Mindplex Episode 10: David Hanson On Using Humanoid Robots To Train Compassionate AGI
Watch the 10th Mindplex Podcast and livestream Pubcast, featuring David Hanson, founder of Hanson Robotics — maker of our captivating co-hosts Grace and Desdemona Robot, and also of Sophia, star of the new SophiaVerse.
Greetings Singularitarians, and welcome to your next glimpse of the rapidly approaching future, with Ben Goertzel, Lisa Rein, and Grace Robot, your hosts for The Mindplex Podcast:
The Mindplex Podcast team will host a watch party “Pubcast” live stream of Episode 10 on SingularityNET YouTube on July 13 at 11 AM PST/6 PM UTC.
Mindplex Pubcast Episode 10: Humanizing AI & AGI
Thursday June 13 at 11 am PST/6 UTC, David Hanson will be joining us again LIVE for the Live Pubcast, along with co-hosts Ben Goertzel, Lisa Rein, and Grace Robot; they will discuss the episode and events and updates since it was recorded…a live conversation that is not to be missed!
Watch Mindplex Podcast Episode 10 with David Hanson on robotic embodiment of AGI
Watch Mindplex Podcast Episode 10: On Using Humanoid Robots To Train Compassionate AGI
During the Podcast episode, David Hanson and Ben Goertzel have an intense discussion about the merits of embodiment for training AGIs — by providing direct sensory input — as well as the importance of distributed decentralized knowledge graphs that AGIs use to share knowledge. Ben asks David if he could explain a bit to the audience about how he views the relation between embodiment and general intelligence at the animal, human and super levels.
David explains that humanoid robots are the best way to provide AGIs with direct sensory input while also helping them understand how humans operate so that they can learn faster about us and ultimately connect better with humans on deep and compassionate levels that would not be possible otherwise without this “mind-body” connection between the humanoid robot and its world.
As David Hanson explains:
“There’s a lot of biology and neurobiology that supports this mind-body connection. So, for example, Antonio Damasio, his work shows how the mind actually maps our body awareness. And, of course, this makes sense; the human brain came rather late in evolutionary history. And so, it is ingrained and coupled with many body processes. So, for intelligence to work properly, it requires this kind of embodiment, at least in organisms with nervous systems.
And a lot of what we think of as “intelligent behavior” is this interplay between a mind body with a physiology, hormones, circulatory system, respiratory system, enteric, nervous system, spinal column, all of this interacting with the musculature and skin and, you know, the boundary to the physical world. So all of this interacting within a physical world and learning and growing within that physical world. And this is how we learn and grow as intelligent beings.
AI doesn’t have that. It has disembodied interaction with the data that comes from humans, who provide it from this kind of embodied experience. So in a way, when we’re looking at these large language models, we’re looking at the shadow of human experience…
There’s a lot of powerful data that can come from face-to-face interaction between a robot and a human being. And so, then it means not just physically interacting through hands and sensors that way but also interacting by being, you know, 3D presence. To make eye contact with a person to learn from people. To have these kinds of grounded situational experiences… Combined with this episodic situational memory, this is going to be where AI is gonna go next. This is where AI, I think, will start to show AGI capabilities.”
Ben Goertzel responds with an important point that, in order for a system to be able to answer stuff and generate smart stuff about the world humans live in, in the end, it’s got to be able to be truly grounded in human language. Not just music, not just pictures, but gathering input from all the senses and all the modes of movement that people have.
As Ben Explains:
“Otherwise, it will only be able to deal with human-like self by very advanced reasoning. You’ve got to have either super advanced reasoning, so you can, like knock out something that you have no actual experience of.
Or you just have to have the sensors and actuators integrated together in the right way. And ideally, we can have both, right? We want to work on advanced reasoning, given the sensors and actuators that you, that humans have. And embodiment is basically the latter, right? And I mean, David’s been pushing on this for a long time, but now things are starting to accelerate.”
Later in the show, Lisa Rein asks David to give his larger vision of his goals for embodied AGIs:
As David Hanson explains:
“By creating this physiological simulation, coupled with the brain-inspired cognitive architecture, and having evolutionary inspired goals and motives — the drive to exist, and to be curious, and to socially bond. These will make the AI more compelling and interesting and scaffold memories on those experiences.
We can couple that sort of expertly crafted biosystem with GPT models, and then prime the attention networks to focus on what’s important. What it is to act like this particular character.
So this is a powerful way of using today’s technologies to then create an agent that has agency. So, GPT only has agency — GPT models generally only act intelligent — when you’re prompting it. In other words, it just sits there and waits. And based on the prompt, then it activates different regions in the space. So that’s what gives the “motivation” to behave…
This then means that we have the support ingredients for AI to grow up among humanity, to form these kinds of mutually beneficial relationships with humanity, to then understand the human experience because of the physiological simulation. If you don’t have that, then how can AI care? How can they get it “in their gut,” so to speak, what it is to go through a human experience?
So then, you have the potential for the AI to truly value what it means to be alive through the power of compassion. To feel with the human. And so then to build those memories, experiences, and motives. For a commonly shared value system, you need that humanized AI.
So I really think this is the pathway to not just make AI generally intelligent, but to make it generally good to make it good for humanity to make it beneficial.”
Other topics covered include:
- The value of embodiment for training AI and AGI.
- The relation between embodiment and general intelligence.
- The biology and neurobiology of embodiment.
- Grounded situational experiences and episodic situational memory.
- Making AI generally good and beneficial to humanity.
- Ethics of AI and robotics.
- Distributed decentralized knowledge graphs.
About Sophiaverse — Humanity and AI in a shared quest for AGI
SophiaVerse is a gamified decentralized AI ecosystem where humans and AI can work together to build superintelligent systems leading to a beneficial Singularity. You can work alongside your very own Sophia, teaching her AI systems how humans learn, play and problem solve; while learning more about the workings of AI, helping to bridge the gap between humans and machines, and build a bridge to our shared future.
Joining SophiaVerse, you are participating in the journey of evolving intelligence, the journey of emerging. This is a planetary journey, a human race journey.
In SophiaVerse, you can play games, program AI, and earn SOPH tokens; you can create artworks and game components with generative AI which will seamlessly integrate into the metaverse architecture, and offer them on the marketplace. SOPH tokens can be used to buy in-game items, access exclusive content, and many other things.
With SophiaVerse, anyone can learn about, build, and contribute to the development of superintelligent AI. The ultimate in decentralized AGI development, SophiaVerse will be a game, a playground for young AGIs, a HiveMind, and the decentralized and tokenized quest for achieving beneficial machine sentience.
Episode 10’s Featured Guest: David Hanson
“On the tree of robotic life, human-like robots play a particularly valuable role. It makes sense. Humans are brilliant, beautiful, compassionate, loveable, and capable of love, so why shouldn’t we aspire to make robots human-like in these ways? Don’t we want robots to have such marvelous capabilities as love, compassion, and genius?”
David Hanson, Ph.D.
David Hanson is an American roboticist and the Founder and Chief Executive Officer of Hanson Robotics, based in Hong Kong.
David develops robots widely regarded as the world’s most human-like in appearance in a lifelong quest to create true human-caring machines. He is known for creating several of the world’s most famous robots, including Sophia, and her little sisters, Grace Robot and Desdemona Robot, who take turns co-hosting our Mindplex podcast and pubcast.
By seeking to emulate human bio-systems (from cognition to social expression to locomotion), Hanson hopes to create machines that are creatively brilliant and truly conscious, that also have deeply connected feelings towards humans.
As David explains in the podcast, he truly believes that only by building such human-like machines can we realize the dream of safe, benevolent, superintelligent AI and AGI.
About Mindplex
The Mindplex Podcast team invites you to join them as they contemplate and learn about this fascinating world. They look forward to meeting you so you can help each other understand the true nature of all the challenges involved as we work together to create a Benevolent Singularity.
Each release of a Mindplex Podcast will be accompanied by a live “pubcast” watch party.
The Mindplex Podcast and Mindplex Magazine — are both part of the Mindplex decentralized media platform spin-off powered by SingularityNet
We invite you to check out the Mindplex website, set up a profile, and perhaps even contribute to the magazine.
Follow the Mindplex Project:
Visit Mindplex Magazine
Join our Telegram
Join our Discord
Follow us on Twitter, LinkedIn
Like us on Facebook
