Category Archives: Ai

Size isnt everything, relationships are

imageBROKER/REX/Shutterstock

By Matt Reynolds

The world is a confusing place, especially for an AI. But a neural network developed by UK artificial intelligence firm DeepMind that gives computers the ability to understand how different objects are related to each other could help bring it into focus.

Humans use this type of inference called relational reasoning all the time, whether we are choosing the best bunch of bananas at the supermarket or piecing together evidence from a crime scene. The ability to transfer abstract relations such as whether something is to the left of another or bigger than it from one domain to another gives us a powerful mental toolset with which to understand the world. It is a fundamental part of our intelligence says Sam Gershman, a computational neuroscientist at Harvard University.

Whats intuitive for humans is very difficult for machines to grasp, however. It is one thing for an AI to learn how to perform a specific task, such as recognising what is in an image. But transferring know-how learned via image recognition to textual analysis or any other reasoning task is a big challenge. Machines capable of such versatility will be one step closer to general intelligence, the kind of smarts that lets humans excel at many different activities.

DeepMind has built a neural network that specialises in this kind of abstract reasoning and can be plugged into other neural nets to give them a relational-reasoning power-up. The researchers trained the AI using images depicting three-dimensional shapes of different sizes and colours. It analysed pairs of objects in the images and tried to work out the relationship between them.

The team then asked it questions such as What size is the cylinder that is left of the brown metal thing that is left of the big sphere? The system answered these questions correctly 95.5 per cent of the time slightly better than humans. To demonstrate its versatility, the relational reasoning part of the AI then had to answer questions about a set of very short stories, answering correctly 95 per cent of the time.

Still, any practical applications of the system are still a long way off, says Adam Santoro at DeepMind, who led the study. It could initially be useful for computer vision, however. You can imagine an application that automatically describes what is happening in a particular image, or even video for a visually impaired person, he says.

Outperforming humans at a niche task is also not that surprising, says Gershman. We are still a very long way from machines that can make sense of the messiness of the real world. Santoro agrees. DeepMinds AI has made a start by understanding differences in size, colour and shape but theres more to relational reasoning than that. There is a lot of work needed to solve richer real-world data sets, says Santoro.

Read more: The road to artificial intelligence: A case of data over theory

Read more: Im in shock! How an AI beat the worlds best human at Go

More on these topics:

Read the original here:

DeepMind's neural network teaches AI to reason about the world - New Scientist

Mashable

An AI server took China's national math exam and did pretty badly Mashable An AI machine that sat the math paper for China's college entrance exam has failed to prove it's better than its human competition. AI-Maths, a machine made of 11 servers, three years in the making, joined almost 10 million high schoolers last week, in ... Even An AI Supercomputer Found This College Entrance Exam ToughIFLScience all 6 news articles »

View original post here:

An AI server took China's national math exam and did pretty badly - Mashable

Artificial intelligence may be all the craze in Silicon Valley, but on Wall Street, well, theres a lot of skepticism.

High-powered algorithms are not a new phenomenon in finance, and for this industry, the name of the game is efficiency and precision.

Quite frankly, finance executives want systems that, in one way or another, make money. Because of this, new wild and flashy AI systems that just make something smart wont fly.

The fintech companies that are successfully leveraging AI today are the ones that have found a very concrete way to apply the technology to an existing business problem. For example, technology such as specialized hardware, big data analytics, and machine learning algorithms are being used in fintech to augment tasks that people already perform.

At the Singularity University Exponential Finance Summit this week, Neil Jacobstein, faculty chair of Artificial Intelligence and Robotics at SU, shared some of the most interesting AI companies in fintech right now.

Not surprisingly, these companies each have a clear market application and reduce friction in the business problems they address.

Numerai is a new kind of hedge fund that is built by crowdsourcing knowledge through a massive network of hedge fundsthe system collects hundreds of thousands of financial models and individual predictions. With this information, Numerai is building their own financial models that incorporate the algorithms submitted through the crowdsourced community. Numerai has already secured funding from First Round Capital and Union Square Ventures, which is no small feat.

In 2010, AlphaSense launched its intelligent search engine, which uses AI, natural language processing algorithms, and advanced linguistic search tools to provide researchers with critical insights with serious accuracy and speed. Financial analysts can pose questions to AlphaSenses systems and get insights that are significantly more customized and accurate than a simple Google search would provide. Its a great example of an AI augmenting a critical task in finance: research.

Opera is helping companies turn their big data into predictive insights and business intelligence. The company uses pattern recognition to identify what they call signals, meaning actionable insights from data. Their signals help researchers understand conditions that may be happening in the market, or the world at large, so they can act quickly on these changes.

AppZen is a very practical solution to one of every executives most arduous taskssubmitting expense reports. The system uses AI to audit 100 percent of employee expenses and then generates an expense report in real time. Automating this process saves companies hours of lost productivity. AppZen also gives companies more confidence in their ability to flag suspicious charges. So, if youve been considering expensing that pricey night out with clients, dont, because AppZen will likely flag it.

CollectAI is a cloud-based software system thats shaking up the collection business. The system is able to mimic the voice and tone of a collection agent to gather important information over the phone about a collections case. With this information, CollectAI uses a self-learning algorithm to learn about the case, and then pulls knowledge from previous successful cases and applies those insights to decide how to best approach the situation at hand. The system gets better and better over time, which is pretty incredible.

Image Credit: Pond5

Go here to read the rest:

5 Fintech Companies Using AI to Improve Business - Singularity Hub

A US intelligence director says a lot of espionage is more boring than you might think, and much of it could be handed over to artificial intelligence.

A significant chunk of the time, I will send [my employees] to a dark room to look at TV monitors to do national security essential work, Robert Cardillo, head of the National Geospatial-Intelligence Agency told reporters including Foreign Policy. But boy is it inefficient.

Cardillo calls out recent advances in artificial intelligence, giving algorithms the ability to analyze vast amounts of images and video to find patterns, give data about the landscape, and identify unusual objects. This kind of work is critical for assessing national security concerns like foreign missile-silo activity, or even just to check in on North Korean volleyball games.

Cardillo has hired a former tech CEO, Anthony Vinci, to lead development of this machine-learning technology. Vinci previously founded Findyr, a company that crowdsources data for companies, like pictures of how products are displayed on shelves or infrastructure development progress.

But the US government is already trailing behind Silicon Valley on this pursuit. Facebook, deep into its crusade to connect the world, was able to apply machine learning to satellite data last year, analyzing buildings likely to contain wireless internet down to five-meter accuracy. Those data were intended to be used to guide Facebooks internet drone, Aquila. (Aviation might prove a more difficult project than machine learning: Facebooks head of global aviation policy indicated that the company doesnt have a timeline to even get one drone in service after last years test.)

Google has also touted its ability to discern details in satellite imagery for accuracy-critical uses, like defense and aviation. Stanford University has used satellite imagery to map poverty.

Cardillos initiatives arent the first use of AI by intelligence or defense agencies. DARPA and IARPA, US defense and intelligence research agencies, have been funding AI research for decades, and the Central Intelligence Agencys venture arm is supporting efforts to apply AI analysis to satellite imagery.

Read more here:

US intelligence agencies are beginning to build AI spies - Quartz

CAMBRIDGE, MA: For all the speculation about where the outer reaches of artificial intelligence may lie, brands can usefully deploy AI today in order to both make and save money.

In an article for the Harvard Business Review site, consultant Brad Power outlined how AI-powered tools are already making a significant difference to businesses as diverse as telecoms providers, software companies and imaging businesses.

"AI tools are the only way I can scale 'helpfulness' to a global community of 200,000-plus users with a team of two," one CMO explained.

Virtual assistants are able to send emails and manage responses, to conduct chats with customers and handle routine inquiries, to identify potential leads and pass the best ones on to humans.

For a business like RapidMiner, whose free trials of an analytical tool for data scientists were attracting 60,000 users a month, chat tools offered a way of managing far more interactions than its existing staff could ever do.

According to chief marketing officer Tom Wentworth, its chosen tool resolves around two-thirds of customer inquiries while also generating sales leads. "It's the most productive thing I'm doing in marketing," he told Power.

"I've learned things about my visitors that no other analytics system would show," he added. "We've learned about new use cases, and we've learned about product problems."

Power suggested that the ability of an AI agent to elicit information like a person rather than simply looking for patterns in data is a particular strength.

And because many of these tools are now offered as-a-service, it is simple for businesses to carry out trials and scale up.

"Clearly, it's worthwhile for companies to test AI-powered chat or email tools to see if they can convert more leads, and improve their understanding of what customers want and need," Power said.

Evidence of just how worthwhile it can be came from Epson America, the printer and imaging business: a 75% increase in qualified leads produced a claimed $2m in incremental revenue in just 90 days.

Data sourced from Harvard Business Review; additional content by WARC staff

Here is the original post:

AI scales 'helpfulness' - Warc

In BriefDeep Blue, IBM's chess computer, caused a worldwide epiphanyregarding the capabilities of AI when it defeated Gary Kasparov in1997. What is the legacy of this match, what other games has AIexcelled in, and what will it succeed in next? Deep Blues Victory

Murray Campbell, a Distinguished Research Staff Member at IBM, recently discussed the legacy and impact of the fateful 1997 chess series in which IBMs Deep Blue beat Gary Kasparov the world number one chess player for 225 out of 228 months between 1986 and 2005.

Campbell was part of the portentous encounter himself. He was a member of the team that helped build Deep Blues university progenitor, Deep Thought, which was the first program to beat a grandmaster in a professional tournament. When IBM took notice, Campbell andhis colleagues were hired by them to build Deep Blue. The system they eventually built was a combination of general-purpose supercomputer processors combined with [] chess accelerator chips.

Although computers had beaten humans in games before such as BKG 9.8s victory over Luigi Villa at backgammon in 1979 and Chinooks domination of Don Lafferty in checkers in 1994 Deep Blues victory was considered so auspicious because it won at chess.

David J. Staley wrote, concerning the match, that Chess represents a domain of human skill that is simple enough to model yet complex enough to reflect deep levels of cognition. Therefore, Deep Blues triumph marked the first true trophy for artificial intelligencebecause it beat the man who many consider to be the best player of all time at a game that has been regarded as a pinnacle of human intelligence since antiquity.

Monty Newborn, Emeritus Professor of McGill Universitys School of Computer Science, makes an apt analogy in his book Kasparov versus Deep Blue: Computer Chess Comes of Age.He states that many advances in the auto world were first tried on racing models and then after refinement incorporated into commercial vehicles. This may be the pattern in the computer field, too, where techniques used by computers to play chess are on the cutting edge of developments in complex problem-solving.

Since Deep Blue established a benchmark, says Campbell, machines have improved in processing speed and memory and so on resulting in them adding more and more gaming jewels to their virtual crown. Additionally, machine learning algorithms have access to a lot more data than they did in the past.

In recent years, the most notable victories have been AlphaGos win over five of the best Go (a game arguably more complicated than chess) players simultaneously, and Libratuss domination over four of the worlds top poker players. In this latter encounter, Dong Kim, one of the contestants, told Wired,I felt like I was playing against someone who was cheating, like it could see my cards. Im not accusing it of cheating. It was just that good.

These developments may be reflections of AIs incremental development towards becoming more human, as these games are similar to the complexities and solutions of life itself. However, Kasparovs point from 2010 that modern technology is a culture of optimization that it is derivative, incremental, profit margin-forced, consumer-friendly technology not the kind that pushes the whole world forward economically applies to these victories.

Perhaps AIs real challenge, and the next paradigm shift, is for it to defeat a game we have developed in modern times like StarCraft II.Oriol Vinyals, a DeepMind researcher and former top-ranked StarCraft player, told The Verge that the game is so complex and multifaceted that the skills required for an agent to progress through the environment and play StarCraft well could ultimately transfer to real-world tasks.

Even though you can play against AI when you play StarCraft, the AI that Vinyals is working on would be modeled afterthe way humans play the game along with usingthe same rules we do. AIs are able to play simple video games (think Atari-level), but nothing as complex as StarCraft yet. The researchers dont know when an AI will be created that is able to best a top-ranked player, but the day will come. This AI will have been taught to make decisions as a human would when playing a game with far more layers and complexities than any gameattempted by AI before. Maybe it will be able to teach players the perfect strategy to defeating a Zerg rush.

Link:

AI Has Beaten Humanity at Our Own Game. Literally. - Futurism

Last year, the artist Ai Weiwei celebrated the the Chinese government returning his passport by putting on no less than four exhibitions in New York, including even a thrift shop in Soho that was in fact stocked with the abandoned belongings of thousands of refugees forced to relocate to a camp on the border of Greece and Macedonia.

Ai's work continues to spotlight sociopolitical crises, of which there is no shortage these days. This week, he unveiled an expansive installation in collaboration with the architects Jacques Herzog and Pierre de Meuron inside the Park Avenue Armory on Manhattan's Upper East Side (this exhibition comes on the heels of the 13 Cate Blanchetts that were projected throughout the cavernous Drill Hall). (It's not the first time Ai has collaboarted with Herzog and de Meuron: They have worked together for the past 15 years on projects like the 2008 Beijing Olympic Stadium which Ai later said he regretted taking part in because the games were "merely a stage for a political party to advertise its glory to the world.")

Hansel & Gretel , as the installation is eerily called, also happens to be interactive, whether visitors like it or not. From the moment they step into Drill Hall, each of their movements is tracked and monitored via drones. Unlike the artist Jordan Wolfson's equally chilling yet slightly more menacing robot , which employed similar technology to lunge at viewers, though, each visitor is then simply projected back onto the installation, as a white light follows them to make sure they won't get lost in the darknessand so they can't avoid the cameras's glare. Still, many of them have taken to throwing up peace signsor, in the case of the artist himself, a middle fingerat the drones. And of course, they're posting about the chilling experience on Instagram . Witness their encounters, here.

Meet the Chameleons of the Art World, aka the Humans of Frieze New York:

Go here to see the original:

Ai Weiwei Is Creeping on New York with an Army of Drones, and Instagram Is Loving It - W Magazine

BGR

Tim Cook says Apple's AI is already watching you BGR When you talk about Silicon Valley tech giants using AI to improve an existing product, there's a decent chance you're thinking of Google. The recent I/O developer conference confirmed that Google thinks AI and machine learning will be all that matters ...

Read the original:

Tim Cook says Apple's AI is already watching you - BGR

In 1997, IBMs Deep Blue beat world chess champion Gary Kasparov, the first time an AI technology was able to outperform a world expert in a highly complicated endeavor. It was even more impressive when you considerthey were using 1997 computational power. In 1997, my computer could barely connect to the internet; long waits of agonizing beeps and buzzes made it clear the computer was struggling under the weight of the task.

Even in the wake of Deep Blues literally game-changing victory, most experts remained unconvinced. Piet Hut, an astrophysicist at the Institute for Advanced Study in New Jersey, told the NY Times in 1997 that it would still be another hundred years before a computer beats a human at Go.

Admittedly, the ancient game of Go is infinitely more complicated than chess. Even in 2014, the common consensus was that an AI victory in Go was still decades away. The world reigning champion, Lee Sedol, gloated in an article for Wired, There is chess in the western world, but Go is incomparably more subtle and intellectual.

Then AlphaGo, Googles AI platform, defeated him a mere two years later. Hows that for subtlety?

In recent years, it is becoming increasingly well known that AI is able to outperform humans in much more than board games. This has led to a growing anxiety among the working public that their very livelihood may soon be automated.

Countless publications have been quick to seize on this fear to drive pageviews. It seems like every day there is a new article claiming to know definitively which jobs will survive the AI revolution and which will not. Some even go so far to express their percentage predictions down to the decimal point giving the whole activity a sense of gravitas. However, if you compare their conclusions, the most striking aspect is how wildly inconsistent the results are.

One of the latest entries into the mire is a Facebook quiz aptly named Will Robots take My Job?. Naturally, I looked up writers and I received back a comforting 3.8%. After all, if a doctor told me I had a 3.8% chance of succumbing to a disease, I would hardly be in a hurry to get my affairs in order.

There is just one thing keeping me from patting myself on the back: AI writers already exist and are being widely used by major publications. In this way, their prediction would be like a doctor declaring there was only a 3.8% chance of my disease getting worseat my funeral.

All this begs the question: why are these predictions about AI so bad?

Digging into the sources from Will Robots take My Job gives us our first clue. The predictions are based on a research paper. This is at the root of most bad AI predictions. Academics tend to view the world very differently from Silicon Valley entrepreneurs. Where in academia just getting a project approved may take years, tech entrepreneurs operate on the idea of what can we get built and shipped by Friday? Therefore, asking academics for predictions on the proliferation of industry is like asking your local DMV about how quickly Uber may be able to gain market share in China. They may be experts in the vertical, but they are still worlds away from the move fast and break stuff mentality that pervades the tech community.

As a result, their predictions are as good as random guesses, colored by their understanding of a world that moves at a glacial pace.

Another contributing factor to bad AI predictions is human bias. When the question is between who will win, man or machine,we cant help but to root for the home team. It has been said, that it is very hard to make someone believe something when their job is dependent on them not understanding it. Meaning the banter around the water-cooler at oil companies rarely turns to concerns about climate change. AI poses a threat to the very notion of human based jobs, so the stakes are much higher. When you ask people who work for a university the likelihood of AI automating all jobs, it is all but impossible for them to be objective.

Hence the conservative estimations to admit that any job that can be taught to a person can obviously also be taught to an AI would fill the researcher with existential dread. Better to sidestep the whole issue and say that it wont happen for another 50 years, hoping theyll be dead by then and it will be the next guys problem.

Which brings us to our final contributing factor, that humans are really bad at understanding exponential growth. The research paper that Will Robots Take My Job was from 2013. The last four years in AI might well have been 40 years based on how much has changed. In fact, their bad predictions make more sense through this lens. There is an obvious bias for assuming jobs that require decision making as more safe than those that are straight routine. However, the proliferation of neural net resources are showing that AI is actually very good at decision making, when the task is well defined.

The problem is our somewhat primitive reasoning tends to view the world in linear reasoning. Take this example often used on logic tests. If the number of lily pads on a lake double every day, and the lake will be full at 30 days, how many days will it take for the lake to be half full? A depressingly high number of peoples knee jerk response would be 15. The real answer is 29. In fact, if you were viewing the pond the lily pads wouldnt appear to be growing at all until about the 26th day. If you were to ask the average person on day 25 how many days until the pond was full they might rightfully conclude decades.

The reality is AI tools are growing exponentially. Even in their current iteration, they have the power to automate at least part of all human jobs. The uncomforting truth that all these AI predictions seek to distract us from is that no job is safe from automation. Collectively we are like Lee Sedol in 2014, smug in our sense of superiority. The coming proliferation of AI is perhaps best summed up in the sentiments of Nelson Mandela: It always seems impossible until is it done.

Aiden Livingston is the founder of Casting.AI, the first chatbot talent agent.

View original post here:

Why are AI predictions so terrible? - VentureBeat

GETTY

As the development of true artificial intelligence (AI) continues and experts work towards the singularity the point where machines will become smarter than humans researchers are examining ways of how to keep humans as the top beings on Earth.

Many experts have warned on the perils of developing machines that are as capable as us, as it could realistically make humans obsolete as they could take our jobs, and eventually see us as more of a hindrance and wipe us off the face of the Earth.

To combat this threat, scientists are proposing ideas that could prevent this, with one new idea being that developers should give AI self-doubt.

The idea goes that if AI has self-doubt, it will need to seek reassurance from humans, much in the same way a dog does, which will consolidate our place as top of the totem on Earth.

GETTY

A team from the University of California has conducted studies and shows that self-doubting robots are more obedient.

The team wrote in a paper published on arXiv: "It is clear that one of the primary tools we can use to mitigate the potential risk from a misbehaving AI system is the ability to turn the system off.

GETTY

"As the capabilities of AI systems improve, it is important to ensure that such systems do not adopt subgoals that prevent a human from switching them off.

Our goal is to study the incentives an agent has to allow itself to be switched off."

In one simulation, a robot mind was turned off by a human and allowed to turn itself back on.

Robots without the self-doubt reactivated themselves, but the one that did have it did not as it was uncertain of the outcome if it went against human wishes.

Asus

1 of 9

Asus Zenbo: This adorable little bot can move around and assist you at home, express emotions, and learn and adapt to your preferences with proactive artificial intelligence.

The team concluded: "Our analysis suggests that agents with uncertainty about their utility function have incentives to accept or seek out human oversight.

Thus, systems with uncertainty about their utility function are a promising area for research on the design of safe AI systems.

Read more:

Give AI self doubt to prevent RISE OF THE MACHINES, experts warn - Express.co.uk