Daily Archives: June 6, 2017

French startup Clustree just raised a $7.9 million Series A round (7 million) from Creandum with Idinvest Partners and Alven Capital also participating. Clustree leverages machine learning to help both good employees who feel stuck in their jobs and HR departments who might not think to check their own companies for the perfect candidate.

In order to do this, Clustree has structured more than 250 million career paths from various sources. Big French companies can then tap into this data to get recommendations about who they should hire next for this job opening. The service combines this data set with internal data as well as continuous feedback from HR managers.

Internal candidates are sometimes better for new job positions, which helps when it comes to employee retention. Clustree can also help you with external recruitment.

Our offer covers the employee lifecycle, from recruitment to succession plan, founder and CEO Bndicte de Raphlis Soissan told me On the recruitment part, Clustree focuses on the natural talent pool of the company: it means that we are analyzing the profiles of their existing employees and the resumes they naturally received. When we deliver recruitment recommendations, our artificial intelligence will analyze that whole unique pool to find interesting candidates.

Companies like Orange, Crdit Agricole, SNCF, Carrefour and LOral are all using Clustree. And they pay quite a lot of money to access the solution.

All our customers are French companies but with an international positioning. Our solution is used across 30 different countries, Bndicte de Raphlis Soissan said. It means that we help recruit and manage careers for American, Japanese, Chinese and German people for instance, even if they all work for a leading french company.

With todays funding round, the company plans to hire more people across the board. The plan is to get more clients in France and then think about international expansion.

Continue reading here:

Clustree grabs $7.9 million for its AI-powered recruitment service - TechCrunch

Like the rest of the tech world, Apple wants to make AI on your mobile device as fast and powerful as possible. Thats why the company unveiled a new machine learning framework API for developers today named Core ML.

The key benefit of Core ML will be speeding up how quickly AI tasks execute on the iPhone, iPad, and Apple Watch. This could cover everything from text analysis to face recognition, and should have an effect on a broad category of apps. It means, says Apple, that image recognition on the iPhone will be six times faster than on Googles Pixel.

Core ML will support a number of essential machine learning tools, including all sorts of neural networks (deep, recurrent, and convolutional), as well as linear models and tree ensembles. And because this is Apple, theres also a privacy focus, too Core ML is for on-device processing, meaning the data that developers use to improve user experience wont leave customers phones and tablets.

Apple isnt the only tech company looking to make AI work better on mobile, though, and this announcement fits an industry-wide trend. Both Google and Facebook have previously announced versions of their machine learning frameworks optimized for mobile devices, and chip-maker Qualcomm has created its own software (named the Neural Processing Engine) to smooth the mobile AI experience. Machine learning: its not just happening in the cloud anymore.

Link:

Apple announces new machine learning API to make mobile AI ... - The Verge

In BriefA startup operating out of Tel Aviv, Israel is looking tochange the way artificial intelligence is able to perceive theworld. Cortica boasts of having a transparent and verifiablemachine learning algorithm that makes it possible for machines tolearn, classify, and represent things without guidance. Seeing the World as We Do

On the second floor of a small office building in the middle of Tel Aviv, the bustling heart of Israels booming tech industry sits the world headquarters for Cortica, an AI company with the ambitiousgoal of getting machines to see the world as well as we do.Click to View Full Infographic

They are one of hundreds of AI startups that have sprouted up all over the world in the last few years. The global AI market has now exceeded a billion dollarsandthe tech giants are racing to acquire them as they are re-positioning themselves as AI companies first.

But its a convoluted space with a high knowledge barrier of entry. Most of these companies use buzz words like machine learning, deep learning or neural networks knowing that most consumers and investors have no idea what they really mean.For someone without a background in the field it can be hard to distinguish substance from snake oil.

What makes Cortica different is that unlike most machine learning algorithms that are essentially blackboxes, systems where even the programmers dont know what the system is doing, Corticas program is transparent and verifiable.

It also relies on a branch of AI research called unsupervised learning. Most AI companies rely on algorithms that have to be taught by humans how to characterize and represent whatever they are learning. It is a much slower and more tedious process that has far less chance of revealing anything novel. Unsupervised learning algorithms, such as the ones Cortica uses, are capable of learning, classifying and representing things without guidance.Cortica was founded by Professor Josh Zeevi and two of his doctoral graduates, Igal Raichelgauz and Karina Odinaev from Technion Israel Institute of Technology. Israel, a tiny country of just over 8 million people, has emerged as a global hub of technical innovation. Israel now has more companies listed on the New York stock exchange than any country besides America and China and 50% of those Israeli firms can trace their origins to Technion. If Corticas aspirations come to fruition, it will catapult itself to the top of a long list of successful

Cortica was founded by Professor Josh Zeevi and two of his doctoral graduates, Igal Raichelgauz and Karina Odinaev from Technion Israel Institute of Technology. Israel, a tiny country of just over 8 million people, has emerged as a global hub of technical innovation. Israel now has more companies listed on the New York stock exchange than any country besides America and China and 50% of those Israeli firms can trace their origins to Technion. If Corticas aspirations come to fruition, it will catapult itself to the top of a long list of successfulIsraeli startups that have earned the country its moniker as the startup nation.

As Prof. Zeevi states, There is no objective reason why computers should not do better than the human brain. Although no one is there yet, Cortica will get there. Their team, with offices in New York and Beijing, believe they have figured out how to reverse engineer the biological visual cortexto enable machines to see the world as well as we can.

Doing so is no trivial task as an incredible amount of processing goes on in your head whenever you open your eyes. As soon as your eyelids flicker open receptors in the back of your eye take in the visible light waves and convert it to electrical signals that relay that message to the back of your brain to sort and analyze the immense amount of information contained into the size, shape, depth, and color of all the objects in view. All that data then gets sent for further processing to your cortex which classifies everything into objects by comparing them to every other object you have ever encountered. This is how you recognize things, make sense of them and determine their function. This happens tens of times per second and gives you a sense of motion as each frame is represented by a selected group of neurons that constitutes a clique a fundamental concept in Corticas technology that was motivated by the neurophysiological function exhibited by cortical networks of neurons. This visual response is triggered instantaneously and gets compared with Corticas extensive data base of previously triggered clustered and stored image concepts represented by highly-compressed signatures of neural cliques. All told, it is a mind boggling amount of processing that happens at every moment your eyes are open and is a process that Cortica believes they have managed to accurately simulate in silicon.

A number of companies now claim to be on the verge of getting machines to do this task with the same ease and fluidity that humans can. Cortica believes that their unique approach to the problem puts them ahead of the crowd and that they will be the first to come up with a visual system on par with our own.

This will have a wide range of applications and revolutionize a number of different fields. Everything from security cameras to autonomous vehicles to satellite imagery to medical diagnostics will be vastly improved by the application of this technology.

But that is just the beginning of what the people at Cortica believe their system will be capable of.

Endowing computers with a sense of sight will be a monumental step forward in AI research. Coupled with advances in natural language processing, AI will then have the fundamental pieces needed to understand and interact with the world. This will enable AI research to move into its next phase which is the search for the holy grail of computer science; AGI, artificial general intelligence.

Unlike narrow AI that is designed to take in and spit out specific information, AGI will be able to take in a variety of inputs and give a variety of outputs. In many ways this is what separates humans from computers, computers need specific instructions and can only output what they are directed to output. But humans can take in a wide range of inputs through our various senses and then do different tasks with that information. In the eyes of programmers this makes us general purpose computers.

But we too are limited, mostly by the amount of information that we can process and the speed at which we can output information. People can really only do one task effectively at a time and can only send information at the speed at which we can talk or write. AGI has the potential to far exceed us in both.

Cortica believes that in time their technology will be a vital part of bringing about AGI. If successful, it will be the last tool we will ever need to create.

Read more:

An Israeli AI Company Is Giving Machines The Gift Of Sight - Futurism

Theres no question that the Mario Karts AI is a big cheater. Anyone who plays the game has probably noticed how the computer-controlled racers tend to catch up to you no matter what and always have an item in hand during the races final moments. But what advantages, exactly, do the robot racers get in comparison to their human opponents? The question has become relevant once more with the new Mario Kart 8 Deluxe for the Switch, in which the artificial intelligence appears to have snapped the rules of fair play in two. Or it could all be a matter of psychology, as some experts have argued.

Ive been trying to unlock every last vehicle and accessory in Deluxe. As I plowed through, I played against a lot of AI competitors. As I played, I planned to earn all three stars for every single prize cup in every level, which you can only achieve through first-place finishes. Heres the thing, though: that Mario Kart AI really doesnt want a lowly human like me to place first.

I mean, sure, human competitors dont want me to win, either. But, I swear, the AI racers in Mario Kart 8 Deluxe appear to ignore the other AI drivers and save their items just for me. Ive seen robot racers hold onto a red shell for ages, only to toss it my way as soon as I finally try to pass them.

It feels personal.

Back in the 1990s and early 2000s, people who played Mario Kart 64 and Mario Kart: Double Dash noticed that the AI had the ability to speed up after falling far behind. It was easy to spot this rubber-band effect in action in those games. More recent Mario Karts have gotten more subtle about providing a helping hand to AI racers. When Mario Kart 8 came out for the Wii U in 2014, some players noticed that the AI was using classic shortcuts on tracks. A computer controlled Mario or Bowser or whoever would use a speed boost right before cutting across a grassy inlet. Human players have done this for years, but AI racers never seemed smart enough to do it before MK8.

Some Mario Kart 8 players have also noticed that the computers speed varies based on how fast the human player goes at the outset of the race. For example, if you hang back and wait for several seconds at the start of a race, your AI opponents will go slower throughout the rest of their laps. Once you finally head out, youll easily be able to outrun them, as this video shows:

That seems only fair, though, right? If you were a bad player, youd be grateful for the AIs rubber-band speed alterations, since it would allow you to catch up and feel like a contender.

With the new version of 8, something seems to be going on with how the computer racers use items. Or its my imagination. Somethings up, though. Heres the deal:

As usual with Mario Kart, item drops in MK8D become less and less useful the longer you stay in the first position. Youll pick up a lot of low-value items like coins and bananas. Every now and then, youll get the best defensive item of all: the Super Horn, which can knock out a Blue Shell if you time it right. MK8D brings back two legacy items from earlier in the series: the Cape Feather and the Boo. That pesky little ghost is a great get for your opponents, because the Boo can steal items. If youre in first, you might have to waste your Super Horn on an incoming Boo, thereby leaving yourself open to a shell takedown. If youre doing well, youre not supposed to get these items, but guess who does get these items when youre doing well?

The higher in speed you go in Mario Kart 8 Deluxe, the more petty the AI gets and the luckier it seems to get with items. Its easy enough to place first in every 50cc and 100cc level and clean up plenty of stars and coins in the process, although the AIs anti-human opportunism will still stick out to you in these early tracks. The 150cc tracks, however? Youll have to get mighty clever in order to beat the AI in those stages. Youll hit the final few turns before crossing the third laps finish line and, somehow, the second-place AI rival will have a Boo and a shell. As soon as you get knocked back to second, youre screwed, because every other AI next in line has also been saving their items for you, and as soon as youre in striking distance, they deploy the shell onslaught. Its as though the robot racers know that theres just one human in the running and theyve got it in for her.

Nintendo didnt respond to my asking if the item distribution in MK8D offers any advantage to the AI racers. I have no proof that my AI rival always gets the Lightning Strike, the Boo, and/or a shell in the final lap. And perhaps its just my imagination, anyway. Perhaps, as a human player, Im biased against robot racers because I feel like theyve got a one-up on me.

Could it all be in my mind?

If a human competitor in Mario Kart gets an amazing item, I chalk that up to the games internal probability and I blame myself for not having a more precise banana-throwing accuracy. But if the computer gets great items? Well, that feels like the game is cheating. The computer is part of the game! It has more information than I do! The AI Mario has a car phone with a direct line to Nintendo and he can get any item he wants!

This episode of Game Psych, which zooms in on Mario Kart 8 Deluxes item distribution, subscribes to the theory that MK8Ds AI racers do indeed get better items, but that this is part of the games attempt to balance out human racers ability to problem-solve and find counter strategies:

According to Game Psych, the computer deserves better items: Humans are better drivers than AIs in Mario Kart. Humans can pull off moves AIs never will. To keep up with the humans, the AI drivers have to rely on the power items they receive from boxes, and that makes them feel cheap or unfair.

Even if thats true, its a delicate balance to strike. If the AIs items are too good, then the game swings back to feeling unfair again. I want to feel challenged, sure, but I dont want to feel like earning three stars in each race is impossible.

Other game developers have discussed this problem with designing AI competitors before. Take Civilization lead creator Sid Meier, who gave a talk at GDC in 2010 titled The Psychology of Game Design (Everything You Know Is Wrong). In the talk, he explained that players want the AI to feel like a human competitor, but that if the AI is too good, the human opponent will suspect the robot peeked behind the curtain.

I rewatched several of my replays to find more definitive proof that the AI players tend to gang up on me. I can only observe what the robots do when theyre close to me, but you know what? They totally hate me. They aim their items at me whenever possible. However, I did notice one instance when AI players too far down the track to identify lobbed shells at one another. So I guess the robots do fight when Im not in shooting distance, but if they can see me, they do all they can to take me out.

But lets look at it from the robots perspective, shall we? Im an unfair opponent myself! I have the ability to drag a banana behind me or save a Super Horn for the perfect moment. Ive got those killer benefits that come with being human: creativity and unpredictability.

Philip Tan, creative director for the MIT Game Lab, described this phenomenon to Compete as an example of how players can do things against AI that AI cant do to players, and players dont consider that cheating. Although the AI has become increasingly humanlike since Mario Kart 8, having gained more drifting and shortcut capabilitiesthe robot racers still cant match my cleverness with counter-items and blocking. If they could, then perhaps they wouldnt need the extra Boo and Lightning Bolt drops. Not that I can prove theyre getting more Boos and Lightning Bolts, because I cant, but come on. They totally are.

Tan also remarked that players rarely notice when AI racers attack other AI racers, which my own experiences have borne out. Ive listed plenty of moments here when the AI racers seemed to gang up on me, but the rest of the time, I pay them no mind because Im focused on my own strategies and racing. I only notice the AI behavior when it affects me.

Above all, I remember the times that the AI felt unfair the most. Those moments stick out the most in my mind.

I dont think about all the races that I won with no trouble. After all, those races were perfectly fair.

See the rest here:

Either I'm Paranoid Or Mario Kart 8 Deluxe's AI Is Cheating - Kotaku

Forbes

AI Influencer Andrew Ng Plans The Next Stage In His Extraordinary Career Forbes Andrew Ng is one of the foremost thinkers on the topic of artificial intelligence. He founded and led the Google Brain project which developed massive-scale deep learning algorithms. In 2011, he led the development of Stanford University's main ...

Continued here:

AI Influencer Andrew Ng Plans The Next Stage In His Extraordinary Career - Forbes

An attack by artificial intelligence on humans, said Google software engineerand University of Michigan professor Igor Markov, would be sort of like when the Black Plague hit Europe in the 14th century, killing up to 50percent of the population.

Virus particles were very small and there were no microscopes or notion of infectious diseases, there was no explanation, so the disease spread for many years, killed a lot of people, and at the end no one understood what happened, he said. This would be illustrative of what you might expect if a superintelligent AI would attack. You would not know precisely whats going on, there would be huge problems, and you would be almost helpless.

Rather than devising technological solutions, in a recent talk about how to keepsuperintelligent AI from harming humans, Markov looked to lessons from ancient history.

Markov joined sci-fi author David Brin and other influential names in the artificial intelligence community Friday at The AI Conference in San Francisco.

One lesson fromearly humans that could help in the fight againstAI: make friends. Domesticate AI the same way Homo sapiens turned wolves into their protectors and friends.

If you are worried about potential threats, then try to use some of them for protection or try to adapt or domesticate those threats. So you might develop a friendly AI that would protect you from malicious AI or track unauthorized accesses, he said.

Markovbegan and ended his presentationby calling himself an amateur and saying he doesnt have all the answers, but he also said he has been thinking about ways to prevent an AI takeover for more than a year.He now believes the most important way for humans to prevent the rise of malicious AIis to put in a series of physical world restraints.

The bottom line here is that intelligence either hostile or friendly would be limited by physical resources, and we need to think about physical resources if we want to limit such attacks, he said. We absolutely need to control access to energy sources of all kinds, and we need to be very careful about physical and network security of critical infrastructure because if that is not taken care of, then disasters can obviously happen.

Calling upon a background in hardware design, Markov suggested steps be taken to separate powerful systems and have deficiencies built in to act as a kill switch, because if superintelligent AI ever arises, it will likely be by accident.

He also implores thatlimits be placed on self repair, replication, or improvement of AI, and urges that specific scenarios be considered, such as a nuclear weapons attack or use of biological weapons.

Generally, each agent, each part of your AI ecosystem needs to be designed with some weakness. You dont want agents to be able to take over everything, right? So you would control agents through these weaknesses and separation of powers, he said. In the discipline of electronic hardware design, we use obstruction hierarchies. We go from transistors to CPUs to data centers, and each level typically has a well-defined function, so if youre looking at this from the perspective of security, if you are defending against something, you would want to limit or regulate every level, and you would want the same type of limitations for AI.

Markovs presentation relies on predictions made by Ray Kurzweil, who believes that in a decade, virtual reality will be indistinguishable from real life, after which computers will surpass humans. Then, through augmentation, humans will become more machine-like until we reach the Singularity.

Markov also pointed out that there is a range ofopinions on malicious AI. Stephen Hawking believes AI will eventually supersede humankind, telling the BBC, The development of full artificial intelligence could spell the end of the human race.

In contrast, former Baidu AI head Andrew Ng said last year that people should be as concerned about malicious AI as they are about overpopulation on Mars.

View original post here:

Watch Google's Igor Markov explain how to avoid the AI apocalypse - VentureBeat