Backchannel has a fascinating profile of DeepMind founder Demis Hassabis which although an interesting read in itself, has a link to a brief, barely mentioned study which may herald a quiet revolution in artificial intelligence.
The paper (available online as a pdf) is entitled “Playing Atari with Deep Reinforcement Learning” and describes an AI system which, without any prior training, learned to play a series of Atari 2600 games to the point of out-performing humans.
The key here is ‘without any prior training’ as the system was not ‘told’ anything about the games. It worked out how to play them, and how to win them, entirely on its own.
The system was created with a combination of a reinforcement learning system and a deep learning network.
Reinforcement learning is based on the psychological theory of operant conditioning where we learning through reward and punishment what behaviours help us achieve certain goals.
One difficulty is that in video games, the reward (points) may only be distantly related to individual actions because strategy is not something that can be boiled down to ‘do this action again to win’. Mathematically there is lots of noise in the link between an action and eventual outcome.
Traditionally this has been solved by programming the structure of the game into the AI agent. Non-player characters in video games act as effective opponents because they include lots of hard-coded rules about what different aspects of the game symbolise, and what good strategy involves.
But this is a hack that doesn’t generalise. Genuine AI would work out what to do, in any given environment, by itself.
To help achieve this, the DeepMind Atari AI uses deep learning, a hierarchical neural network that is good at generating its own structure from unstructured data. In this case, the data was just what was on the screen.
To combine ‘learning effective action’ and ‘understanding the environment’ the research team plumbed together deep learning and reinforcement learning with an algorithm called Q-learning that is specialised for ‘model-free’ or unstructured learning.
So far, we have performed experiments on seven popular ATARI games – Beam Rider, Breakout, Enduro, Pong, Q*bert,Seaquest, Space Invaders. We use the same network architecture, learning algorithm and hyperparameters settings across all seven games, showing that our approach is robust enough to work on a variety of games without incorporating game-specific information…
Finally, we show that our method achieves better performance than an expert human player on Breakout, Enduro and Pong and it achieves close to human performance on Beam Rider.
The team note that the system wasn’t so good at Q*bert, Seaquest and Space Invaders, and it wasn’t asked to battle the real Ko-Dan Empire after playingStarfighter, but it’s still incredibly impressive.
It’s an AI that worked out its environment, its actions, and what it needs to do to ‘survive’, without any prior information.
Given, the environment is an Atari 2600, but the AI is a surprisingly simple system that ends up, in some instances, outperforming humans from a standing start.
Essentially, the future of humanity now rests on whether the next system is given a gun or a dildo to play with. Link to Backchannel profile of Demis Hassabis. pdf of paper “Playing Atari with Deep Reinforcement Learning”
There’s a key difference between knowledge and experience and it’s best described like this:
The image is from cartoonist Hugh MacLeod, who came up with such a brilliant way to express a concept that’s often not that easy to grasp.
The image makes a clear point—that knowledge alone is not useful unless we can make connections between what we know. Whether you use the terms “knowledge” and “experience” to explain the difference or not, the concept itself is sound.
Lots of great writers, artists and scientists have talked about the importance of collecting ideas and bits of knowledge from the world around us, and making connections between those dots to fuel creative thinking and new ideas.
This is a really fun, inspiring topic to read about, so I collected some quotes and advice from my favorite creative thinkers about the importance of making connections in your brain. I’ve added emphasis to the important parts, but if you have time I’d recommend reading the whole post and even digging into the sources I’ve linked to.
To start with though, I want to look at some research that shows intelligence is closely linked with the physical connections in our brains.
INTELLIGENCE AND CONNECTIONS: WHY YOUR BRAIN NEEDS TO COMMUNICATE WELL WITH ITSELF
The researchers found that, rather than residing in a single structure, general intelligence is determined by a network of regions across both sides of the brain.
One of the researchers explained that the study showed the brain working as a distributed system:
“Several brain regions, and the connections between them, were what was most important to general intelligence,” explains Gläscher.
The study also supported an existing theory about intelligence that says general intelligence is based on the brain’s ability to pull together and integrate various kinds of processing, such as working memory.
At Washington University, a research study found that connectivity with a particular area of the prefrontal cortex has a correlation with a person’s general intelligence.
This study showed that intelligence relied partly on high functioning brain areas, and partly on their ability to communicate with other areas in the brain.
Aside from physical connectivity in the brain, being able to make connections between ideas and knowledge we hold in our memories can help us to think more creatively and produce higher quality work.
CONNECTIONS FUEL CREATIVITY: NOTHING IS ORIGINAL
Steve Jobs is an obvious person to reference whenever you’re talking about creativity or innovation, so I wasn’t surprised to find that he has spoken about making connections before. This great quote is from a Wired interview in 1996:
Creativity is just connecting things. When you ask creative people how they did something, they feel a little guilty because they didn’t really do it, they just saw something.
That’s because they were able to connect experiences they’ve had and synthesize new things. And the reason they were able to do that was that they’ve had more experiences or they have thought more about their experiences than other people.
… in order for us to truly create and contribute to the world, we have to be able to connect countless dots, to cross-pollinate ideas from a wealth of disciplines, to combine and recombine these pieces and build new castles.
She’s given a talk on this at a Creative Mornings event before, and made some great points. Being able to read about a wide range of topics is often one of of the most important elements. I really liked how she pointed out the way our egos affect our willingness to build on what others have done before:
… something we all understand on a deep intuitive level, but our creative egos sort of don’t really want to accept: And that is the idea that creativity is combinatorial, that nothing is entirely original, that everything builds on what came before…
My favorite part of this talk is Popova’s Lego analogy, where she likens the dots of knowledge we have to Lego building blocks:
The more of these building blocks we have, and the more diverse their shapes and colors, the more interesting our castles will become.
Author Austin Kleon is someone who immediately comes to mind whenever the topic of connections and remixing art comes up. Kleon is the author of Steal Like An Artist, a book about using the work of others to inspire and inform your own.
It starts off like this:
Every artist gets asked the question, “Where do you get your ideas?”
The honest artist answers, “I steal them.”
Kleon is inspiring because he’s so upfront about how the work of other people has become part of his own work. He’s also keen on the phrase I quoted from Maria Popova above, that “nothing is original”:
Every new idea is just a mashup or a remix of one or more previous ideas.
If you’re looking for advice on creating more connections between the knowledge you have (and collecting even more knowledge), Kleon’s book is a great place to start. He offers suggests like:
HOW SCIENTIFIC THINKING IS ALL ABOUT MAKING CONNECTIONS
When it comes to the field of science, making connections between those dots of knowledge seems to be just as important. In The Art of Scientific Investigation,Cambridge University professor W. I. B. Beveridge wrote that successful scientists “have often been people with wide interests,” which led to their originality:
Originality often consists in linking up ideas whose connection was not previously suspected.
He also suggested that scientists should expand their reading outside of their own field, in order to add to their knowledge (so they would have more dots when it came time to connect them, later):
Most scientists consider that it is a more serious handicap to investigate a problem in ignorance of what is already known about it.
Nature no more obeys the territorial divisions of scientific academic disciplines than do continents appear from space to be colored to reflect the national divisions of their human inhabitants. For me, the great scientific satoris, epiphanies, eurekas, and aha! moments are characterized by their ability to connect.
START MAKING CONNECTIONS AND GETTING CREATIVE
I’ll leave you with some suggestions for improving your own ability to make connections.
1. ADD TO YOUR KNOWLEDGE—THE POWER OF BRAND NEW EXPERIENCES
After all, the more knowledge you have, the more connections you can make. Start by reading more, reading more widely, and exploring new opportunities for gathering knowledge (for instance, try some new experiences—travel, go to meetups or take up a new hobby).
As researcher Dr.Duezel explained when it comes to experiencing new things:
“Only completely new things cause strong activity in the midbrain area.”
So trying something new and forcing a gentle brain overload can make a dramatic improvement for your brain activity.
2. KEEP TRACK OF EVERYTHING – ESPECIALLY IN THE SHOWER
As Austin Kleon suggests, take a notebook (or your phone) with you everywhere and take notes. Don’t expect your brain to remember everything—give it a hand by noting down important concepts or ideas you come across. As you do this, you may remember previous notes that relate (hey, you’re making connections already!)—make a note of those as well.
You can do this even when you’re in the shower with something like Acqua Notes. The shower is especially a place that has proven to make us more creative.
Going over your notes often can help you to more easily recall them when you need to. Read through what you’ve made notes of before, and you might find that in the time that’s passed, you’ve added more knowledge to your repertoire that you can now connect to your old notes!
In the early 1900s, researchers looked into the psychology of magic as a way of understanding how the human mind works. That interest petered out over the next few decades, possibly due to the rise of behaviourism in the early 20th century, according to Ronald Rensink, a psychology professor at UBC.
“They were trying to break everything down into simple bits and pieces,” Rensink said. “There’s a bell and you expect meat. Simple associations, simple stimuli, and magic did not fit in to that.”
Rensink and co-researchers recently published a study showing how subtle cues can influence people to choose a desired card in a magic trick.
In this photo taken Feb. 11, 2015, Angelica Lopez, 3, smiles during a therapy session at the University of Southern California in Los Angeles. Angelica was born deaf and received an auditory brainstem implant to allow her to hear some sounds in a research study at USC. U.S. researchers are implanting a device on the brain stems of a small number of deaf children to see if it will help them learn to hear. The studies are aimed at children who don’t have working hearing nerves and thus don’t qualify for a more common technology, cochlear implants. The implants stimulate brain cells that those nerves normally target. (AP Photo/Damian Dovarganes)
Associated PressFeb. 13, 2015 | 3:08 p.m. EST+ More
By LAURAN NEERGAARD, AP Medical Writer
WASHINGTON (AP) — At age 3, Angelica Lopez is helping to break a sound barrier for deaf children.
Born without working auditory nerves, she can detect sounds for the first time — and start to mimic them — after undergoing brain surgery to implant a device that bypasses missing wiring in her inner ears.
Angelica is one of a small number of U.S. children who are testing what’s called an auditory brainstem implant, or ABI. The device goes beyond cochlear implants that have brought hearing to many deaf children but that don’t work for tots who lack their hearing nerve.
When the ABI is first turned on, “she isn’t going to be hearing like a 3-year-old. She’ll be hearing like a newborn,” audiologist Laurie Eisenberg of the University of Southern California tells parents. She outlined the research Friday at a meeting of the American Association for the Advancement of Science.
The children don’t magically understand and use those sounds. “It’s going to take a lot of work,” Eisenberg cautioned.
Angelica cried when her ABI first was switched on, scared by the sounds. But five months later, her mother says the youngster uses sign language to identify some sounds — that was a cough, that’s a dog barking. And she’s beginning to babble like hearing babies do, as therapists work to teach her oral speech.
“It’s just so awesome to hear her little voice,” said Julie Lopez of Big Spring, Texas, who enrolled her daughter in the study at USC, where researchers say she’s progressing well.
Many children born deaf benefit from cochlear implants, electrodes that send impulses to the auditory nerve, where they’re relayed to the brain and recognized as sound. But the small fraction born without a working hearing nerve can’t make that brain connection.
The ABI attempts to fill that gap by delivering electrical stimulation directly to the neurons on the brainstem the nerve normally would have targeted. Here’s how it works: The person wears a microphone on the ear to detect sound, and a processer changes it to electrical signals. Those are beamed to a stimulator under the skin, which sends the signals snaking through a wire to electrodes surgically placed on the brainstem.
The Food and Drug Administration approved the device in 2000 specifically for adults and teenagers whose hearing nerves had been destroyed by surgery for a rare type of tumor. It doesn’t restore normal hearing, but can help to varying degrees.
Then about a decade ago, an Italian surgeon started trying the ABI in deaf children, whose younger brains are more flexible and might better adapt to this artificial way of delivering sound.
Now, spurred by some successes abroad, the first U.S. studies in young children are underway at a handful of hospitals. Hearing specialists are watching the work closely.
There are children “who are not being helped in any other way,” said Dr. Gordon Hughes of the National Institutes of Health, which is funding Eisenberg’s study. And cochlear implants proved “there’s a critical time window when the brain is very receptive to auditory stimulation and can develop speech communication in ways that are surprisingly good, if the stimulation is started early enough.”
The studies are small, each enrolling 10 to 20 children. Ages vary; the Los Angeles study will implant starting at age 2, while some others try earlier. Children then receive intensive therapy, to learn to hear.
The studies must prove safety, since the ABI requires delicate brain surgery in healthy children.
“We’re talking about real surgery to go into a deep area of the brain,” said Dr. Marc Schwartz, a neurosurgeon with the House Clinic and Huntington Medical Research Institutes in Los Angeles, who is part of the USC study. “This is a precise operation that requires exacting technique.”
FIRST POSTED: FRIDAY, FEBRUARY 13, 2015 12:14 PM EST | UPDATED: FRIDAY, FEBRUARY 13, 2015 12:22 PM EST
Dogs can tell whether humans are happy or angry, a new study suggests.
Scientists from the aptly named clever dog lab at the Messerli Research Institute in Vienna, Austria, trained dogs to associate the pictures of happy or angry faces with a reward.
In another test, researchers showed the dogs pictures from a different batch of men’s and women’s faces, suggesting the animals could tell the difference between emotions.
Researchers showed 20 dogs images of either the upper half or the lower half of the faces of happy or angry people.
Half the dogs were given a treat when they touched an angry face, while the other group were rewarded for touching happy faces.
The dogs responded to the happy faces better than the sad ones, the study says.
“It seems that dogs dislike approaching angry faces,” study director Ludwig Huber told Science Daily.
More than half of the dogs learned the task very well.
“We believe that dogs draw on their memory during this exercise. They recognize a facial expression which they have already stored,” co-author Corsin Muller said, Science Daily reported.
“We suspect that dogs that have no experience with people would perform worse or could not solve the task at all.”
The study was published in the journal Current Biology.
BY SHERYL UBELACKER, THE CANADIAN PRESSFEBRUARY 12, 2015
TORONTO – Canadian researchers have developed a fibre-optic probe they believe could help neurosurgeons track down stray cancer cells in patients with invasive brain tumours, thereby potentially reducing the risk of recurrence and improving life expectancy.
The hand-held device, called a Raman spectroscopy probe, was developed by researchers at McGill University for use by neurosurgeons in the removal of brain tumours called gliomas, which typically infiltrate healthy brain tissue.
The difficulty for surgeons is that current imaging technologies like MRI are unable to detect all the malignant cells that have seeded themselves within normal tissue outside the solid tumour, said Dr. Kevin Petrecca, the chief of neurosurgery at the Montreal Neurological Institute who helped design and test the probe.
“So being able to visualize the tumour in its entirety is one area we’re unsuccessful at and the second one is that even if you could visualize the entirety of the tumour, it doesn’t mean that you could remove all of those areas,” Petrecca, who performs hundreds of brain tumour surgeries each year, said from Montreal.
“As a glioma surgeon, that’s all I think about every day: I wish I could see those cells.”
By using Raman spectroscopy — a technique that measures the way an object’s molecules scatter light — Petrecca said neurosurgeons can pinpoint cancer cells in small spots about a millimetre deep where the probe comes in contact with brain tissue.
In a study published Wednesday in Science Translational Medicine, the researchers showed the device detected not only the dense tumour mass, but also individual cancer cells that had penetrated surrounding tissue in 17 patients with advanced gliomas.
Since then, researchers have used the probe on a total of 40 patients without adverse effects. The team is planning a follow-up clinical trial that will assess whether use of the device increases patient survival.
Petrecca said the beauty of the probe is its simplicity: results from the probe are shown on a laptop within a second of it contacting the targeted area, allowing the surgeon to remove a tiny section of brain tissue if it is unlikely to cause physical or cognitive impairments in the patient.
Minimizing the amount of residual cancer left after surgery can affect survival, he said.
“The better resection you can do, the longer the patient lives. We’re really talking about the difference of living six months versus three, four, five, six years. So it’s a big, big deal.”
There are different types of gliomas: some are more invasive and aggressive than others, usually strike at different points in life, and have varied life expectancies that range from months to decades, depending on the tumour type and at what stage it is diagnosed.
Petrecca believes that with further development, the Raman-based probe could become a valuable tool in the operating room, with a cost that’s a tiny fraction of equipment like intraoperative MRIs and surgical robots.
“So do I think it’s going to be transformational? Absolutely.”
Dr. Mark Bernstein, a neurosurgeon at Toronto Western Hospital who specializes in brain tumours, called the Raman-based probe “an innovative application of sophisticated scientific theory and technology.”
However, he said it would be impossible to identify every cancer cell. And even if it were possible, surgically chasing down all the cells could end up destroying vital brain tissue, potentially leaving the patient unable to speak, walk or see.
The tumour also might still recur because the genetic underpinnings that initially gave rise to the cancer still remain, said Bernstein, who was not involved in the research.
“I truly applaud the authors’ idea and excellent and careful study of it, and await further data and advancement in techniques. But the chance of this technology translating into improved outcomes for patients with primary brain tumours is very low in my opinion,” he said.
In fact, Bernstein believes that if a cure for these brain tumours is ever found, the solution won’t be surgical but pharmaceutical, in the form of drugs that target the cancer cells at the molecular level while leaving healthy neurons unharmed.
“We think these lenses hold a lot of promise for low vision and age-related macular degeneration (AMD),” Tremblay said. “It’s very important and hard to strike a balance between function and the social costs of wearing any kind of bulky visual device. There is a strong need for something more integrated, and a contact lens is an attractive direction. At this point this is still research, but we are hopeful it will eventually become a real option for people with AMD.”
The telescopic contact, which magnifies 2.8 times, works by incorporating a thin reflective telescope into a 1.55 millimeter thick lens. Tiny mirrors inside the lens bounce the light around, expanding objects and magnifying the view.
The researchers also created a unique method for electronically switching views between unmagnified and telescopic vision “on demand.” In this system, the accompanying electronic glasses use a small light source and light detector to “recognize winks and ignore blinks.” If the wearer winks their right eye, the lenses will switch to the magnified view, and winking the left eye switches them back to normal vision.
The glasses electronically select a polarization of light to reach the contact lenses, allowing the polarization in the 1x aperture and another in the 2.8x aperture. The wearer can then see the view where the polarization of the glasses meets the aperature.
“The combination of the telescopic contact lenses and optional blink-controlled eyewear represent a huge leap in functionality and usability in vision aid devices and a major feat for optics research,” the researchers said.
“We think these lenses hold a lot of promise for low vision and age-related macular degeneration (AMD),” Tremblay said. “It’s very important and hard to strike a balance between function and the social costs of wearing any kind of bulky visual device. There is a strong need for something more integrated, and a contact lens is an attractive direction. At this point this is still research, but we are hopeful it will eventually become a real option for people with AMD.”
The telescopic contact, which magnifies 2.8 times, works by incorporating a thin reflective telescope into a 1.55 millimeter thick lens. Tiny mirrors inside the lens bounce the light around, expanding objects and magnifying the view.
The researchers also created a unique method for electronically switching views between unmagnified and telescopic vision “on demand.” In this system, the accompanying electronic glasses use a small light source and light detector to “recognize winks and ignore blinks.” If the wearer winks their right eye, the lenses will switch to the magnified view, and winking the left eye switches them back to normal vision.
The glasses electronically select a polarization of light to reach the contact lenses, allowing the polarization in the 1x aperture and another in the 2.8x aperture. The wearer can then see the view where the polarization of the glasses meets the aperature.
“The combination of the telescopic contact lenses and optional blink-controlled eyewear represent a huge leap in functionality and usability in vision aid devices and a major feat for optics research,” the researchers said.
Google has given fellow tech companies an ultimatum: patch your software vulnerabilities within 90 days or we’ll make them public.
An elite team of Google hackers and programmers scrub their own and competitors’ software for security flaws, giving companies a deadline to issue a fix. Google says it wants software makers to move fast because cybercriminals act with lightning speed when they spot bugs.
It’s a sensitive topic – rivals Microsoft and Apple declined to talk about the tactic – though others in the industry say the help isn’t always welcome, usurps a role best left to government and can jeopardise security.
“I’m not sure who made Google the official referee of the marketplace for vulnerability notification,” said John Dickson, a principal with software security company Denim Group in San Antonio. He said pressuring companies to fix flaws is a good idea, but “what noble motives they had in mind could be called into question given the fact that they essentially outed vulnerabilities for two of their biggest rivals.”
Google established the team in July, calling it Project Zero after the much-feared “zero day” security flaws that are exploited before developers learn of them. It says it is trying to help everyone as well as protect its own products that run on others’ devices and software.
That’s an activity some security experts say is more appropriate for a government agency. The respective roles of the private and public sectors is on the agenda at a cybersecurity summit Friday in Palo Alto, California, where President Barack Obama will call on technology leaders to improve cooperation and share more information.
Some researchers are wondering aloud, however, how much cooperation can be expected if the biggest internet companies can’t play nice together.
If these companies can’t even get along, that’s just bad for security for the whole ecosystem Jake Kouns, chief information security officer for Risk Based Security
Jake Kouns, chief information security officer, Risk Based Security
“We support a variety of efforts, including Project Zero and our Security Reward Programs, to find and fix online threats,” Aaron Stein, spokesman for the Mountain View, California-based Google said in an email.
Apple declined to comment while Microsoft would only refer to a previous statement in which it said Google’s tactics felt like a game of “gotcha,” illustrating how divisive the issue is.
“If these companies can’t even get along, that’s just bad for security for the whole ecosystem,” said Jake Kouns, chief information security officer for Risk Based Security Inc. in Richmond, Virginia.
Opponents of Google’s practice say it puts online security at risk by revealing gaps before they can be plugged.
Hackers work fast to exploit problems when they become known. Chinese-backed intruders exploited a Web-security flaw known as Heartbleed last year to attack Community Health Systems Inc. more than a week after the hole was publicised.
In January, Apple pleaded with Google to wait about a week before going public so it could fix three flaws in the Mac OS X operating system, according to a person familiar with the request who wasn’t authorised to speak publicly.
Google knew the fix was coming and had possession of the updated software because it serves as a developer for Apple, the person said. Regardless, Google refused and released details of the flaws.
The decision feels less like principles and more like a ‘gotcha,’ with customers the ones who may suffer as a result
Chris Betz, senior director of Microsoft’s Security Response Center
Microsoft, meanwhile, requested two additional days to fix a flaw in Windows. Google refused and publicised the bug.
“The decision feels less like principles and more like a ‘gotcha,’ with customers the ones who may suffer as a result,” wrote Chris Betz, senior director of Microsoft’s Security Response Center, in a January 11 blog post, which has remained the company’s only public comment on the issue to date. “What’s right for Google is not always right for customers.”
App users: Tap here to view the ‘Fighting Cybercrime: A New Era of Collaboration’ videoMicrosoft asks that researchers privately disclose flaws to software providers, working with them until a fix is made available, Betz said. “Policies and approaches that limit or ignore that partnership do not benefit the researchers, the software vendors, or our customers. It is a zero sum game where all parties end up injured,” he wrote.
Google supporters say the hard-line approach may fundamentally alter software industry practices in which companies can take months or years to patch bugs.
According to an analysis by Risk Based Security, Project Zero has identified 39 vulnerabilities in Apple products and 20 in Microsoft products. The team also has found 37 flaws in Adobe Systems software and 22 in the FreeType software development library for rendering fonts.
Project Zero publicly released details before a fix became available about Apple flaws 16 times, Microsoft three times and Adobe once, Kouns said in a phone interview.
Google’s “strict policy is good for the industry,” and the company should be praised because they “stuck to their guns,” said Tom Gorup, a manager with Rook Security based in Indianapolis.
“A regular Joe on the street doesn’t have the clout that Google does,” Gorup said in a phone interview. “If we have huge companies like Microsoft, Apple and Google going at each other and pushing for better security, it’s a win across the board.”
Google created Project Zero after revelations about the Heartbleed bug and spying by the National Security Agency and other governments.
“You should be able to use the Web without fear that a criminal or state-sponsored actor is exploiting software bugs to infect your computer, steal secrets or monitor your communications,” according to a July 15 blog post announcing Project Zero. “Our objective is to significantly reduce the number of people harmed by targeted attacks.”
Google also is helping to spur the market for managing and patching software vulnerabilities, which is expected to grow to $1 billion in value by 2018 from about $600 million in 2014, said Christopher Kissel, a network security industry analyst with research company Frost & Sullivan Inc.
We’ve had a lot of experiences where vendors will seemingly not care about something unless it’s in the headlines or unless there’s something out there that people see as an immediate threat
Craig Young, senior security researcher with Tripwire
Companies that provide vulnerability management services like Hewlett-Packard, Tenable Network Security Inc. and Qualys Inc. stand to gain from the increased spending, Kissel said in a phone interview.
The number of internet flaws being found surged to 7,903 in 2014 from 5,174 in 2013, he said. It took companies 205 days on average in 2014 to learn that hackers had infiltrated their networks, according to cybersecurity company FireEye Inc.
“While a few adversaries use zero-day exploits to target victims, many adversaries still target known vulnerabilities for which patches have been released, capitalising on slow patch processes and risk decisions by network owners not to patch certain vulnerabilities or systems,” the FBI said in an alert at the end of January obtained by Bloomberg News.
A 90-day deadline might not be practical for large companies that have to search through thousands lines of code and make sure patches don’t negatively affect other software, said Craig Young, a senior security researcher with Tripwire Inc. based in Portland, Oregon, in a phone interview.
Other times, however, a company may be negligent. “We’ve had a lot of experiences where vendors will seemingly not care about something unless it’s in the headlines or unless there’s something out there that people see as an immediate threat,” Young said.
Young reported a bug to Apple in October 2012 that could let hackers attack a file server in OS X. Although the flaw wasn’t critical, Apple didn’t issue a final patch until Jan. 27 of this year, Young said.
The flaws exposed by Project Zero without fixes so far haven’t been very serious, Young said. He said he would have more concerns if Google published details about a critical vulnerability that put users at a high risk.
“Microsoft is using this opportunity to kick some sand up in Google’s face and attack their mantra of ‘Do no evil’,” said Gorup with Rook Security. “If it was a government entity, Microsoft wouldn’t be able to make the case.”
A UBC researcher explains why some lacy new lingerie might just do the trick
Is your relationship going stale? UBC psychology researcher Heather Morton explains the research behind flagging libidos, and offers tips on reigniting the flames of passion.
Why did you want to study partner novelty?
If you Google “How to enhance sexual desire” or “How to spice up your relationship,” you will find a large number of pop psychology and self-help sites will suggest the idea of novelty. Even couples counsellors will recommend these kinds of strategies: “Take salsa dancing” or “Buy new lingerie.” However, there is a lack of research examining the efficacy of this strategy.
Many non-human animal studies suggest that novelty may plan an important role in sexual functioning. This has been frequently demonstrated by the Coolidge effect, where a male will copulate with a female repetitively until he’s satiated and is no longer interested in continuing. But when a new female is introduced, he’s interested again. Although you clearly can’t do this with humans, there are some areas of research that have examined this phenomenon indirectly.
What have you found?
In our recent review of the literature, we found evidence that suggests men and women’s sexual desire declines over time with familiar partners and returns with novel partners. For instance, research on long-term relationships has shown that sexual frequency, sexual satisfaction, and sexual desire decline over time with the same partner. Remarriage has been shown to result in an increase in sexual frequency, therefore this effect cannot simply be attributed to age.
Laboratory studies examining habituation to erotica also support this idea. If you show someone the same erotic clip over and over again, his or her level of arousal decreases. When you show someone a new clip, his or her arousal is renewed. It appears that the Coolidge effect does occur, to some extent, in both men and women.
Is this depressing news for long-term couples?
It may sound depressing, but it can also be empowering. Men and women are choosing, for the most part, to enter monogamous relationships. There are a lot of benefits to these unions with regards to physical and mental health as well as monetary benefits.
I think it’s also helpful to know the potential downsides associated with these relationships, so you can create strategies to overcome them. What we’re now in the process of studying is whether couples can reverse the effect of declining desire by increasing novelty within their relationship, such as engaging in a greater variety of leisure activities together or in a wider sexual repertoire.
It can also be helpful to understand that this is natural—that there’s not something wrong with you or your partner. My hope is that we can use this information to benefit long-term relationships. Maybe you can have an affair within your own relationship.
Do you have any tips for couples on Valentine’s Day?
Talk with your partner about exciting activities you may want to try together this Valentines, and new sexual experiences you’d like to have together.
Couples may be able to gain the benefits in sexual functioning that accompany a new partner by introducing novelty in other ways. Previous studies have shown that engaging in exciting activities together increases relationship satisfaction, however the impact on sexual functioning was not examined.
Early results from one of our studies suggest that men and women who engage in a greater variety of sexual activities experience greater sexual desire and satisfaction. It appears as though the frequently given advice on “spicing up your relationship” may in fact be spot on.
Jiebo Luo, professor of computer science at the University of Rochester, in collaboration with researchers at Adobe Research has come up with a more accurate way than currently possible to train computers to be able to digest big data that comes in the form of images.
In a paper presented at the recent American Association for Artificial Intelligence (AAAI) conference in Austin, Texas, the researchers describe a “progressive training deep convolutional neural network (CNN).”
Once trained, a computer can be used to determine what sentiments (feelings) that a given image is likely to elicit. Luo says that this information could be useful for things as diverse as measuring economic indicators and predicting elections.
Sentiment analysis of text by computers is itself a challenging task; in social media, where many people express themselves using images and videos, sentiment analysis is even more complicated.
For example, during a political campaign voters will often share their views through pictures. Two different pictures might show the same candidate, but they might be making very different political statements. A human could recognize one as being a positive portrait of the candidate (e.g. the candidate smiling and raising his arms) and the other one being negative (e.g. a picture of the candidate looking defeated).
But no human could look at every picture shared on social media — it is truly “big data.” To be able to make informed guesses about a candidate’s popularity, computers need to be trained to digest this data, which is what Luo and his collaborators’ approach can do more accurately than was possible until now.
Image classification
The researchers treat the task of extracting sentiments from images as an image classification problem. This means that somehow each picture needs to be analyzed and labels applied to it.
To begin the training process, Luo and his collaborators used a huge number of Flickr images that have been loosely labeled by a machine algorithm with specific sentiments, in an existing database known as SentiBank (developed byProfessor Shih-Fu Chang’s group at Columbia University).
This gives the computer a starting point to begin understanding what some images can convey. But the machine-generated labels also include a likelihood of that label being true, that is, how sure is the computer that the label is correct?
The key step of the training process comes next, when the researchers discard any images for which the sentiment or sentiments with which they have been labeled might not be true. So they use only the “better” labeled images for further training in a progressively improving manner within the framework of the powerful convolutional neural network. They found that this extra step significantly improved the accuracy of the sentiments with which each picture is labeled.
They also adapted this sentiment analysis engine with some images extracted from Twitter. In this case they employed “crowd intelligence,” with multiple people helping to categorize the images via the Amazon Mechanical Turk platform. They used only a small number of images for fine-tuning the computer and yet, by applying this domain-adaptation process, they showed they could improve on current state of the art methods for sentiment analysis of Twitter images.
One surprising finding is that the accuracy of image sentiment classification has exceeded that of the text sentiment classification on the same Twitter messages.
Bruce's Blog 