Amazon Offers $2.5M To Make Alexa Your Friend

Amazon has unveiled the Alexa Prize, a $2.5 Million purse for the first team to turn Alexa, the voice service that powers the Amazon Echo, into a ‘socialbot’ capable of, “conversing coherently and engagingly with humans on popular topics for 20 minutes”.

The Alexa Prize is only open to teams from colleges or universities, with the winning team taking home $500,000 USD, with $1M awarded to the team’s college or university in the form of a research grant. Of course, the Alexa Prize grants Amazon a perpetual, irrevocable, worldwide, royalty-free license to make use of the winning socialbot.

It may be argued the Alexa Prize is a competition to have a chat bot pass a Turning Test. This is a false equivalency; the Turing Test, as originally formulated, requires a human evaluator to judge between two conversation partners, one of which is a human, one of which is a computer. Additionally, the method of communication is text-only, whereas the Alexa Prize will make use of Alexa’s Text to Speech functionality. The Alexa Prize is not a Turing Test, but only because of semantics. If you generalize the phrase, ‘Turing Test’ to mean a test of natural language conversation, the Alexa Prize is a Turing Test.

This is not the first prize offered for a computer program that is able to communicate with a human in real time using natural language. Since 1990, the Loebner Prize, cosponsored by AI god Marvin Minsky, has offered a cash prize of $100,000 (and a gold medal) to the first computer that is indistinguishable from a human in conversation. Since 1991, yearly prizes have been awarded to the computer that is most like a human as part of the competition.

For any team attempting the enormous task of developing a theory of mind and consciousness, here are a few tips: don’t use Twitter as a dataset. Microsoft tried that, and their chatbot predictably turned racist. A better idea would be to copy Hackaday and our article-generating algorithm. Just use Markov chains and raspberry pi your way to arduino this drone.

Hackaday Prize Entry: Explore M3 ARM Cortex M3 Development Board

Even a cursory glance through a site such as this one will show you how many microcontroller boards there are on the market these days. It seems that every possible market segment has been covered, and then some, so why on earth would anyone want to bring another product into this crowded environment?

This is a question you might wish to ask of the team behind Explore M3, a new ARM Cortex M3 development board. It’s based around an LPC1768 ARM Cortex M3 with 64k of RAM and 512k of Flash running at 100MHz, and with the usual huge array of GPIOs and built-in peripherals.

The board’s designers originally aimed for it to be able to be used either as a bare-metal ARM or with the Arduino and Mbed tools. In the event the response to their enquiries with Mbed led them to abandon that support. They point to their comprehensive set of tutorials as what sets their board apart from its competition, and in turn they deny trying to produce merely another Arduino or Mbed. Their chosen physical format is a compact dual-in-line board for easy breadboarding, not unlike the Arduino Micro or the Teensy.

If you read the logs for the project, you’ll find a couple of videos explaining the project and taking you through a tutorial. They are however a little long to embed in a Hackaday piece, so we’ll leave you to head on over if you are interested.

We’ve covered a lot of microcontroller dev boards here in our time. If you want to see how far we’ve come over the years, take a look at our round up, and its second part, from back in 2011.

Earliest Recorded Computer Music Restored

You want old skool electronic music? How about 1951?

Researchers at the University of Canterbury in New Zealand have just restored what is probably the oldest piece of recorded, computer-generated music. Recorded in 1951, the rendition of “God Save The King”, “Baa-Baa Black Sheep” and “In The Mood” was produced by a computer built by none other Alan Turing and other researchers at the Computing Machine Research Laboratory in Manchester.

These phat beats were captured by the BBC for broadcast on an acetate disk that the researchers found in an archive. They sampled and restored the recording, fixing the rather poor quality recording to reproduce the squawky tones that the computer played. You can hear the restored recording after the break.

It halts apparently unexpectedly in the middle of a stanza, sounds essentially horrible, and goes out of tune on the higher notes. But you gotta learn to crawl before you can walk, and these are the equivalent of the grainy 8mm films of baby’s first steps. And as such, the record is remarkable.

Via ABC News

Continue reading “Earliest Recorded Computer Music Restored”

Boombox Doorjam Plays Your Theme Song When You Step in the Ring

Although many of us may have had childhood aspirations to be a famous wrestler in the WWE, not very many of us will ever realize those dreams. You can get close, though, if you have your own epic intro music theme that plays anytime you walk into a room. Although it’s not quite the same as entering a wrestling ring, [Matt]’s latest project will have you feeling just as good whenever you enter a room to your own theme song.

The core of the build consists of a boom box with an auxiliary input. The boom box is fed sound via a Raspberry Pi which also serves as the control center for the rest of the project. It runs Node.js and receives commands via websockets from a publicly accessible control server. The Pi is also running Spotify which allows a user to select a theme song, and whenever that user’s iBeacon is within range, the Pi will play that theme song over the stereo.

The project looks like it would be easy to adapt to any other stereo if you’re looking to build your own. Most of the instructions and code you’ll need are available on the project’s website, too. And, if you’re a fan of music playing whenever you open a door of some sort, this unique project is clearly the gold standard. It might even make Stone Cold Steve Austin jealous.

Hallucinating Machines Generate Tiny Video Clips

Hallucination is the erroneous perception of something that’s actually absent – or in other words: A possible interpretation of training data. Researchers from the MIT and the UMBC have developed and trained a generative-machine learning model that learns to generate tiny videos at random. The hallucination-like, 64×64 pixels small clips are somewhat plausible, but also a bit spooky.

The machine-learning model behind these artificial clips is capable of learning from unlabeled “in-the-wild” training videos and relies mostly on the temporal coherence of subsequent frames as well as the presence of a static background. It learns to disentangle foreground objects from the background and extracts the overall dynamics from the scenes. The trained model can then be used to generate new clips at random (as shown above), or from a static input image (as shown in pairs below).

Currently, the team limits the clips to a resolution of 64×64 pixels and 32 frames in duration in order to decrease the amount of required training data, which is still at 7 TB. Despite obvious deficiencies in terms of photorealism, the little clips have been judged “more realistic” than real clips by about 20 percent of the participants in a psychophysical study the team conducted. The code for the project (Torch7/LuaJIT) can already be found on GitHub, together with a pre-trained model. The project will also be shown in December at the 2016 NIPS conference.

Testing the Speed-of-Light Conspiracy

There are a number of ways to measure the speed of light. If you’ve got an oscilloscope and a few spare parts, you can build your own apparatus for just a few bucks. Don’t believe the “lies” that “they” tell you: measure it yourself!

OK, we’re pretty sure that conspiracy theories weren’t the motivation that got [Michael Gallant] to build his own speed-of-light measurement rig, but the result is a great writeup, and a project that includes one of our favorite circuits, the avalanche transistor pulse generator.

setupThe apparatus starts off with a very quickly pulsed IR LED, a lens, and a beam-splitter. One half of the beam takes a shortcut, and the other bounces off a mirror that is farther away. A simple op-amp circuit amplifies the resulting pulses after they are detected by a photodiode. The delay is measured on an oscilloscope, and the path difference measured with a tape measure.

If you happen to have a photomultiplier tube in your junk box, you can do away with the amplifier stage. Or if you have some really fast logic circuits, here’s another project that might interest you. But if you just want the most direct measurement we can think of that’s astoundingly accurate for something lashed up on breadboards, you can’t beat [Michael]’s lash-up.

Oh and PS: He got 299,000 (+/- 5,000) km/sec.