Heather Knight - Seth Lloyd - Geordie Rose -
Stephen Wolfram - Kevin Jain - Jeff Lieberman -
Darlene Cavalier - Michael Smolens - Itamar Arel - Ben Goertzel
Heather Knight runs Marilyn Monrobot Labs in NYC, which creates socially intelligent robot performances, sensor design and electronic art. |
Heather Knight on Why Robots Need to Spend More Time in the Limelight: People Tracking and Artificial Personality
She’s talking about the intersection of robots and entertainment. She asks are robots the new vampires?
She has a robot prototype she’s setting on the table next to her.
What’s great about robots is they have AI and sensors and can learn stuff from us. She has a strong history with robots.
How could robots understand human gestures? She worked on this at the MIT media lab, and made a design so a robot could feel tickles and hugs.
They’re have been previous projects doing this kind of things. For instance, there was one robot that lost its desire to work at Osaka University in Japan.
Now her robot is moving on the table in front of us: “What do you want me to do?”
She asks it to tell us a story, but it just asks what she wants it do again.
Now it seems to be telling a story about Star Wars, with Darth Vader breathing and lightsaber sounds, R2D2 noises and now the Star Wars theme music.
The crowd claps, which scares the robot.
All these advances will let us integrate robots into our lives, or give astronauts someone to interact with.
And play. The robots can learn something about us during play.
This was a really well done presentation, very professional and cool demo.
Finally she asks, which is cuter, a vampire bunny or a robot bunny?
Seth Lloyd is a Professor of Quantum-Mechanical Engineering at thre Massachusetts Institute of Technology. |
Seth Lloyd: The Democracy of Knowledge
What role does science have in society?
We can define science as knowledge that can be verified by anyone, not just a scientists.
Science belongs to everybody, not just to scientists and engineers.
There’s a lot of science performed in secret government labs and the like, which is fine, but that’s not really science.
There are some questions from the crowd, someone asks about open access and he says his computer is a Microsoft free zone.
He says if you have to have the word science in your name, maybe you’re not a science.
There’s a terrible trend where the government is moving away from funding basic science to applied science, which he says is a mistake.
He says he’s going to use the examples of photosynthesis and time travel as instances of quantum computing. Photosynthesis actually involves quantum computing, which he thought was ridiculous at first but it’s true.
That’s an example that’s actually useful, now onto time travel. He says anger is a great motivator in science to explore this kind of thing.
He talks about a conversation that Richard Feynman had. The theory was that positrons are electrons are going back in time.
We could send photons back in time in the lab. He jokes about the photon killing itself, and the grandfather paradox. When they sent the photon back in time and tried to have it kill itself, it never did kill itself.
Dr. Geordie Rose is the founder and CTO of D-Wave. |
Geordie Rose: Building and using superconducting quantum circuits that learn
D-Wave is a large project, not like citizen science, more like the LHC.
When they first built their quantum computers, they weren’t sure what to do with them.
Their machine isn’t what people normally think of quantum computers as. It tackles optimization problems.
The slide says “learning is the key to AGI,” which is “regardless of your definition of intelligence.”
He wanted to figure out what intelligence was, so he went online and got hundreds of definitions. They were contradictory and several of them wouldn’t include humans, but they all included learning and adapting to the environment.
Is quantum mechanics required for transhuman learning ability?
Some people believe you should never talk about intelligence and quantum computing in the same sentence.
One possibility is that evolution always leads to classical brains like we have. However, there are quantum algorithms for machine learning (these algorithms are simply prescriptions for solving hard problems).
On the remote possibility that our brain uses quantum mechanics, we definitely need to study quantum mechanics and physics.
Now he’s talking about deep neural nets and unsupervised learning. Unsupervised learning would involve, for example, giving an AI a large input of images, and it categorizes them by whether they have cars, without you telling it to.
Now some pictures of quibits. There’s been a lot of progress in teaching them how to learn. Quantum neural nets have created the best car detector program ever.
Learning underpins AGI, we need to understand learning to build AGI.
Kevin Jain is an undergraduate at Harvard University, and is Founder and President of the Harvard College Future Society. |
Kevin Jain: Transhumanism & Education
Now we’re going to hear from Kevin Jain, founder of the Harvard Future Society. He’s the reason we’re here at Hartford. His course of study at Harvard included work from IEET’s James Hughes and Nick Bostrom. He had a special study track in transhumanism, and Lightman says it would be great if Harvard were the first university to make this into a degree program.
Jain lists five of the most popular majors at Harvard: economics, psych, bio, CS, government. Economics is based on the assumption of scarcity. Psychologists assume the limitations of the human, and the inevitably of suffering. Biologists assume life is carbon based and naturally selected. And all the other disciplines have similar assumptions.
But what happens when those assumptions are broken? How will nanotech, AGI, cryonics, the end of death, etc. affect these assumptions? What happens if nanotech makes scarcity irrelevant? What happens to psychology if the inevitability of death is no longer present?
Jain questioned whether what he was learning in school, with all these assumptions, would be relevant to the future. So he made an independent study to discover whether his education would still be relevant.
Jain resolved to expand his curriculum, and now he’s working on a new type of textbook. It would function to explain tech development and progress, engage assumptions, empower new leaders, and educate.
Is it relevant to get a Ph.D. in economics if its assumptions will no longer be relevant in the future.
Imagine if instead of statically assuming scarcity, we could adapt. What if we combined emerging technologies into dialogue and question whether these assumptions are relevant?
He’s working with a large team of professors and students to make this happen, so if you’re interested contact him for more info.
Jeff Lieberman is a musician wrapped in a roboticist sculptor wrapped in a photographer. |
Jeff Lieberman on A New Kind of Citizen:Where We Haven't Looked
Lieberman wants to ask a fundamental question: why do humans have an instinct to change things around them?
What if the deepest way we can change things is to change the way we think?
All the senses and streams of information we take for granted, but they’re only a small window of reality.
“We do not perceive reality - we perceive functionality.”
He’s asking the audience to close their eyes and imagine yourself as a child in the house that you grew up in. He says you’re clearly not the mental image, you’re observing it. Now he asks us to focus exclusively on the air going through our nostrils for 30 seconds.
99% percent of the human species found the first task easy, but focusing just on our breathing is next to impossible. This shows we’re very good at telling our brain to do a task, but very bad at doing nothing.
“The Human Operating System (1.0)”
He says he just wants to propose that we’ve evolved a tool that allows us to create alternate realities.
We think that what it means to be us is to have thoughts, but for most of our evolution we couldn’t do that. We couldn’t create alternate realities for survival benefit.
To do these simulation we have to devalue what’s going on right now.
We tend to be most happy when 100% of our attention is given to the task at hand (”flow”). Withouth the capacity for alternate realities, there’d be no way for negative emotions like guilt, jealousy, anxiety.
How could this change?
“The Human Operating System (2.0?)”
Western science has studied mindfulness, and the people who’ve worked on that are less stressed, happier, etc.
The ironic thing it’s one of the oldest pieces of tech on the planet. Mystics have been talking about this forever, but with language that isn’t scientific.
Our OS keeps pushing us to change things for the better, but we have to learn how to use this as a tool. Right now, that tool is out of of control.
A citizen scientist can take action right now, by closing their eyes and examining how their mind works.
Darlene Cavalier is the founder of ScienceCheerleader.com, a blog that promotes the involvement of citizens in science and science-related policy. She is also the cofounder of ScienceForCitizens.net. |
Darlene Cavalier on Citizen Scientists: Disrupting Science... In A Good Way!
Alex Lightman gives Darlene a nice introduction as a science cheerleader.
Darlene says citizen scientists go by several names, and they disrupt science in a good way. She wanted to see if she could get them involved in policy.
She ran into some challenges in Congress and in the scientific community. People think the public is too stupid for science, but she’s implying that’s not the case.
She says there are over million citizen scientists worldwide. Most of them have advanced degrees, are affluent, active users of the Internet, etc.
Most citizen scientists are from Generation Jones.
Why volunteer for science? To help the environment, advance science, improve the community, to expand your knowledge, etc., but not for a living.
She talks about Galaxy Zoo, which involves sorting through images of galaxies. This activity attracts mostly males, but nature activities are 50-50.
The number one reason people contribute to Galaxy Zoo is to contribute to research.
What’s the impact? Peer-reviewed papers, and discoveries in astronomy, “Climategate,” etc. There’s an increased public understanding of science. For policy, there’s the Expert and Citizen Assessment of Science and Technology.
Back to what she’s doing. We enable regular people to contribute to real science.
The problem is that millions of people are eager to explore science but can’t find citizen science projects. So they’ve put together a database of projects.
By harnessing America’s greatest resource, they hope to improve the world.
Michael Smolens is founder of dotSUB is a revenue generating, global market leading, technology/media company. |
Michael Smolens on Removing Language as a Barrier to Cross Cultural Communication Using the Crowd
He starts with a slide saying the Tower of Babel needs to come tumbling down. He has impressive history, 9 startups.
There’s a huge problem with people communicating with each other, and we need to remove language as a barrier.
People have to learn a certain language if they want to learn something that’s only in a certain language, but technology means it doesn’t have to be that way.
Some examples: twitter had a video using dot SUB, it had 9.1 million views in a plethora of different languages and several other impressive stats.
One English film that was translated with tech had 91% of its views in other languages.
TED launched an open translation problem. 9,900 videos into 85 languages by 4500 volunteers who’d never subtitled before in a year.
Even though machine translation is fast and free, it’s not good enough.
He shows that you can watch a TED and select a transcript in virtually any language. Searchable Arabic transcription.
There’ll be translation on mobile devices in multiple languages in any font. You can watch children in a village speaking something that’s not even a real language, and see it translated into English and other languages.
plugthatwell.com has citizen scientists helping the Gulf in real time. Volunteer programmers have put together a game there, which shows how quickly you can develop these things that raise awareness.
Language does not have to be a barrier, Smolens is devoting his life to removing that barrier, and he wants to hear from people who want to help.
Itamar Arel is an Associate Professor of Electrical Engineering and Computer Science and Director of the Machine Intelligence Laboratory at the University of Tennessee. |
Itamar Arel: Is Deep-Layered Machine Learning the Catalyst for an Artificial General Intelligence Revolution?
I was distracted for the first bit of this talk, so definitely check out the slides and the video. But right now he’s talking about goal of learning to perceive and represent the world.
We know that learning is driven by rewards and reinforcements. Intelligence implies “strategic thinking.” Strategic involves not just short term rewards, but rewards down the line.
There’s been headway in deep learning like speech analytics and image recognition. They’ve been able to generate human level capabilities in these settings.
Some closing thoughts: progress toward AGI is being made. It’s the time to discuss moral implications, socioeconomic impact, and regulatory policies. We need to discuss these things now, not when AGI is happening. He thinks AGI will be here soon, he plans to demonstrate something to us in a year or two.
Ben Goertzel is CEO of AI software company Novamente LLC and bioinformatics company Biomind LLC; leader of the open-source OpenCog AI software project, Vice Chairman of Humanity+, and a Fellow of the IEET. |
IEET’s Ben Goertzel: The Future History of Artificial General Intelligence
Narrow AI is still dominant. Google, driving cars, and playing chess are some examples, but these are all task specific.
But narrow AI is useful. He’s used narrow AI to find genes that predict Parkinson’s.
In 2009, Biomind used AI to infer aging networks from gene expression data.
In 2010, AGI is on the rise. He talks about his project OpenCog.
There are some technical difficulties trying to get a video to play. We have such a wonderful audience, that somebody is able to bring up a min-VGA cord that fixes things.
Now we see a video of an AI dog. It’s learning about it’s environment, and can learn behaviors from reinforcement etc.
He’s been working at Xian University to advance his research. There’s a video of robot that is navigating around it’s environment. It can detect obstacles in its path. The cool thing is it’s using the same code that the dog uses in the virtual world.
There’s a bit more overhead, because in RL you need code to distinguish a chair instead of just using metadata.
What will happen in 2020? One of Ben’s goals is to make an AI toddler, which we could have in 2-3 years with adequate funding, but even without that we could have it in a decade.
A truly powerful artificial scientist that can uses bio instruments and read bio research papers will eventually emerge.
If we just had an AI as smart as a human biologist, but it could have access to all the bio data on the internet, who knows what it would discover?
In 2030, everything will all be on one big network. People will be jacked into network.
In 2040, as we all know, the Singularity is near. Kurzweil has said with “iron-clad certainty” we’ll have a Singularity in 2045. Ben thinks Kurzweil is roughly right.
But we won’t able to predict what an AI that smart will do, any more than a cockroach could predict the genes involved in longevity.
We’re going to be growing along with the Singularity. Ben doesn’t think it’s us vs. the machines, we’ll be part of them.
He’s written a new book, the Cosmist Manifesto which is now available (and parts of his manifesto have been on the IEET blog).