Deep Neural Networks can be pretty good at identifying images — almost as good as they are at attracting Silicon Valley venture capital. But they can also be fairly brittle, and a slew of research projects over the last few years have been working on making the networks’ image classification less likely to be deliberately fooled.
One particular line of attack involves adding particularly-crafted noise to an image that flips some bits in the deep dark heart of the network, and makes it see something else where no human would notice the difference. We got tipped with a YouTube video of a one-pixel attack, embedded below, where changing a single pixel in the image would fool the network. Take that robot overlords!
Or not so fast. Reading the fine-print in the cited paper paints a significantly less gloomy picture for Deep Neural Nets. First, the images in question were 32 pixels by 32 pixels to begin with, so each pixel matters, especially after it’s run through a convolution step with a few-pixel window. The networks they attacked weren’t the sharpest tools in the shed either, with somewhere around a 68% classification success rate. What this means is that the network was unsure to begin with for many of the test images — making it flip from its marginally best (correct) first choice to a second choice shouldn’t be all that hard.
It’s the story of the moment, isn’t it. As the price of Bitcoin continues on its wild and crazy rollercoaster ride, everyone’s talking about cryptocurrencies, and in almost mystical terms, about blockchains. Perhaps to be a little more accurate, we should report that they are talking about The Blockchain, a single entity which it seems is now the answer to all ills.
Of course, there is no single blockchain, instead blockchain technologies form the underpinnings of the cryptocurrency boom. Since little dollar signs seem to be buzzing around in front of everyone talking about that subject, it has attracted the attention of hordes of people with little understanding of it. APNIC have a good article aimed at those people: Don’t Get Caught Up In Blockchain Hype, which is worth a read even if you do understand blockchain technologies.
It makes the point that many large enterprises are considering investments in blockchain technologies, and lists some of the potential pitfalls that they may encounter. There may be a slight element of schadenfreude for some of the technically literate in seeing this in action, but given that such things can have consequences for those among us it’s too important to ignore.
As an analogy of a relatively clueless executive jumping on a tech-driven bandwagon, a software company of our acquaintance had a boss who decided in the heady days before the dotcom crash that the organisation would fully embrace open-source. Something to be welcomed, you might think, but given that the software in question was a commercially sensitive asset upon which all company salaries depended, it was fortunate that he listened to his developers when they explained to him exactly what open source entails.
Whether you are a blockchain savant or an uninterested bystander, it’s worth a read as you may sometime need its arguments to save someone from their own folly. If you fancy a simple example to help understand something of how blockchains work, we’ve got that covered for you.
Snark aside, to get you through the phase of inflated expectations and on down into disillusionment, we’re going to use just one word: “security”. (Are you disillusioned yet? We’re personally bummed out anytime anyone says “security”. It’s a lot like saying “taxes” or “dentist’s appointment”, in that it means that we’re going to have to do something unpleasant but necessary. It’s a reality-laden buzzkill.)
Researchers at the Hasso Plattner Institute have created “Scotty,” a so-called teleportation system. While the name is a clear homage to the famous Star Trek character, this is not the Sci-Fi teleporting you may be expecting. The system is composed of two 3D printers (they used a pair of MakerBot Replicators). The “sender” printer has a camera and built-in milling machine. It uses deconstructive scanning – taking the picture of an object’s layer, then grinding that layer down to expose the next layer – and then sends the encrypted data to a “receiver” printer with a RasPi to decrypt the data so that it can immediately print the object. The ultimate idea behind this is that there is only one object at the end of the process.
It’s a disservice describing Scotty as a teleporter. By the researchers’ definition of a teleporter, the lowly fax machine is on par with Scotty – and it doesn’t destroy the original. The researchers claim that this destructive-reconstuctive method preserves the uniqueness of a given object, as long as any sentimentality. We can agree with the unique aspect: the less copies of something means it retains it intrinsic value in the marketplace. The sentimentality – not so much. We’ve all had a moment in our lives where a treasured item of ours, worthless to everyone else, was destroyed. Either we’d get a replacement or someone else would give us one to silence our wailing, but it wasn’t quite the same. If you could clone your dead pet, subconsciously you’d know it’s not going to be the same Fluffy. It’s that exact thing, atoms and all, that has the emotional attachment. Trying to push that psychological perspective onto Scotty’s purpose is irksome.
[HyPe] over at the Natural User Interface Group developed this concept as part of his Master’s Degree in Industrial Design. This suitcase sized projector and computer allows people to have a 60″ multitouch screen available wherever there is a large enough surface. The current software is designed for ad-hoc meetings about large-scale construction plans. The rolling case includes a short-throw projector and webcam. Just set it on top of your work surface, lift the lid, and it’s ready to go.