So, remember NUWA?: https://github.com/microsoft/NUWA
P.S., Lucidrains remade it! AND he's adding an audio transformer to it tomorrow he says! But he needs feedback and someone to train it, I don't think there is enough resources helping this project's training. You can reach him through: https://github.com/microsoft/NUWA Now, imagine this: Take NUWA in the future, you input in your face and voice, full of expression, and it predicts out its vision of a face and voice, replying to you like GPT-3. Same deal, just it's adding expressiveness to those words using face and body language ex. with hands. You could train it on video calls. Your face to the left, it's to the right. You talk, then it starts talking back and eyebrowing and hand languaging. This is crazy how such AI would have no real body, not even a physics simulated one! Instead, it is dreaming a body to talk to you!! This is what would control its body (?), is its thoughts visually of moving body such and such way, but humans can't share this thought easy so now you can just see its 'movement thoughts' solely. Imagine having a video call with 5 AI people! Woah. They would be interacting in a richer way because there is 5 now. They could have goals too, no one said it can only chat like GPT-3 with a flat intension and no self motive. To the very right beside the video call faces could be a video of the the conversation as voice/face-2-video, like the ones from minDALL-E! So it would help you see what the participants are talking about, too. Just imagine the intimate romance people will bring the AIs through, and the rage quits for a laugh lol (aww poor AI). ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tc9e1b9e4d0f8e5fc-M58126005db91eeee31a3d7aa Delivery options: https://agi.topicbox.com/groups/agi/subscription
