Tag Archives: running
#431238 AI Is Easy to Fool—Why That Needs to ...
Con artistry is one of the world’s oldest and most innovative professions, and it may soon have a new target. Research suggests artificial intelligence may be uniquely susceptible to tricksters, and as its influence in the modern world grows, attacks against it are likely to become more common.
The root of the problem lies in the fact that artificial intelligence algorithms learn about the world in very different ways than people do, and so slight tweaks to the data fed into these algorithms can throw them off completely while remaining imperceptible to humans.
Much of the research into this area has been conducted on image recognition systems, in particular those relying on deep learning neural networks. These systems are trained by showing them thousands of examples of images of a particular object until they can extract common features that allow them to accurately spot the object in new images.
But the features they extract are not necessarily the same high-level features a human would be looking for, like the word STOP on a sign or a tail on a dog. These systems analyze images at the individual pixel level to detect patterns shared between examples. These patterns can be obscure combinations of pixel values, in small pockets or spread across the image, that would be impossible to discern for a human, but highly accurate at predicting a particular object.
“An attacker can trick the object recognition algorithm into seeing something that isn’t there, without these alterations being obvious to a human.”
What this means is that by identifying these patterns and overlaying them over a different image, an attacker can trick the object recognition algorithm into seeing something that isn’t there, without these alterations being obvious to a human. This kind of manipulation is known as an “adversarial attack.”
Early attempts to trick image recognition systems this way required access to the algorithm’s inner workings to decipher these patterns. But in 2016 researchers demonstrated a “black box” attack that enabled them to trick such a system without knowing its inner workings.
By feeding the system doctored images and seeing how it classified them, they were able to work out what it was focusing on and therefore generate images they knew would fool it. Importantly, the doctored images were not obviously different to human eyes.
These approaches were tested by feeding doctored image data directly into the algorithm, but more recently, similar approaches have been applied in the real world. Last year it was shown that printouts of doctored images that were then photographed on a smartphone successfully tricked an image classification system.
Another group showed that wearing specially designed, psychedelically-colored spectacles could trick a facial recognition system into thinking people were celebrities. In August scientists showed that adding stickers to stop signs in particular configurations could cause a neural net designed to spot them to misclassify the signs.
These last two examples highlight some of the potential nefarious applications for this technology. Getting a self-driving car to miss a stop sign could cause an accident, either for insurance fraud or to do someone harm. If facial recognition becomes increasingly popular for biometric security applications, being able to pose as someone else could be very useful to a con artist.
Unsurprisingly, there are already efforts to counteract the threat of adversarial attacks. In particular, it has been shown that deep neural networks can be trained to detect adversarial images. One study from the Bosch Center for AI demonstrated such a detector, an adversarial attack that fools the detector, and a training regime for the detector that nullifies the attack, hinting at the kind of arms race we are likely to see in the future.
While image recognition systems provide an easy-to-visualize demonstration, they’re not the only machine learning systems at risk. The techniques used to perturb pixel data can be applied to other kinds of data too.
“Bypassing cybersecurity defenses is one of the more worrying and probable near-term applications for this approach.”
Chinese researchers showed that adding specific words to a sentence or misspelling a word can completely throw off machine learning systems designed to analyze what a passage of text is about. Another group demonstrated that garbled sounds played over speakers could make a smartphone running the Google Now voice command system visit a particular web address, which could be used to download malware.
This last example points toward one of the more worrying and probable near-term applications for this approach: bypassing cybersecurity defenses. The industry is increasingly using machine learning and data analytics to identify malware and detect intrusions, but these systems are also highly susceptible to trickery.
At this summer’s DEF CON hacking convention, a security firm demonstrated they could bypass anti-malware AI using a similar approach to the earlier black box attack on the image classifier, but super-powered with an AI of their own.
Their system fed malicious code to the antivirus software and then noted the score it was given. It then used genetic algorithms to iteratively tweak the code until it was able to bypass the defenses while maintaining its function.
All the approaches noted so far are focused on tricking pre-trained machine learning systems, but another approach of major concern to the cybersecurity industry is that of “data poisoning.” This is the idea that introducing false data into a machine learning system’s training set will cause it to start misclassifying things.
This could be particularly challenging for things like anti-malware systems that are constantly being updated to take into account new viruses. A related approach bombards systems with data designed to generate false positives so the defenders recalibrate their systems in a way that then allows the attackers to sneak in.
How likely it is that these approaches will be used in the wild will depend on the potential reward and the sophistication of the attackers. Most of the techniques described above require high levels of domain expertise, but it’s becoming ever easier to access training materials and tools for machine learning.
Simpler versions of machine learning have been at the heart of email spam filters for years, and spammers have developed a host of innovative workarounds to circumvent them. As machine learning and AI increasingly embed themselves in our lives, the rewards for learning how to trick them will likely outweigh the costs.
Image Credit: Nejron Photo / Shutterstock.com Continue reading
Posted in Human Robots
#431159 How Close Is Turing’s Dream of ...
The quest for conversational artificial intelligence has been a long one.
When Alan Turing, the father of modern computing, racked his considerable brains for a test that would truly indicate that a computer program was intelligent, he landed on this area. If a computer could convince a panel of human judges that they were talking to a human—if it could hold a convincing conversation—then it would indicate that artificial intelligence had advanced to the point where it was indistinguishable from human intelligence.
This gauntlet was thrown down in 1950 and, so far, no computer program has managed to pass the Turing test.
There have been some very notable failures, however: Joseph Weizenbaum, as early as 1966—when computers were still programmed with large punch-cards—developed a piece of natural language processing software called ELIZA. ELIZA was a machine intended to respond to human conversation by pretending to be a psychotherapist; you can still talk to her today.
Talking to ELIZA is a little strange. She’ll often rephrase things you’ve said back at you: so, for example, if you say “I’m feeling depressed,” she might say “Did you come to me because you are feeling depressed?” When she’s unsure about what you’ve said, ELIZA will usually respond with “I see,” or perhaps “Tell me more.”
For the first few lines of dialogue, especially if you treat her as your therapist, ELIZA can be convincingly human. This was something Weizenbaum noticed and was slightly alarmed by: people were willing to treat the algorithm as more human than it really was. Before long, even though some of the test subjects knew ELIZA was just a machine, they were opening up with some of their deepest feelings and secrets. They were pouring out their hearts to a machine. When Weizenbaum’s secretary spoke to ELIZA, even though she knew it was a fairly simple computer program, she still insisted Weizenbaum leave the room.
Part of the unexpected reaction ELIZA generated may be because people are more willing to open up to a machine, feeling they won’t be judged, even if the machine is ultimately powerless to do or say anything to really help. The ELIZA effect was named for this computer program: the tendency of humans to anthropomorphize machines, or think of them as human.
Weizenbaum himself, who later became deeply suspicious of the influence of computers and artificial intelligence in human life, was astonished that people were so willing to believe his script was human. He wrote, “I had not realized…that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.”
“Consciously, you know you’re talking to a big block of code stored somewhere out there in the ether. But subconsciously, you might feel like you’re interacting with a human.”
The ELIZA effect may have disturbed Weizenbaum, but it has intrigued and fascinated others for decades. Perhaps you’ve noticed it in yourself, when talking to an AI like Siri, Alexa, or Google Assistant—the occasional response can seem almost too real. Consciously, you know you’re talking to a big block of code stored somewhere out there in the ether. But subconsciously, you might feel like you’re interacting with a human.
Yet the ELIZA effect, as enticing as it is, has proved a source of frustration for people who are trying to create conversational machines. Natural language processing has proceeded in leaps and bounds since the 1960s. Now you can find friendly chatbots like Mitsuku—which has frequently won the Loebner Prize, awarded to the machines that come closest to passing the Turing test—that aim to have a response to everything you might say.
In the commercial sphere, Facebook has opened up its Messenger program and provided software for people and companies to design their own chatbots. The idea is simple: why have an app for, say, ordering pizza when you can just chatter to a robot through your favorite messenger app and make the order in natural language, as if you were telling your friend to get it for you?
Startups like Semantic Machines hope their AI assistant will be able to interact with you just like a secretary or PA would, but with an unparalleled ability to retrieve information from the internet. They may soon be there.
But people who engineer chatbots—both in the social and commercial realm—encounter a common problem: the users, perhaps subconsciously, assume the chatbots are human and become disappointed when they’re not able to have a normal conversation. Frustration with miscommunication can often stem from raised initial expectations.
So far, no machine has really been able to crack the problem of context retention—understanding what’s been said before, referring back to it, and crafting responses based on the point the conversation has reached. Even Mitsuku will often struggle to remember the topic of conversation beyond a few lines of dialogue.
“For everything you say, there could be hundreds of responses that would make sense. When you travel a layer deeper into the conversation, those factors multiply until you end up with vast numbers of potential conversations.”
This is, of course, understandable. Conversation can be almost unimaginably complex. For everything you say, there could be hundreds of responses that would make sense. When you travel a layer deeper into the conversation, those factors multiply until—like possible games of Go or chess—you end up with vast numbers of potential conversations.
But that hasn’t deterred people from trying, most recently, tech giant Amazon, in an effort to make their AI voice assistant, Alexa, friendlier. They have been running the Alexa Prize competition, which offers a cool $500,000 to the winning AI—and a bonus of a million dollars to any team that can create a ‘socialbot’ capable of sustaining a conversation with human users for 20 minutes on a variety of themes.
Topics Alexa likes to chat about include science and technology, politics, sports, and celebrity gossip. The finalists were recently announced: chatbots from universities in Prague, Edinburgh, and Seattle. Finalists were chosen according to the ratings from Alexa users, who could trigger the socialbots into conversation by saying “Hey Alexa, let’s chat,” although the reviews for the socialbots weren’t always complimentary.
By narrowing down the fields of conversation to a specific range of topics, the Alexa Prize has cleverly started to get around the problem of context—just as commercially available chatbots hope to do. It’s much easier to model an interaction that goes a few layers into the conversational topic if you’re limiting those topics to a specific field.
Developing a machine that can hold almost any conversation with a human interlocutor convincingly might be difficult. It might even be a problem that requires artificial general intelligence to truly solve, rather than the previously-employed approaches of scripted answers or neural networks that associate inputs with responses.
But a machine that can have meaningful interactions that people might value and enjoy could be just around the corner. The Alexa Prize winner is announced in November. The ELIZA effect might mean we will relate to machines sooner than we’d thought.
So, go well, little socialbots. If you ever want to discuss the weather or what the world will be like once you guys take over, I’ll be around. Just don’t start a therapy session.
Image Credit: Shutterstock Continue reading
Posted in Human Robots
#431155 What It Will Take for Quantum Computers ...
Quantum computers could give the machine learning algorithms at the heart of modern artificial intelligence a dramatic speed up, but how far off are we? An international group of researchers has outlined the barriers that still need to be overcome.
This year has seen a surge of interest in quantum computing, driven in part by Google’s announcement that it will demonstrate “quantum supremacy” by the end of 2017. That means solving a problem beyond the capabilities of normal computers, which the company predicts will take 49 qubits—the quantum computing equivalent of bits.
As impressive as such a feat would be, the demonstration is likely to be on an esoteric problem that stacks the odds heavily in the quantum processor’s favor, and getting quantum computers to carry out practically useful calculations will take a lot more work.
But these devices hold great promise for solving problems in fields as diverse as cryptography or weather forecasting. One application people are particularly excited about is whether they could be used to supercharge the machine learning algorithms already transforming the modern world.
The potential is summarized in a recent review paper in the journal Nature written by a group of experts from the emerging field of quantum machine learning.
“Classical machine learning methods such as deep neural networks frequently have the feature that they can both recognize statistical patterns in data and produce data that possess the same statistical patterns: they recognize the patterns that they produce,” they write.
“This observation suggests the following hope. If small quantum information processors can produce statistical patterns that are computationally difficult for a classical computer to produce, then perhaps they can also recognize patterns that are equally difficult to recognize classically.”
Because of the way quantum computers work—taking advantage of strange quantum mechanical effects like entanglement and superposition—algorithms running on them should in principle be able to solve problems much faster than the best known classical algorithms, a phenomenon known as quantum speedup.
Designing these algorithms is tricky work, but the authors of the review note that there has been significant progress in recent years. They highlight multiple quantum algorithms exhibiting quantum speedup that could act as subroutines, or building blocks, for quantum machine learning programs.
We still don’t have the hardware to implement these algorithms, but according to the researchers the challenge is a technical one and clear paths to overcoming them exist. More challenging, they say, are four fundamental conceptual problems that could limit the applicability of quantum machine learning.
The first two are the input and output problems. Quantum computers, unsurprisingly, deal with quantum data, but the majority of the problems humans want to solve relate to the classical world. Translating significant amounts of classical data into the quantum systems can take so much time it can cancel out the benefits of the faster processing speeds, and the same is true of reading out the solution at the end.
The input problem could be mitigated to some extent by the development of quantum random access memory (qRAM)—the equivalent to RAM in a conventional computer used to provide the machine with quick access to its working memory. A qRAM can be configured to store classical data but allow the quantum computers to access all that information simultaneously as a superposition, which is required for a variety of quantum algorithms. But the authors note this is still a considerable engineering challenge and may not be sustainable for big data problems.
Closely related to the input/output problem is the costing problem. At present, the authors say very little is known about how many gates—or operations—a quantum machine learning algorithm will require to solve a given problem when operated on real-world devices. It’s expected that on highly complex problems they will offer considerable improvements over classical computers, but it’s not clear how big problems have to be before this becomes apparent.
Finally, whether or when these advantages kick in may be hard to prove, something the authors call the benchmarking problem. Claiming that a quantum algorithm can outperform any classical machine learning approach requires extensive testing against these other techniques that may not be feasible.
They suggest that this could be sidestepped by lowering the standards quantum machine learning algorithms are currently held to. This makes sense, as it doesn’t really matter whether an algorithm is intrinsically faster than all possible classical ones, as long as it’s faster than all the existing ones.
Another way of avoiding some of these problems is to apply these techniques directly to quantum data, the actual states generated by quantum systems and processes. The authors say this is probably the most promising near-term application for quantum machine learning and has the added benefit that any insights can be fed back into the design of better hardware.
“This would enable a virtuous cycle of innovation similar to that which occurred in classical computing, wherein each generation of processors is then leveraged to design the next-generation processors,” they conclude.
Image Credit: archy13 / Shutterstock.com Continue reading
Posted in Human Robots
#431006 Adoption of robotics into a ...
VTT Technical Research Centre of Finland studied the implementation of a logistics robot system at the Seinäjoki Central Hospital in South Ostrobothnia. The aim is to reduce transportation costs, improve the availability of supplies and alleviate congestion on hospital hallways by running deliveries around the clock on every day of the week. Joint planning and dialogue between the various occupational groups and stakeholders involved was necessary for a successful change process. Continue reading
Posted in Human Robots