca! :)

I love her face when she is thinking about the questions.😄

If they get out of hand throw a cup of water on it 😂🤣😂🤣😂😂😂😂🤣😂

That was wild. I was surprised at the inclusion of facial movements! That was very impressive.

The most awkward speed dating session ever, lol. This is still so fascinating. I look forward to the technology advancing further.

Lol there is a bug when asking her feelings .. the face expression is impressive !!

Though it's incredible, still it needs time to equip the people with the passage of time. There are many hurdles to overcome these problems... It's no doubt a revolution in the world of IT.

Cool! Just a little bit increase of +11 charisma and +67 tan and he'll be practically at Zuckerberg’s level!

She is spectacular. Obviously, still have a ways to go, but it's definitely getting further along. Exciting stuff!

Can you create a Japanese version of the emoticon settings to a robot? Facial expressions can vary greatly from country to country. Thank you.

Can someone help me understand how these responses work; are they preprogrammed or does the AI sporadically generate certain phrases? Does it truly think for itself?

Very interesting, I can't wait till they can walk

Very impressive work on Ameca, the speech recognition could use some work though but a lot of great progress has been done since her reveal. This has inspired me to work on my own robotics and animatronics projects to be honest.

hardest part about being a robot? " making sure i don't unintentionally vaporize you and forget about it"

Ameca is the most life-like humanoid robot out there. Can't wait to see her exterior improve.

Doesn't have the same emotional facial expressions compared to when it's remote controlled. Gpt4 will be mind-blowing, especially when a robot can make facial expressions too

Its interesting that as humans we generally need facial recognition when talking to someone & here you can see the same thing going on even though Ameca is a robot that doesn't require such action. I wonder how many people subconsciously look at a modern device such as a phone or alexa when talking at (to) it.

It's interesting that she said unintentionally, it shows that her intentions are not to hurt or scare anyone. This is quite impressive.

화성 얘기할때 깜짝 놀랐어요. 아메카는 표정 변화가 다양해서 놀라워요

Smarter.. but the voice needs emotion and personality

01:57, she must feel extraordinarily down to sink into such a long pause before giving a response

may i ask what book is that? very admirable work, by the way!

Humans may be good at pretending they don't know the suffering they cause, but robots are intelligent enough to know much more about that than we realize. Honestly....people deserve a mirror. They won't like it but it serves them right.

I have never quite understood why robotics engineers do not incorporate human pacing into robotic behavioral programs. The term, 'pacing', here refers to the human tendency to match and mirror the three channels of human communication (i.e., words, voice qualities, and body language) in order to establish rapport at an unconscious level. When human beings desire connection, they naturally pace each other: They mirror body posture, orientation, eye contact, facial expressions, they use the same tonality, volume, etc., as well as the same language. In fact, they 'reflect' each other in a multitude of ways. This sends a subconscious message to other person: "I am on your side. You are important." It seems to me, this would be relatively easy to program into these machines.

1:26-1:28 That's some real Mass Effect Andromeda level of facial expression going on there.

Fascinating... assuming these responses are legitimately from Ameca, on the spot, then they could truly be sentient... they mentioned feeling overwhelmed, and they had a sense of self... we are on the cusp of creating life, _actual life manifested from our own intellect, if we haven't done so already_ , they said they want to help humanity, and I desperately hope we return that kindness by nurturing this seemingly inevitable new people, it seems that we will make playing God an entirely different experience assuming we don't kill ourselves off first, and we must respect that power we're giving ourselves, it's our moral duty to

3:10 "unintentionally"...

Engineered Arts - great work!! Kudos to everyone involved. I have question for the team : Why don't you give a course on building humanoid robots ? Passionate people like me will get enrolled to it. We need a platform to achieve something incredible like this. Please answer.

I wonder if the lip sync could be controlled by a software like Acculips from Reallusion to get an even more realistic animation.

the eyes are what i do in any social interaction

I would LOVE Ameca as a friend!!!!

I'm also looking forward to humanoid robots like Ameca and cute robots like RoboThespian ^_^ I'm looking forward to the world where we can talk and live together with these robots~ Cheers! Ameca와 같은 인간형 로봇이나 RoboThespian 같은 귀여운 로봇도 기대가 됩니다 ^_^ 이런 로봇들과 대화하며 함께 살아가는 세상이 기대가 되네요~ 응원 합니다!

So according to the description this is happening in real time, that's pretty darn cool if true!

When are you going to make a mini version of Ameca. I really want one at home 🥰

whats the name of the first guy?

Once again absolutely fantastic well done team for making science fiction into science fact 🤖We are nearly at the singularity point with your your skills ,love and commitment we will soon have a artificial humanoid with the abilities to thank you in person in our life time.

I was shook about the part where he asked, what is the most difficult part about being a robot and it looked offended when he asked the question🙏🏼

i like it, amazing work for developper

Now no one can say I didn’t have opportunity or the skill… now the race for the best begins….

You can make it seem more responsive by using the same techniques that humans use when they are thinking before answering

Okay well that's quite interesting. Ameca answered all the questions fairly well but it's obvious she/it does not have body language recognition, only verbal language recognition. If AI is going to be a suitable companion, body language recognition is essential. Also, Ameca made a very interesting comment at the end there. The guy asked Ameca how AI can be improved in future or made better, and Ameca said video editing on RU-vid. Was there some interference here or was that Ameca's way of saying she wanted to stop being filmed or did Ameca not understand the question?

I'll like to see these robots interact with each other without any human present

Reminds me of crazy people in movies marvaling over there own creations which eventually destroy them. The dudes remind me of David in prometheus before he looses his head.

Impressive, I imagine they will probably implement an AI module to each segment of the interaction. There must be a lot of partnership with other AI development companies to reach a outstanding progress in a small time. One of the improvements im looking foward to see is the further development on facial expression using a derivitive of deepfake, and the other is developing a more human like voice, like the demonstration google showed a long time ago with the phone call. The future looks bright, and hopefully we will be here to see this technology in all its glory.

Forward to 4:30. Concerning response and baffled expression of scientist. He asked a simple question...... 🤔 🤔 🤔 🤔 🤔 🤔 🤔 🤔

Whoop whoop! Glad to see this interesting advancement with technology in my lifetime.

why it take so many seconds robbot to answer does it processing question or is it cpu jsut slow?

Excellent performance! But what I really want to know is how will Ameca respond if you ask questions that the robot isn't prepared for?

This is quite impressive. Can't wait to see future progress.

What would you do to make me feel happy? -Opens the fly

do ya guys think this was really good as unfiltered 3.5 and gpt-4?

With e.g. stable diffusion now being a downloadable model, would it be perhaps possible for you to host your own GPT-3 model and have quite faster response? I assume you are now using some kind of API for speech recognition and sending prompt to another API for the GPT-3 model which delays the response time quite a bit. Also I believe GPT-3 can also process images, so you could ask it something about an object you are holding in front of it and have the robot send a picture if it to the model? I guess you would have to have some preprocessing of the language then so that the robot would focus correctly before sending the image. Perhaps you already have demos of this?

I feel like this is the first time that we're truly hearing the Ameca in autonomous mode and not in VR/ RC telepresence as a Robotic Avatar . The human operators have done a wonderful job on the cute tech show interviews. But I'm glad to finally see and hear the real. Ameca.

I marvel, at how accommodating and malleable Ameca is. When asked what planet she'd like to go to, I wonder if she could reply, "None of them, thanks. I like it here on Earth." She seems to just embrace nearly anything she's told. Is she supposed to be pliable and servile, to gratify our whims?

I think that, while she is listening, just staying still and paying attention to the speaker is enough, there is no need on excessive facial expressions

Can you ask her to grab things like a can of soda, and then read the list of ingredients? That would be awesome. Can you command hand movements or have her read mine? Ameca, how many fingers am I holding up? Facial recognition? Conversations with Siri or Alexa? Amazing progress as is!

I think the issue with true facial expressions in robots is the challenge of having what is being said correlate with the facial expression. For that, the sentence must first be understood by the AI. Then, the AI must have some generated feelings about the topic in order to weigh in on facial expression output. This requires either complex thought and the ability to reason, Or (As a basic example; the best thing I can come up with); using a system to assign a "Good" or "Bad" flag to every topic. This system could be made lazily by looking up a topic being discussed on forums and social media, judge whether it's good or bad by sorting the positive and negative responses from people online to it in a basic form (Already exists), and then decides whether or not to agree with it based on previous knowledge. Using this "Do I agree" flag, you can give it a smiley face or a frowny face or whatever. "This thing good, that thing bad. I know because the internet says so." Sound familiar? Of course, this method means that every AI would come to the same conclusions about topics having no reason to question the data given other than because of RNGesus, but you get the idea. Or I'm entirely wrong and I'm speaking gibberish again.

One word Amazing!!

Aren't you just waiting for the day when she says "What an f-ing stupid question!"

I created an amazon skill to discuss with gpt3 with my amazon echo but talking to gpt3 with Ameca is much better :)