Daily Archives: June 24, 2017

Artificial intelligence has made great strides in the past few years, but its also generated much hype over its current capabilities.

Thats one takeaway from a Friday panel in San Francisco involving leading AI experts hosted by the Association for Computing Machinery for its 50th annual Turing Award for advancements in computer science.

Michael Jordan, a machine learning expert and computer science professor at University of California, Berkeley, said there is way too much hype regarding the capabilities of so-called chat bots. Many of these software programs use an AI technique called deep learning in which they are trained on massive amounts of conversation data so that they learn to interact with people.

Get Data Sheet , Fortunes technology newsletter.

But despite several big tech companies and new startups promising powerful chat bots that speak like humans when prodded, Jordan believes the complexity of human language it too difficult for bots to master with modern techniques like deep learning. These bots essentially perform parlor tricks in which they respond with comments that are loosely related to a particular conversation, but they cant say anything true about the real world.

We are in era of enormous hype of deep learning, said Jordan. Deep learning has the potential to change the economy, he added, but we are not there yet."

Also in the panel, Fei-Fei Li, Googles ( goog ) machine learning cloud chief and Stanford University Professor, said We are living in one of the most exciting and hyped eras of AI. Li helped build the ImageNet computer-vision contest, which spurred a renaissance in AI in which researchers applied deep learning to identify objects like cats in photos.

But while everyone talks about ImageNets success, we hardly talk about the failures, she said, underscoring the hard work researchers have building powerful computers that can see like humans.

For more about technology and finance, watch:

Still, Li is excited that current AI milestones will eventually lead to more breakthroughs that will touch every single industry, like healthcare. We are entering a new phase in AI, she said.

What will help usher more breakthroughs in deep learning will be the continuing advancements in powerful computing hardware, like Nvidia's GPUs that make it possible to crunch tremendous amounts of data faster than ever, explained Ilya Sutskever, the research director of Elon Musk-backed AI research group OpenAI . Deep learning will keep booming in tandem with advancements in computing hardware that shows no signs of slowing down .

"Compute has been the oxygen of deep learning," Sutskever said.

See the original post here:

Beware the Hype of Artificial Intelligence - Fortune

Is AI over-hyped in 2017? originally appeared on Quora - the place to gain and share knowledge, empowering people to learn from others and better understand the world.

Answer by Joanne Chen, Partner at Foundation Capital, on Quora:

To quote Bill Gates We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten. Don't let yourself be lulled into inaction.

In short, over the next ten years, I dont believe AI will be overhyped. However, in 2017, will all of our jobs be automated away by bots? Unlikely. I believe the technology has incredible potential and will permeate across all aspects of our lives. But today, my sense is that many people dont understand what the state of AI is, and thus contribute to hype.

Artificial intelligence, a concept dating back to the 50s, is simply the notion that a machine can performance tasks that require human intelligence. But AI today is not what the science fiction movies portray it to be. What we can do today falls in the realm of narrow AI (vs general intelligence), which is the idea that machines can perform very specific tasks in a constrained environment. With narrow AI, there are a variety of techniques that you may have heard of. Ill use examples to illustrate differences.

Lets say you want to figure out my age (which is 31).

1) Functional programming: what we commonly know as programming, a way to tell a computer to do something in a deterministic fashion. I tell my computer that to compute my age, it needs to solve AGE = todays date birth date. Then I give it my birth date (Dec 4, 1985). There is 0% chance the computer will get my age wrong.

2) Machine learning: an application of AI where we give machines data and let them learn for themselves to probabilitically predict an outcome. The machine improves its ability to predict with experience and more relevant data. So take age for example. What if I had 1,000 data sets of peoples ages and song preferences? Song preference is highly correlated with generation. For example, Led Zeppelin and The Doors fans are mostly 40+ and Selena Gomez fans are generally younger than 25. Then I could ask the computer given that I love the Spice Girls and Backstreet Boys, how old does it think I am? The computer then looks at these correlations and compares it with a list of my favorite songs to predict my age within x% probability. This is a very simple example of using machine learning..

3) Deep Learning: is a type of machine learning emerged in the last few years, and talked widely about in the media when Google DeepMinds AlphaGo program defeated South Korean Master Lee Se-dol in the board game Go.

Deep learning goes a step further than ML in that it enables the machine to learn purely by providing examples. In contrast, ML requires programmers to tell the computer what it should look for. As a result, deep learning functions much more like the human brain. This especially works well with applications like image recognition.

4) Deep reinforcement learning: DRL goes one step further and combines deep learning with reinforcement learning which is the notion of learning by trial-and-error, solely from rewards or punishments. DRL mimics how children learn they see observe other people doing things, they try things out and depending on the reward, they either repeat them or not!

Machine learning technologies have become more available (and the reason why there has been increasing media hype around this space) has been driven by advancements in three areas:

1) Infrastructure to run ML algorithms massive improvements in storage, processing capabilities (i.e. GPUs that speed up parallel processing), and accessibility for rapid innovation (cloud).

2) New available algorithms developed.

3) Data proliferation to train algorithms.

Between algorithms innovation and data availability, I believe data plays a more crucial role in advancements. If you look at the chart below, breakthroughs in AI have been quickly followed by availability of datasets, while many of the corresponding algorithms have been available for over a decade.

AI will permeate our lives in the next ten years. Think of the possible time, money, and manpower saved by automating simple processes. And as the technology becomes more advanced, the use cases will get even more exciting. I think its a wonderful time as an entrepreneur to be able to leverage this technology, and I couldnt be more excited as an investor.

This question originally appeared on Quora - the place to gain and share knowledge, empowering people to learn from others and better understand the world. You can follow Quora on Twitter, Facebook, and Google+. More questions:

The Morning Email

Wake up to the day's most important news.

Go here to read the rest:

Is Artificial Intelligence Overhyped in 2017? - HuffPost

When you think of artificial intelligence, you may think of Lt. Commander Data or C-3PO, but this AI will actually be the spacecraft rather than on board.

Exploring space has some far-out challengesand this is after weve sent robots to Mars and all manner of probes and orbiters to other planets, including Venus and Saturn. Future missions will venture deeper and deeper into unexplored star systems and galaxies that have only been observed via telescope. This is easier dreamed than done. Too many unforeseen obstacles could cause a craft to glitch or break down hundreds of thousands of miles away, which is why scientists developing these future missions need to be paranoid.

Space scientists Steve Chien and Kiri Wagstaff of NASAs Jet Propulsion Laboratory suggest that programming probes with advanced artificial intelligence will largely eliminate the need for prompts from the home planet that would have increasing difficulty reaching out to them the further they ventured into space. Not to mention that probes on more daring missions will have to be able to think for themselves, because they even more of them will be required and they will probably not be able to receive any intervention from Earth. It gets creepier with the realization that the capacity to learn will need to be wired into their computerized brains to make them adaptable.

"The goal is for A.I. to be more like a smart assistant collaborating with the scientist and less like programming assembly code," said Chien, who collaborated with Wagstaff on an article recently published in the journal Science Robotics. "It allows scientists to focus on the 'thinking' thingsanalyzing and interpreting datawhile robotic explorers search out features of interest."

Autonomous probes should be able to function on a hypersensitive level that includes understanding and carrying out mission requirements, recognizing geological phenomena and identifying differences between what passes for normal planetary conditions (depending on the planet) and extreme space weather. They should also be able to reprioritize if they eye something spontaneous, like ocean plumes erupting on watery worlds similar to Enceladus. Advancing the science of AI enough may even make them able to use their findings for future studies. Not having infinite fuel means the robo-brains will also need to make the call on which regions are worth delving into the most.

AI is already being prototyped for the Mars 2020 rover and could someday make once-impossible endeavors, like a mission to Alpha Centauri, possible, but even the researchers themselves admit it still has to level up.

"For the foreseeable future, there's a strong role for high-level human direction," Wagstaff said. "But A.I. is an observational tool that allows us to study science that we couldn't get otherwise."

(via Phys.org)

Go here to see the original:

Space probes of the future will have artificial intelligence, and it's kind of creepy - SYFY WIRE (blog)

John Kelleher is a partner at McKinsey & Co. and the co-chair of Next Canada. Laura McGee is an engagement manager at McKinsey & Co. and co-founder of #GoSponsorHer.

John Kelleher is a partner at McKinsey & Co. and the co-chair of Next Canada. Laura McGee is an engagement manager at McKinsey & Co. and co-founder of #GoSponsorHer.

Theres no doubt that Canada could lead the planet in artificial intelligence (AI). Canadian academics such as Geoffrey Hinton and Yoshua Bengio essentially created the field of deep learning and put Canada on the map; today, Edmonton, Toronto and Montreal are globally important centres of AI research. The best AI talent in the world is also increasingly coming to Canada to launch AI businesses such as integrate.ai and others.

All these companies and researchers are convinced of the technologys enormous commercial potential. If AI develops like other technologies, most of these benefits will flow to the country that builds the first good ecosystem. This is a huge opportunity for Canada.

At the same time, AI poses clear challenges to business and government. Over the next 10 to 20 years, nearly half of Canadas jobs are at high risk of being affected by automation. Women hold a lot of these jobs and are especially at risk the World Economic Forum says that globally, women will face about twice the rate of job loss as men in what it calls the fourth industrial revolution.

How can Canadian companies gain the benefits of this disruptive technology while ensuring that large segments of society are not left behind? In our view, the public and private sectors should take six steps to outsmart AI and avoid its dislocations:

Commit to building the worlds best AI ecosystem: The winning AI cluster will create many high-paying jobs and create spillover effects for the middle class but the also-rans will not. Half-measures wont work. Canada must play to win. If there is going to be a steam engine that disrupts the status quo and AI is shaping up that way then Canada should develop and build the very best steam engine it can, right here at home.

Create at-scale AI training programs: Industry can form coalitions to collect data, oversee curriculum development and rapidly retrain workers in the skills needed to succeed in nascent AI applications.

Take Generation, a McKinsey-supported initiative that works with employers to quickly train and place young workers in sectors like health care and technology. Graduates have an 84 per cent employment rate within 90 days of completing the program and earn two to six times more income than before. Similarly, Prominp in Brazil trains 30,000 youth each year for positions in the oil and gas industry, with 189 skill-profile tracks and an 80-per-cent postgrad employment rate.

In Canada, such a program could be built in partnership with new research groups such as the Vector Institute in Toronto or with incubators such as Communitech, Next Canada and the Creative Destruction Lab.

Launch innovative new training models: The government could launch and fund a venture capital lab to create innovative training programs, so new training ideas can be tested, validated and scaled up (as recommended by the Advisory Council on Economic Growth). Startups such as Ryersons Magnet have great potential to address labour-market challenges. A so-called FutureSkills Lab could help scale great ideas and share learnings across provinces.

Build real links between companies and research schools: Large companies could partner with universities and vocational schools to provide equipment, facilities and expertise to prepare students for AI. In exchange, these companies could receive preferential recruiting.

For example, TAFE SA in South Australia trains approximately 500,000 students each year in high-demand areas such as aged care and nursing, trades and information technology. It partners with hundreds of businesses each year, which provide apprenticeships and traineeships. TAFE also orchestrates reverse co-op program where large corporations and small-to-medium-sized enterprises send workers back to campus for a term to learn critical AI skills.

Urgently reinvent curriculums for software and AI: Elementary, high-school and university programs have to develop the skills that empower students to be leaders in the coming AI tsunami critical thinking, teamwork, coding, algorithmic understanding and math. Some jurisdictions (e.g., Chicago and Queensland, Australia) are already moving to make software-coding classes mandatory. Canada should consider doing the same.

Government may want to consider practising what it preaches and adopt AI itself: A technology-enabled, AI-smart public service could not only be more efficient and provide better services. It might also create a product that Canada could export to the world.

Canadian companies have a real opportunity to leverage AI for growth but not without an inclusive work force. We all have a stake in getting this right.

Follow us on Twitter: @GlobeBusiness

Continued here:

Canada has a chance to monopolize the artificial intelligence industry - The Globe and Mail