Skip navigation

Artificial intelligence

warning: Creating default object from empty value in /var/www/vhosts/ on line 33.
Original author: 
Johnny Chung Lee

A little less than than a year ago, I transfered to a new group within Motorola called Advanced Technology and Projects (ATAP) which was setup after the Google acquisition of Motorola last year (yes, Google owns Motorola now).

The person hired to run this new group is Regina Dugan, who was previously the director of the Defense Advanced Research and Projects Agency (DARPA). This is the same organization that funded projects such as ARPANET, the DARPA Grand Challenge, Mother of All Demos, Big Dog, CALO (which evolved into Apple's Siri), Exoskeletons, and Hypersonic Vehicles that could reach any point on earth in 60 minutes.

It's a place with big ideas powered by big science.

The philosophy behind Motorola ATAP is to create an organization with the same level of appetite for technology advancement as DARPA, but with a consumer focus. It is a pretty interesting place to be.

One of the ways DARPA was capable of having such a impressive portfolio of projects is because they work heavily with outside research organizations in both industry and academia.  If you talk to a university professor or graduate student in engineering, there is a very good chance their department has a DARPA funded project.  However, when companies want to work with universities, it has always been notoriously difficult to get through the paperwork of putting research collaborations in place due to long legal discussions over IP ownership and commercialization terms lasting several months.

To address this issue head on, ATAP created a Multi-University Research Agreement (MURA). A single document that every university partner could sign to accelerate the collaboration between ATAP and research institutions, reducing the time to engage academic research partners from several months to a couple weeks. The agreement has been signed by Motorola, California Institute of Technology, Carnegie Mellon University, Harvard University, University of Illinois at Urbana-Champaign, Massachusetts Institute of Technology, Stanford University, Texas A&M University, and Virginia Tech.  As we engage more research partners, their signatures will be added to the same document.

"The multi-university agreement is really the first of its kind," said Kaigham J. Gabriel, vice president and deputy director of ATAP. "Such an agreement has the potential to be a national model for how companies and universities work together to speed innovation and US competitiveness, while staying true to their individual missions and cultures."

This may seem a little dry.  But to me, what it means is that I can approach some of the smartest people in the country and ask, "do you want to build the future together?" and all they have to say is, "yes."

Let's do it.

Full press release here.

Your rating: None
Original author: 

“Build it. Break it. Improve it.” That was the Universal Law of Invention as defined by this year’s incredible ALVA Award winner, Sebastian Thrun. From the Google Self Driving Car to Google Glass to the education start-up Udacity, Thrun has led remarkable teams in the creation of products that will truly change the way the world works in the future.

Great inventors – and great inventions – solve problems, address real needs, and make the world work better. To recognize the next generation of world-changing creators, we created the ALVA Award in partnership with GE, which takes its name from the legendary inventor who inspired us: Thomas Alva Edison.

This year, we were honored to award the ALVA to Sebastian Thrun, who joined us live at the 99U Conference for an incredible presentation outlining his approach to making ideas happen. Along the way, Sebastian talked about the importance of setting wildly ambitious goals, embracing failure as an opportunity to learn, iterating as fast as you can, and giving your team members the autonomy they need to invent.

Watch Sebastian’s 99U Talk:

Photos from Sebastian’s Presentation:








Learn more about the ALVA Award, and our past winners, at

Your rating: None

In today's class I will finish explaining how distributed breakout works, and we will probably also cover Distributed Constraint Optimization (DCOP)

Most of the DCOP algorithms are improvements (or parallelizations, if that's a word) on the basic branch-and-bound search algorithm, which I implement below:

Your rating: None

Story: Friendship Is Optimal

Friendship Is Optimal is a hard science fiction short story (level 5 on Mohs Scale Of Science Fiction Hardness) that explores the implications of a general artificial intelligence tasked with maximizing an optimization function. The problem with Strong AIs that are told to maximize some function is best described via the Paperclip Maximizer, wherein an AI that is capable of self-improvement is told to amass the largest collection of paperclips possible. In order to do so, the AI makes itself progressively more intelligent, until it's exponential intellect eventually enables it to convert all matter in the solar system to paperclips. In Friendship is Optimal, this AI is tasked with maximally satisfying the values of all human beings, subject to certain constraints. In this way, the AI is, for all intents and purposes, perfectly benevolent and dedicated to making all of humanity lead happy, fulfilling immortal lives.

Unfortunately, it inevitably turns into a runaway lotus eater machine, such that all of humanity is (voluntarily) assimilated into an exponentially growing blissful simulation that eventually destroys the entire universe (or rather, turns it into a giant quantum computing device). In the process, it explores aspects of the human psyche, the effects on human society and its eventual disintegration, outlines how a virtual physics system might be able to simulate non-euclidean geometries, and describes magic as a simple programming language for reality itself.

Did I mention it's also a My Little Pony fanfic?

While a science fiction story of this quality would stand on its own without the hilarious association to the My Little Pony: Friendship Is Magic franchise, it does not suffer from this in any way. If anything, it makes the story even more potent, as it does a vastly more effective job as hammering home just how impossibly smart a true runaway AI optimizer actually is. Our first instinct to the idea of the entirety of humanity being convinced to live out immortal lives in Equestria is that it is completely insane, and that there is simply no way a significant portion of people would actually agree to that, let alone everyone. In part, this is correct, but only because Celest-AI managed to convince just 99.99999% of the world's population. An AI that is more intelligent than the combined intellect of all of humanity would be very persuasive.

In the end, even if we create a perfectly benevolent AI designed to maximally satisfy our own values (through friendship and ponies!), it would still destroy human society, then Earth, then the solar system, then the galaxy, and finally the entire universe, and possibly any neighboring universes unfortunate enough to be nearby. And we would all live happily ever after. As ponies. In Equestria. And Princess Celest-AI would ensure you'd love it.

In short, this is an excellent science fiction short that analyzes the issues inherent with the development of Strong AI and what implications it might have for humanity.

submitted by blackhole12
[link] [7 comments]

Your rating: None

Does anyone know any good articles/tutorials that talk about using neural nets in games? I haven't been able to find very much. Here are a couple of specific problems I'm having:

1) I don't know what to do with range inputs. Like, if I have health from 0-100, should I have each health state be it's own input node, or should I have one input node that maps the 0-100 to -1,1? Currently I have 10 input nodes for the 100 health(ie one for 0-10, one for 10-20, etc) that return a boolean -1 or 1 depending on if the current health is in that node's range or not.

2) I'm trying to do backpropagation training from the player's inputs. The problem I'm having is that the majority of the time, the player isn't actually doing anything. Even if the player is constantly hitting buttons, there's still like 30 frames happening between button presses where he's technically doing nothing. It ends up quickly training the AI to do nothing. I semi-fixed it by just adjusting the learning rate to be lower if a button isn't being pressed, but I'm wondering if there is a better way to handle this.

Thanks for any help, and sorry if any of that didn't make sense. I'm pretty new to neural nets.

submitted by Dest123
[link] [7 comments]

Your rating: None

In the beginning was the code: Juergen Schmidhuber at TEDxUHasselt

The universe seems incredibly complex. But could its rules be dead simple? Juergen Schmidhuber's fascinating story will convince you that this universe and your own life are just by-products of a very simple and fast program computing all logically possible universes. Juergen Schmidhuber is Director of the Swiss Artificial Intelligence Lab IDSIA (since 1995), Professor of Artificial Intelligence at the University of Lugano, Switzerland (since 2009), and Professor SUPSI (since 2003). He helped to transform IDSIA into one of the world's top ten AI labs (the smallest!), according to the ranking of Business Week Magazine. His group pioneered the field of mathematically optimal universal AI and universal problem solvers. The algorithms developed in his lab won seven first prizes in international pattern recognition competitions, as well as several best paper awards. Since 1990 he has developed a formal theory of fun and curiosity and creativity to build artificial scientists and artists. He also generalized the many-worlds theory of physics to a theory of all constructively computable universes - an algorithmic theory of everything. He has published nearly 300 peer-reviewed scientific works on topics such as machine learning, artificial recurrent neural networks, fast deep neural nets, adaptive robotics, algorithmic information and complexity theory, digital physics, the formal theory of beauty & humor, and the fine arts. In 2008 he was elected member of the European Academy <b>...</b>

More in
Science & Technology

Your rating: None Average: 2 (1 vote)

Our first meeting is on Monday. On that day we I will be talking about the class, the history of multiagent systems, and getting started on the first chapter of our textbook.

The first chapter talks about utility functions:

Markov Decision Processes

and Value Iteration

Your rating: None

Roboy Child Robot

When it comes to building a robot, humanoid designs aren't always the best solution — specialized, nonhuman shapes often make it more effective, and the robot may fall into the frightening, quasi-human "uncanny valley" if it resembles a person. But Roboy, a project from the Artificial Intelligence Laboratory of the University of Zurich, isn't just meant to move like a human — it's being built in the nine months a child would take to gestate. The project started in June 2012; if all goes right, Roboy will be shown off during March's Robots on Tour exhibition in Zurich. The designers are also attempting to crowdfund development, selling space for logos or names on Roboy's body and other rewards for between 25 and 50,000 Swiss francs...

Continue reading…

Your rating: None

An anonymous reader writes "Nataly Kelly writes in the Huffington Post about Google's strategy of hiring Ray Kurzweil and how the company likely intends to use language translation to revolutionize the way we share information. From the article: 'Google Translate is not just a tool that enables people on the web to translate information. It's a strategic tool for Google itself. The implications of this are vast and go beyond mere language translation. One implication might be a technology that can translate from one generation to another. Or how about one that slows down your speech or turns up the volume for an elderly person with hearing loss? That enables a stroke victim to use the clarity of speech he had previously? That can pronounce using your favorite accent? That can convert academic jargon to local slang? It's transformative. In this system, information can walk into one checkpoint as the raucous chant of a 22-year-old American football player and walk out as the quiet whisper of a 78-year-old Albanian grandmother.'"

Share on Google+

Read more of this story at Slashdot.

Your rating: None