Carnegie Mellon’s Mayhem AI takes home $2 million from DARPA’s Cyber Grand Challenge

Carnegie Mellon’s Mayhem AI takes home million from DARPA’s Cyber Grand Challenge

The 2016 DARPA Cyber Grand Challenge has concluded, and the winning team — Carnegie Mellon University’s ForAllSecure — is taking home the $2 million grand prize. What’s that you say — what’s the Cyber Grand Challenge? I’m glad you asked!

Like the other challenges the Defense Department’s R&D arm has offered, the CGC is focused on autonomy — but where the original Grand Challenge and Robotics Challenge were about intelligently navigating the real world, this one is about operating in a threat-filled internet.

Seven teams were invited to Las Vegas to compete on the floor in a 96-round game of “Capture the Flag.” It’s a time-tested competitive hacking game in which teams are assigned servers which must perform certain tasks while constantly being fed new code filled with bugs, security holes, and inefficiencies. Teams must protect their own data while attempting to access that of the others — much like real-life CTF.

The difference in this game is that the players in the game were totally autonomous. Normally a human would be looking at and correcting the code, choosing whether and whom to attack, and so on — but for the CGC, all that has to be done by the system.

The idea is, of course, to produce systems that can patch themselves, watch for intrusions, and so on, with minimal human interaction. It’d be nice to know that your computer is looking out for itself.

After some 8 hours of battle at a ballroom in the Paris hotel (some highlights), the victor emerged: ForAllSecure’s “Mayhem,” with second place going to TechX’s Xandra. That’s $2 million and $1 million respectively, on top of the $750,000 each of the 7 finalists already received.


What’s more, Mayhem is, as of this writing, the first autonomous CTF system to play against humans. The team was invited to enter the CTF tournament at the neighboring DEF CON, and the game is afoot.

The manager of the CGC program, Mike Walker, promptly threw Mayhem under the bus.

“I don’t expect Mayhem to finish well,” he said in the DARPA press release. “This competition is played by masters and this is their home turf. Any finish for the machine save last place would be shocking.”

Not the nicest thing to say about a champion AI that just took first place in an incredibly sophisticated virtual game, but he probably knows what he’s talking about. I’ll update this post when we find out how Mayhem fares against flesh and blood.

Featured Image: DARPA CGC
Source: TechCrunch

NASA funds long shots aiming to make electric and eco-friendly aircraft

NASA funds long shots aiming to make electric and eco-friendly aircraft

One of these days we’re going to see aircraft go electric, just as cars are — but not for a while. In the meantime, we can’t slack off when it comes to the research that will make it possible. NASA has announced five research projects that may help make planes more efficient and green.

It’s part of the organization’s Transformative Aeronautics Concepts Program, which has dozens of such projects aimed at improving or (as the name suggests) transforming aviation. These five were picked from a series of pitches made to the TACP team, Startup Battlefield style (or close, anyway).

The thing preventing electric aircraft from taking off, so to speak, boils down to a lack of energy: even the best batteries can’t store nearly as much power as an equivalent volume of jet fuel. We can make better use of the energy we can store by making aircraft lighter and their propulsion more efficient, but ultimately the batteries themselves need to get better.

One project is aimed at accomplishing that last goal with the creation of Lithium-Air batteries, which use oxygen as the active electron carrier, sucking it in when they’re being discharged and expelling it when charging. Li-Air batteries could theoretically achieve fossil fuel-like energy density, but there are serious technical challenges — ones the NASA team will be looking at.

Replacing the battery with a fuel cell is another option; one project is looking into the possibility of a super-efficient cell that would combine oxygen from the air with the hydrogen from jet fuel to produce electricity, then use the exhaust to drive a turbine. It would still be using fossil fuels, but it would be extracting more energy from them and doing it in a much cleaner way.

The electric engine being driven by that energy could also be more efficient; 3D printing might be a solution. Improved additive manufacturing techniques may enable smaller, more power-dense motors — or if not, at least make them lighter.

Airplane Skyline Horizon Flight Cloud ConceptAnother way to save weight is to lose some off the body of the plane itself. The vertical tail fin is used to center the plane, but it also adds drag. NASA is looking into the possibility of using adjustable winglets that can serve as stabilizers during take-off and landing, but flatten out for better aerodynamics during flight. This would reduce the need for a large tail fin, so the plane could be both lighter and present a lower profile.

The last project isn’t quite green, but it’s still interesting. Unmanned aerial vehicles currently need to be operated within radio line of sight; satellite tracking would be better, but the antennas required are bulky. Researchers are investigating the possibility of an aerogel-based antenna that conforms closely to the contours of the aircraft yet enables transmission and receiving in multiple directions.

NASA doesn’t expect all or even any of the projects to bear fruit — “Is failure an option? It depends on your definition of failure,” said the TACP manager, Doug Rohn, in the news release. Really, the point isn’t to make the things described above, but to determine whether they realistically can be made, and the answer may well be no — at least for now.

Featured Image: NASA
Source: TechCrunch

App lets visually impaired in India hear books in their native language

App lets visually impaired in India hear books in their native language

For the millions upon millions of visually impaired people in India, it can be difficult getting hold of the audiobook they want in the language they need it in. A project from Carnegie Mellon University and partners aims to fix that with a free, easily extensible Android app that can be quickly trained to read texts aloud in local languages.

The app, Hear2Read, had its first release today, supporting Tamil, with Hindi, Bengali, Punjabi, and other languages and dialects coming over the course of the year. A few hours of talking from a native speaker is the raw data, which is then fed into a machine learning system.

“Each language is different and historically TTS systems have been done one at a time. We looked at commonalities of Indian languages and developed tools to apply the same technology to multiple languages,” said Suresh Bazaj, founder of the project, in a CMU blog post.


The resulting language database is small enough that it can be stored on the phone, meaning no internet connection is needed to translate texts. It’ll run on low-end phones, as well, an important factor in a country where budget devices and spotty connectivity are the rule. (You’ll need Android 4.1 or higher, though.)

The Hear2Read software also integrates with Android’s built-in accessibility functions, letting browsers, email apps, and others integrate text to speech.

More info on Hear2Read, the app, and the partners that helped make it possible can be found at the company’s website.

Featured Image: Liba Taylor / Getty Images
Source: TechCrunch

This neural network tries to tell if you’re being sarcastic online

This neural network tries to tell if you’re being sarcastic online

One of the perils of text-based communication is the lack of cues that clearly signal irony being employed — no doubt we’ve all had our own mishaps behind this particular issue. Researchers from Portugal have had enough, though, and built a neural network that tries to determine whether you or your virtual interlocutor is being sarcastic.

It’s not just to avoid awkward missteps in our everyday conversations online, though: computationally determining the tone and meaning of a given message is important for many things.

You can’t do accurate sentiment analysis, for instance, if you don’t know when someone is kidding around when they say they love or hate something. And knowing the difference between an affirmative “great!” and a sarcastically disappointed one is important for natural language processing.

But it’s not en easy problem. After all, the same exact sentence and punctuation may mean very different things coming from different speakers. Imagine, for example, “Make America great again” being tweeted by a Trump supporter and one who opposes him. Same words, wildly different meanings.

“Lexical cues alone are insufficient to discern ironic intent,” reads the introduction of the paper, written by Silvio Amir at the University of Lisbon. “Appreciating the context of utterances is critical for this, even for humans.”

To learn that context, the paper describes a method by which the neural network finds the user’s “embeddings” — i.e. contextual cues like the content of previous tweets, related interests and accounts, and so on. It uses these various factors to plot the user with others, and (ideally) finds that they form relatively well-defined groups.

For instance, these little clouds of users are the result of looking at what politicians a set of users followed on Twitter.


If the sentiment of the tweet seems to disagree with the bulk of what is expressed by similar users, there’s a good chance sarcasm is being employed.

Combined with textual factors that give a hint of irony, this approach proved more capable than other models out there — barely. It was right about 87 percent of the time versus 85 in the other system. However, the neural network setup required far less manual configuration and monitoring, and could be more easily extended for deployment on different social networks.

We’re still a little short of a universal sarcasm indicator, but we’ll get there. Amir’s paper is due to be presented at CoNLL, a natural language learning conference taking place next month.

Featured Image: majcot/Shutterstock
Source: TechCrunch

Latest Olympics media rules nix GIFs, Vines, and streaming apps

Latest Olympics media rules nix GIFs, Vines, and streaming apps

rio_sparkleThe Olympics Committee has always been restrictive in who can use what footage and when, but a couple new rules introduced for the 2016 games take things even further. The official social media guidelines now prohibit streaming apps, Vines, and even the lowly GIF.

Part of the new restriction appears in the official broadcast rules (PDF), under “Internet and Mobile Platforms.”

…the use of Olympic Material transformed into graphic animated formats such as animated GIFs (i.e. GIFV), GFY, WebM, or short video formats such as Vines and others, is expressly prohibited.

“Olympic Material” is defined as pretty much any images of the games, whatever the source.

Then, in the FAQ for the social and digital media guidelines (PDF):

Broadcasting images via live-streaming applications (e.g. Periscope, Meerkat) is prohibited inside Olympic venues.

The versions of these documents updated for the 2014 games in Sochi don’t have any comparable language, or at least nothing this specific. A possible exception is the “Photographer’s Undertaking,” which states:

The dissemination of moving images or sound  captured  in  an  Olympic venue, through any media, including display on the Internet, Mobile Platform and other interactive media or electronic medium, is strictly prohibited.

But that document isn’t dated, so it’s hard to say when the IOC might have made that language more inclusive.

In general the rule boils down to: Video taken in the Village or at a venue is to be taken by accredited media and used within the broadcast rules, or else kept for personal purposes only. That outlawed streaming in effect, but the new rules make it more clear. GIFs weren’t mentioned, let alone GIFVs, GFYs, or any other new custom looping image format.

There's lots more where this came from.

There’s lots more where this came from.

Of course, any attempt to crack down on the inevitable meme-making that accompanies such a major media event is doomed to hilarious failure.

Hell, even the pros are likely to get tut-tutted by the IOC for not following every rule to the letter. Just try making sure your broadcast, front page, segment, vlog, or whatever meets all the restrictions set out here.

Featured Image: IOC
Source: TechCrunch

Banking trojan Zeus Panda shambles into Brazil ahead of Olympics

Banking trojan Zeus Panda shambles into Brazil ahead of Olympics

It seems there’s no limit to the perils being faced by athletes at the Rio 2016 Olympics: not just their competitors, but toxic water, poor accommodations, and impressive mobile bills. Add malicious pandas to the menu — virtual ones, at least. A nasty Trojan known as Zeus Panda has made its way to the Olympic host just in time for an influx of visitors.

IBM’s X-Force Research discovered that the Trojan, a variant of the Zeus variety that’s been kicking around for the last few years, had spread to Brazil in July. Zeus and its relatives — the pantheon, if you will — target transactions such as online banking logins, payment portals, and Bitcoin exchanges. Basically anywhere they can slip in and steal a login with the power to approve more such transactions.

The specifics of this Panda variant are discussed here by Arbor Networks. It seems to be largely the same as previous malware in this lineage, albeit modified to frustrate the latest detection packages and target Brazilian banks and services specifically.

Deployment appears to be done professionally, as well — the Trojan is likely being sold in the usual nooks of the Dark Net where such sundries are found. So far the preferred delivery mechanism has been Word docs with embedded code that activates the malware, but other vectors are of course in play as well. One-time passwords for two-factor authentication are acquired via a fraudulent 2FA pop-up that forwards that data on to the hackers.

IBM notes that while the software behind Zeus Panda isn’t particularly new, nor is the cybercrime scene in Brazil particularly advanced — so Panda may be as a wolf (or rather bear) among lambs.

You can avoid Trojans like this by not opening strange attachments or following suspicious links, but it can also be addressed at a systematic level by the banks being targeted. The methodology of this malicious software is well understood, but it takes vigilance (and savvy IT) to keep it at bay.

Featured Image: Leungchopan/Shutterstock
Source: TechCrunch

Average broadband speed in US rises above 50 megabits for the first time

Average broadband speed in US rises above 50 megabits for the first time

Speedtest has released its mid-year broadband speeds report, and there’s actually quite a bit of good news in there. Speeds are steadily increasing despite machinations within the industry, and in fact the average in the U.S. just went north of 50 megabits for the first time ever.

That average, 54.97 megabits per second is 42 percent higher than the same period last year, and upload jumped even more — 18.88 is 51 percent higher year over year. This is all based on the 8 million or so daily tests conducted on Speedtest’s website and apps, by the way, so the data is pretty sound.

Comcast Xfinity took the honors for fastest speed on average, but its 125 megabits wasn’t that much higher than the competition: Cox with 118 and Spectrum with 114.


Does that last name fail to ring a bell? Spectrum is the new company formed by Charter, Time Warner Cable, and Bright House Networks when the first bought the other two earlier this year.

“Market consolidation by large ISPs doesn’t typically bode well for innovation and increased speeds,” Speedtest observes, although TWC and Bright House did see performance improvements in the last 6 months, so perhaps this will be an exception.

Fiber providers like Google and Verizon (which owns Aol, which owns TechCrunch, by the way) are continuing to roll out their networks, pushing services using traditional connections to provide competitive speeds — and, critically, bringing solid broadband to remote areas where it has been hard to come by for decades.


On mobile, Verizon and T-Mobile are tied for first place with 21 megabits and change download speed on average, though the latter beats the competition by a long shot with upload speeds averaging 11.59 megabits. Poor Sprint, though.

Speedtest notes that the mobile world has seen serious competition over the last couple years, with consumers the gainers as all the networks invest in their networks and push each other on prices.

It’s good stuff, but don’t lose track of the fact that our speeds still lag behind many other countries, and many companies exert ongoing strangleholds on cities or neighborhoods. Plenty of work to do!

Featured Image: Flickr UNDER A CC BY 2.0 LICENSE
Source: TechCrunch

The White House requested input on artificial intelligence, and IBM’s response is a great AI 101

The White House requested input on artificial intelligence, and IBM’s response is a great AI 101

The field of artificial intelligence is so huge, and the potential applications so numerous, that it would be folly to try to explain it all in one — no, wait, IBM just did.

So back in June, the White House published a formal request for information, or RFI, about the possibilities and risks of AI.

“The views of the American people, including stakeholders such as consumers, academic and industry researchers, private companies, and charitable foundations, are important to inform an understanding of current and future needs for AI in diverse fields,” the summary read.

This was followed by a numbered list of specific topics to be addressed, from where research should be focused to the potential for abuse or for public good.

IBM took the suggestion and ran with it: each topic is given a solid enough explanation that after reading through the whole thing, you’ll more than likely be the best-informed person in the room on it. (Depends on the room, of course.)

There’s no sense in my summarizing it here; there’s too much and it’s mostly to the point, non-controversial, informative stuff. That said, this is a good quote:

We believe that many of the ambiguities and inefficiencies of the critical systems that facilitate life on this planet can be eliminated. And we believe that AI systems are the tools that will help us accomplish these ambitious goals.

It’s a fundamentally optimistic look, and a bit IBM-focused, of course, but beyond that it really is worth reading to get a sense of the current state, future, and risks of AI. Don’t skip the “see more here” links — that’s where the good stuff is.

Featured Image: Omelchenko/Shutterstock
Source: TechCrunch

Watch wall-walking spiderbots weave ‘impossible’ structures with carbon fiber

Watch wall-walking spiderbots weave ‘impossible’ structures with carbon fiber

Normally, the notion of setting a pair of spiderbots free in your house would seem alarming. But if you knew that they were weaving you a strange and mathematical hammock while you were gone? Suddenly that fear turns to wonder — and eventually, a nap.

The Mobile Robotic Fabrication System for Filament Structures was developed at the University of Stuttgart’s Institute for Computational Design. The concept behind it is “swarm construction”: many small robots working together to produce a single item. In this case, the bots are like wall-hugging Roombas weaving a carbon fiber nook that looks like it sprang fully formed out of a Spirograph. This project in particular was built on the work of ICD graduate student Maria Yablonina.

“We are only at the very beginning of exploring the true architectural potential of this fabrication system,” ICD director Achim Menges told Dezeen. “But we are convinced that its main advantage is that you can build entirely new structures that would be impossible to materialize otherwise.”

[embedded content]

It is, at the very least, very cool to watch. The robots have spools of carbon fiber thread that they pass back and forth after affixing to points on the wall, like two hands putting together a cat’s cradle. (“See the cat? See the cradle?”)

Being able to crawl along the walls and interact with one another in swarm fashion could indeed produce some truly unique structures and utilize spaces in quite a different way.

Menges is inspired by nature’s economy and ingenuity; his team has based buildings on insect and lobster shells, and the bots described here clearly are influenced by arachnids and other silk-weaving animals. (There’s an amazing episode of Life in the Undergrowth dedicated to them, incidentally, which you should watch.)

The plan now is to increase the number of robots and allow them to maneuver and attach the fibers to other surfaces, like ceilings or curved walls. That should allow for even more unusual creations. Whether you’d want to live or work in a mind-warping expanse of black yarn, of course, is a separate question altogether.

Featured Image: University of Stuttgart
Source: TechCrunch

Living Computer Museum restores Xerox Alto and debuts new emulator

Living Computer Museum restores Xerox Alto and debuts new emulator

Anyone interested in the history of personal computing will surely have heard of the Xerox Alto, but when’s the last time you got to play with one? It’s been a while even for Paul Allen — long enough that he decided to have a couple restored at the Living Computer Museum in Seattle.

Don’t worry if that’s too far for you to visit (it’s right down the road for me, just saying) — the team also created a highly accurate emulator so you can experience Xerox’s UI revolution from the safety of your home.

Allen, who founded the museum, gives a brief history of the Alto and its tremendous effect on personal computing (far too extensive to go into here) before showing off the two restored units.

xerox alto 1“It’s one thing to read about a true breakthrough, something else to see it in action,” he wrote.

The Alto used Ethernet to communicate with other computers (another innovation), and the restoration team made sure that worked, too — and not just so the two could talk to each other. They also made a 3-megabit Ethernet bridge that allows the Altos to communicate with modern computers. That is, if a new PC can understand the old-timers’ crazy lingo.

Y Combinator has also been working on getting an old Alto working, and has documented the process pretty extensively if you’re interested in the details (and why wouldn’t you be?). Lots of posts to read through: day 1, day 2, day 3, and day 4 — and more to come.

Their work was also a collaboration with the museum; there will soon even be an gateway to allow Altos to network over the internet. Playing MazeWar with someone on another continent will be an original experience, but likely one that Xerox engineers would have predicted back in the ’70s.


Lenna, of course. The Pinball game cleverly uses the mouse as the ball (you can see it blazing by just right of center).

The time was also ripe for the LCM to release its emulator, ContrAlto, which according to Allen “simulates the Alto at the microcode level, providing a very accurate simulation of the original computer.”

There’s already an Alto emulator, Salto, but its author admits it’s “kind of buggy” and not feature complete. ContrAlto appeared, in my brief testing, to work quite well — games and apps loaded up, though of course some crashed, owing to my not knowing the boot order of this or that. You can work it out and do a little early WYSIWYG editing in Bravo or play Breakout. You’ll need to start the OS and load the disk – drop a ? in the command line to get a list of executables.

If you’re from the area or just visiting, drop by the Living Computer Museum and check out the Altos and dozens of other interesting (and working) artifacts from computing history. It’s in SoDo and free on 1st Thursdays.

Featured Image: Living Computer Museum
Source: TechCrunch