In a recent paper published on the preprint server Arxiv.org, researchers at Carnegie Mellon, Facebook, and the University of Illinois Urbana-Champaign propose Active Neural Simultaneous Localization and Mapping (Active Neural SLAM), a hierarchical approach for teaching AI agents to explore environments. They say that it leverages the strength of both classical and AI-based path- and goal-planning methods, making it robust against errors and sidestepping the complexities associated with previous approaches.

Techniques like those underpinning Active Neural SLAM could greatly advance the state of the art in robotics. Navigation, which in this context refers not only to coordinate navigation but to pathfinding (i.e., finding paths to objects), is a critical task for autonomous machines. But training those machines to learn about mapping requires a lot of computation.

Active Neural SLAM, then, works with raw sensory inputs such as camera images and exploits regularities in the layouts of environments, enabling it to achieve the same or better performance than existing methods while requiring a fraction of the training data.

The neural SLAM module within Active Neural SLAM comprises a Mapper and a Pose Estimator. The Mapper is responsible for generating a top-down spatial map of a given environment and predicting obstacles and explored areas, while the Pose Estimator anticipates the agent’s pose based on past pose estimates. The spatial map — where each element corresponds to a cell size of 25 square centimeters in the physical world — is ingested along with the agent pose by a global policy to produce various long-term goals. A Planner model then takes the goals, the spatial obstacle map, and the agent pose estimates to compute short-term goals, or the shortest paths from the current location to the long-term goals. Lastly, a local policy outputs navigational actions using camera data and the short-term goals.

In experiments, the researchers paired Facebook’s open source Habitat platform, a modular high-level library for training agents across a variety of tasks, environments, and simulators, with data sets (Gibson and Matterport’s MP3D) consisting of 3D reconstructions of real-world environments like office and home interiors. Agents could make one of three moves — forward 25 centimeters, leftward 10 degrees, or rightward 10 degrees — in the environments and were trained in 994 episodes consisting of 1,000 steps or 10 million frames, such that all of Active Neural SLAM’s components — the Mapper, the Pose Estimator, the global policy, and the local policy — were trained simultaneously.

The team reports that the Active Neural SLAM managed to almost completely explore small scenes in around 500 steps versus the baselines’ 85% to 90% exploration of the same scenes in 1,000 steps. The baseline models also tended to become stuck in areas, indicating that they weren’t able to “remember” explored areas over time — a problem that Active Neural SLAM didn’t exhibit.

Encouraged by these results, the coauthors deployed the trained Active Neural SLAM policy from simulation to a real-world Locobot robot. After adjusting the camera height and vertical field-of-views to match those of the Habitat simulator, they say that it successfully explored the living area in an apartment.

“In the future, [Active Neural SLAM] can be extended to complex semantic tasks such as semantic goal navigation and embodied question answering by using a semantic Neural SLAM module, which creates a … map capturing semantic properties of the objects in the environment,” wrote the coauthors. “The model can also be combined with prior work on localization to relocalize in a previously created map for efficient navigation in subsequent episodes.”

The post Facebook’s AI teaches robots to navigate environments using less data appeared first on Artificial Intelligence.

Marilyn Monroe is back starring in movies via an avatar program that talks and acts like her, with machines having learned her speech and mannerisms from her films.

Indeed, there will be a bot version of us that lingers on after we die, that reads our will to relatives and friends, and consoles them. Maybe you can toast yourself at your wake. A bad bot version may be out-and-about settling old scores.

Computers hire and fire, and planes, trains and ghost ships crisscross the world without any humans aboard.

Do you like this future? Is it plausible? Or is it the work of an overly fertile imagination?

Much depends on who makes the predictions because futurists are a mixed bunch. Some are fiction writers with a sense of tomorrow and beyond. Some have a sociology background and see us moving in these directions.

Some have a technology background and their predictions are an extrapolation of what’s possible with tech today. Their predictions are scary because they are likelier to be right.

When the futurist is an eminent Australian professor of artificial intelligence, it’s even more frightening. Such is the case in It’s Alive! Artificial Intelligence From the Logic Piano to Killer Robots, a new book by Toby Walsh, professor of artificial intelligence at the University of NSW and a leader in research at Data61, Australia’s centre for information and communications technology research.

“There are few other human inventions that are likely to have as large an impact on our lives as machines that can think,” he says. The frightening truth, he adds, is that AI is already an indispensable part of our lives without most of us realising it.

Walsh’s book goes back to 1950, when there were fewer than a dozen computers worldwide. “Each was a glorious combination of vacuum tubes, relays, plug boards and capacitors that filled the room,” Walsh says.

He credits Alan Turing, the man who led Britain’s successful effort to crack the Nazi Enigma code in World War II, as behind many ideas permeating computer science today, and helping start the field of artificial intelligence.

“I predict that, when the 21st-century ends, Alan Turing will be most remembered for laying the foundations to this field that is trying to build thinking machines.”

Walsh’s book is pitched at everyday readers, not just geeks. It includes a history of advances in mathematics leading up to AI as a study in its own right, circa 1956. He says using computers to translate from one language to another was proposed as early as 1946, and translation projects were launched in the 1950s and 60s.

He says speech recognition systems also were developed in the 50s and 60s but failed. In 1952, Bell Labs had built a system that could recognise single digits, although it was limited to a single speaker.

In 1969, John Pierce, who led the development of the first commercial communication satellite, Telstar, likened speech recognition to “schemes for turning water into gasoline, extracting gold from the sea, curing cancer or going to the moon”.

But there were successes such as Shakey, the first robot to perceive its environment, ELIZA, a 1960s attempt at a computerised psychotherapist. Walsh says Eliza cheated. She’d invert what a patient said into the next question and had limited understanding of the semantics of a conversation.

IBM Watson’s ability to beat humans at the TV game Jeopardy! and become quiz champion in 2011 isn’t the first example of computers triumphing over humans. Walsh writes that in 1979, a computer program called BKG 9.8 beat the world’s backgammon champion in a $5000, winner-take-all match in Monte Carlo.

The second part of Walsh’s book discusses the present, including the state of machine learning. “It would be wrong to conclude, however, that machine learning has brought us very close to thinking machines and that, with a little more refinement, techniques such as deep learning will ‘solve’ intelligence,” Walsh says.

“One reason that deep learning is not the end of the game is that it needs loads of data.”

But machine learning is maturing as a technology, and could solve many problems without too much help from humans.

Walsh discusses automating reasoning, the state of robotics, and robots in theatres of war. “An army race is under way to automate warfare. The Pentagon has allocated $18 billion in its current budget for the development of new types of weapons, many of them autonomous,” he says.

There are surveillance drones, and on land Boston Dynamics has developed two-legged and four-legged robots that carry loads for soldiers. In the ocean, there are robotic minesweepers and even autonomous submarines.

Walsh mentions Boeing’s 15.6m-long autonomous submarine, Echo Voyager, which can spend six months underwater with a range of about 12,000km.

Walsh has been active in advocating the banning of autonomous weapons that can select and engage targets without human intervention. In 2015, he and fellow researchers collected 1000 signatures from those in university AI and robotics labs globally, calling for the ban.

In the domestic sphere, he cites recent research that will make it possible for robots to run, iron and fold clothes, and catch balls. If robots can do laundry, I won’t mind that sort of future.

He profiles several professions where AI is playing an increasing role and asks whether the machines should or can take over, and the idea of “universal basic income” in the machine age, given many jobs are automated.

As for his 10 predictions about the future, he is probably right. Hollywood actors may indeed play roles in new films long after their death, although by 2050 there may not be so much interest in reviving Monroe, an actor from 100 years before, as there will be for actors of that time.

Walsh’s idea that we’ll send our cars off to earn money as taxis while we work is well within the sights of the autonomous car industry. “Within 15 to 20 years, most of us can expect to be driven around in autonomous cars,” he says.

The role of fitness watches and computers monitoring our health daily is on track, too. The technology is there but not properly organised. Walsh predicts smart­phones will regularly take photos of you, to identify melanomas and monitor the health of your eyes. Your voice will be analysed to identify colds, and even a stroke or Parkinson’s disease.

His prediction that computers will hire and fire you is an extrapolation of applications already being developed using IBM Watson intelligence. But it won’t end there.

“Computers will increasingly take over many of the tasks of managing you during your employment. Programs will schedule your activities, approve your holidays, as well as monitor and reward your performance,” Walsh writes.

“In December 2016, Bridgewater Associates, one of the world’s largest hedge funds with over $100bn under management, announced a project to automate the day-to-day management of the firm, including hiring, firing and other strategic decision-making.”

He says our device interfaces will disappear, and in their place will be conversations. “No more typing. No more pointing. Just speak, and the device will perform complex tasks for you.”

Walsh predicts that by 2050 a robot will have robbed a bank, but it will be achieved by stealth rather than an armed robbery.

And by that time the nightly television news will be made without a single human being involved in the production, he says. That is plausible, as newer smartphones can automatically edit videos. The tech is there.

As for virtually living as a bot after you die, the prospects are frightening. For example your social media presence could continue on. And celebrities will use bots to create a social media presence, responding to Facebook messages, tweeting and Instagramming in response to events.

Who knows: by 2050 Facebook may be mainly a machine social network, with most posts, comments and likes made by machines. This could be a worrying proposition for Mark Zuckerberg. That’s unless he leaves the running of Facebook to a Zuckerberg bot, who may love it.

The post Artificial intelligence gives robots and autonomous machines a future appeared first on Artificial Intelligence.