Daily Archives: May 6, 2017

Reid Hoffman at the Tech Alliance annual luncheon in Seattle today. (GeekWire Photo / Todd Bishop)

What would happen if we put artificial intelligence on a path to enlightenment?

LinkedIn founder and Microsoft board member Reid Hoffman shared a number of thoughts on the future of artificial intelligence research during the annual State of Technology Luncheon Friday in Seattle, as part of a wide-ranging discussion hosted by the Tech Alliance.

Hoffman bypassed the more near-term concerns of how artificial intelligence and machine learning will be implemented in todays technology to think about how this headlong plunge into AI research will affect society.

Hoffman and eBay founder Pierre Omidyar recently started a $27 million fund for AI research called Ethics and Governance of Artificial Intelligence. What does AI for the public interest mean? Hoffman was asked by moderator Sarah Imbach, an investor and former LinkedIn executive. Im not sure anybody knows, he joked.

He went on to explain that ones level of anxiety about artificial intelligence research probably depends on how soon you think it will arrive, Hoffman said. If you think were 10 years away from developing computers with human-level intelligence, youre absolutely totally frenetic about it, probably a little crazy. If you dont think this will arrive for another 50 years or longer, youre probably a little more blase about the whole thing.

But in the grand scheme of things, do you know how short 50 to 100 years is on the human timeframe? he said. If theres even a 20 percent chance human-level artificial intelligence arrives by that point, we need to start thinking now about the impact it will have on society and the ethics that govern that research.

After all, computers only do what they are trained to do. If you could train an AI to be a Buddhist, it would probably be pretty good, Hoffman said.

Or, you could go in a different direction. Hoffman said he is particularly interested in what happens as artificial intelligence research starts to intersect with biology.

With the growing sophistication of neural networks, massive amounts of data, and nearly unlimited computing resources, its hard to see an easy ceiling for artificial intelligence research right now, he said. At the same time, biological research is taking advantage of many of the same resources to advance its own research, maybe increasing at a slower curve than AI, but increasing nonetheless.

At some point, those curves will intersect, and the growth of specialized artificial intelligence systems means you more or less get to a point where youre also saying that, deliberately or accidentally, you could be engineering different versions of the homo genus, like Neanderthals.

Hopefully theyll be Buddhists.

Originally posted here:

Reid Hoffman: 'If you could train an AI to be a Buddhist, it would probably be pretty good' - GeekWire

Photo courtesy ofPivotal

Artificial intelligence has a diversity problem. Too many of the people creating it share a similar background. To renowned researcher Fei-Fei Li, this paucity of viewpoints constitutes a crisis: As an educator, as a woman, as a woman of color, as a mother, Im increasingly worried, she says. AI is about to make the biggest changes to humanity, and were missing a whole generation of diverse technologists and leaders.

From the chair next to her, Melinda Gates affirms this, adding, If we dont get women and people of color at the tablereal technologists doing the real workwe will bias systems. Trying to reverse that a decade or two from now will be so much more difficult, if not close to impossible.

Both women are powerful technologists. As chief scientist of artificial intelligence and machine learning for Google Cloud, Li is currently on sabbatical from Stanford, where she directs the Stanford Artificial Intelligence Lab. Gates studied artificial intelligence in the early days of the 1980s when when she was learning to code at Duke University. She spent a decade at Microsoft before leaving and later pursuing philanthropy. Now Gates is putting her mind and her money behind a national nonprofit that Li is helping launch: AI4All.

The name says it all. AI4All will support educational programs designed to expose underrepresented high school students to artificial intelligence. I sat down with Gates and Li last week at Stanford University to talk about how to make AI research more appealing to women, why hoodies shouldnt be techs status symbol, and what it takes to work in AI.

Jessi Hempel: How did you get to know each other?

Melinda Gates: If youre at all interested in artificial intelligence, youre going to hear about Fei-Feis work. I wanted to meet her and understand what she was doing, in particular, with some of her PhD students, and what it was like for a group of females to be in the field of AI. We met. Then Fei-Fei pulled together a group of women [studying AI].

Fei-Fei Li: Melinda, when I heard that you were starting to pay attention to AI, I really had that moment of thinking, Finally. Finally, a world leader whose voice can be heard is a woman technologist and she is now paying attention to AI!

I have been in this space for many, many, many years as an educator as well as a technologist, and Ive been having this increasing worry. As a technologist, I see how AI and the fourth industrial revolution will impact every aspect of peoples lives. If you look at what AI is doing at amazing tech companies like Microsoft, Google, and other companies, its increasingly exciting.

But in the meantime, as an educator, as a woman, as a woman of color, as a mother, Im increasingly worried. AI is about to make the biggest changes to humanity and were missing a whole generation of diverse technologists and leaders. So when I heard Melinda was paying attention to this, and your people reached out to meyou dont know this, Melinda, but they reached out to me when my daughter was about four months old and I was home nursing.

Melinda Gates: So been there.

Fei-Fei Li: I was just so happy. We immediately arranged your visit and wanted to have a candid conversation. And I told the students, You guys are all extremely passionate technologists, but you are also still blazing the trail. Be candid with Melinda about your experiences.

Melinda Gates: And that was fantastic. I just want to echo one thing that Fei-Fei said: If we dont get women and people of color at the tablereal technologists doing the real workwe will bias systems. Trying to reverse that a decade or two from now will be so much more difficult, if not close to impossible. This is the time to get women and diverse voices in so that we build it properly, right? And it can be great. Its going to be ubiquitous. Its going to be awesome. But we have to have people at the table.

Fei-Fei Li: Exactly, because AI is a technology that gets so close to everything we care about. Its going to carry the values that matter to our lives, be it the ethics, the bias, the justice, or the access. If we dont have the representative technologists of humanity sitting at the table, the technology is inevitably not going to represent all of us.

Jessi Hempel: We have already seen some of the consequences of not including diverse voices in the beginning stages of development. Is it already too late?

Melinda Gates: I wouldnt say its too late but I would say that that car is speeding down the road very quickly. This is one of the reasons Fei-Fei and I are so interested in thinking about how you get female technologists into this field.

Jessi Hempel: What came of your conversations?

Fei-Fei Li: When I was coming out of maternity leave, I was thinking deeply about what I could do to really help this generation. I see this as one of the most important efforts I can make. Three years ago, I had started a test program along with my former PhD student, Olga Russakovsky. It was a pilot program called SAILORS, Stanford Artificial Intelligence Lab Outreach Summer Program. We invited high school ninth graders in the Bay Area. Its a non-residential program focusing on young women, and inviting them to spend two weeks within the AI lab.

There are two pieces of SAILORS. One: We have a strong hypothesis that the pipeline issue is deeply affected by the way that technology is presented to young students. In Silicon ValleyIve lived here for 10 years. I love Silicon Valley, but there is a dominant voice of, Tech is cool. Tech is geeky. Tech is a guy with a hoodie.

Melinda Gates: Yes!

Fei-Fei Li: The guys with hoodies have changed our world. But theyre not the only technologists. Thats not the only way to motivate people, especially young women with many choices. [Theyre thinking], I can be a doctor at the bedside saving peoples lives. I can be a journalist in the most needed area of the world giving the people a voice. Why should I be in AI or CS if all I heard is you can have a hoodie and look cool?

We add a humanistic mission into the teaching of the technology that goes to the core of what these young people are longing for. So for example, as a research project, were doing self-running cars in the robotics team for SAILORS. We wrapped it in the context of aging society, because a self-driving car, of course, is cool technology, but one of the populations its going to help the most is our increasing aging society.

Jessi Hempel: How did you decide to target ninth graders?

Fei-Fei Li: We spent a lot of time looking at past data. We realized that around early high school years is when students start to think about their college major. Theyre questioning: Who am I; what impact can I make on the world?

The program was very popular and successful. We have amazing young women. The only problem is, its not big enough. So then I started thinking, we really ought to start to spread it nationally. And this is when we started collaborating with Melinda. We started this organization called AI4All. Its still, I would say, stealth-ish.

Melinda and Jensen Huang, the founder of NVIDIA, are putting in the seed money for us. AI4All is focusing on spreading a SAILORS modelthe education of AI with humanistic mission to diverse studentsto different campuses and companies.

Jessi Hempel: Have you started rolling it out?

Fei-Fei Li: We officially started in March. Five universities are partnering with us: Berkeley, Carnegie Mellon University, Princeton, Boston University, and Simon Fraser. Theyre going to start their own chapters of SAILORS. They will tailor it to different local communities. For example, Berkeley will be more robotics focused, and will focus on low-income students. The Princeton program will be more about racial diversity, because New Jersey has a strong African-American community.

Jessi Hempel: What are the major barriers to launching something like this?

Fei-Fei Li: We have so few AI technical leaders who are diverse themselves. Also, theyre busy doing the things like building a startup or making money off publishing papers. This kind of education is longterm. Education is thankless for a long time.

Jessi Hempel: Melinda, what insight do you have from funding other organizations that could help AI4All be successful?

Melinda Gates: Fei-Fei is in the process of hiring an executive director, and shes in a very fortunate situation. Shes got a couple of really strong candidates. But were talking about the skills you need in that executive director. Because sometimes, and [Bill and I] certainly made this mistake, both ourselves and with other organizations, you think you know what you want. You have this really shiny candidate, and they have all these other skills. But if theyre actually not good at hiring, recruiting, retention, and building an organization, youre not going to succeed.

Jessi Hempel: Melinda, when we spoke last fall, you put out a call out to figure out where to focus your resources as you turned your attention to helping women succeed in tech. Is this the first piece of that effort? Are we going to see more of it?

Melinda Gates: This is one piece of it. Youll see more of it. Definitely. Since you and I talked, were funding Girls Who Code more, because I think thats another model, for sure, for getting the pipeline filled.

But Im also looking at workplace diversity. Ill make some investments there. Theres a fantastic economist at Harvard, Iris Bohnet. She does behavioral economics and shes looking at how you design diversity into a system. Shes the person who has talked about how in orchestras, women couldnt get a first chair. Finally, when [audition judges] put a curtain down so that people on the other side of the curtain couldnt see who was playing the violin on the other side, the numbers went up a little bit. But they didnt go up as much as she thought they might. She realized that the person on the other side of the curtain interviewing could hear the footsteps of the person walking across the stage. Once they fixed that, the number of female first chairs went up significantly.

So in coding, when a professor looks at a females code or a males code, weve seen the bias numbers. You just have an inherent bias. When its anonymized, guess what? The women do just as well as the men. I know a young man whos working on a fantastic young startup where you submit your code with no name. There are seven great coders reviewing the code on the other side that was submitted anonymously.

And the last thing Ill just say, the other place that Im investing is NCWIT [National Center for Women and Information Technology]. Theyre doing a great job of designing things into that first computer science course a student takes that attract women.

Jessi Hempel: There are already many women and people of color working in the field. How do we draw attention to their work?

Fei-Fei Li: Oh boy. I just tell media, please find a list of AI technologists and give them a voice, because its so convenient to pick up the phone and call that guy that is always out there. There are women and other diverse technologists. And if you need help finding them, there are people like me. Im happy to supply you with a list of AI technologists who have diverse backgrounds. I think that voice needs to be heard.

Melinda Gates: And the other thing I would just say for readers is that this is an exciting field. AI is going to change so much. So we shouldnt be afraid of it. We have to be smart about how its done. But you can learn AI. And you can learn how to be part of the industry. Go find somebody who can explain things to you. If youre at all interested, lean in and find somebody who can teach you.

Jessi Hempel: Im so glad you said that, because I think sometimes we think, well, youve got to get the ninth grader interested because its too late for the rest of us who are mid-career.

Melinda Gates: And I think sometimes when you hear a big technologist talking about AI, you think, Oh, only he could do it. No. Everybody can be part of it.

Fei-Fei Li: Our culture has a tendency to call a few of them geniuses. And then mortals just think, Were not geniuses. Its not true. If someone has a fantastic biology background, he or she can contribute in AI and health care. AI has many aspects. AI is everywhere. Its not that big, scary thing in the future. AI is here with us.

More here:

Melinda Gates and Fei-Fei Li Want to Liberate AI from Guys With Hoodies - Backchannel

According to IBM, we produce 2.5 exabytes of data each day. This is equivalent to 250,000 Libraries of Congress or 90 years of HD video each day.This data exhaust results from our continual digital interactions, whether explicit such as typing a search into Google or implicit, like the location signals we give off as we move through the world with our smartphones.

In advertising, we use these data signals for demand capture which is less expensive than demand generation usually in programmatic contexts, for targeting and analytics and attribution. From a users point of view, much of that data output is visual and text-based, such as a list of Google search results. This visual environment is a good brand too, since it contextualizes and dimensionalizes its offering.

Big Data is also an important driver of advances in artificial intelligence. AI is nothing if not data-hungry, and cheap access to the exponentially growing, cloud-stored motherlode of data means that machine learning and deep learning systems, with their sophisticated algorithms and parallelized processors, can be trained that much more effectively.

advertisement

advertisement

Its easy to get breathless when talking about AI (it happens to me all the time). What people spend less time talking about are the near-term collisions between AI and various professional fields, like medical diagnostics or finance or media and marketing.

Take Alexa, for example, the brain inside Amazons Echo and Dot. Alexa lives in your device but also in the cloud, and without sophisticated deep-learning algorithms trained on massive amounts of data, Alexa wouldnt exist at all. But when you ask Alexa for the best restaurant in Brooklyn, she names exactly four restaurants, one of them Shake Shack. And thats it. When you type the same results into Google, you get a huge number of results, richly contextualized with ratings and descriptions and locations on a map.

The results returned in text and image are far more useful to the user, and the brand, than those returned in voice. The point isnt that Alexa should use Google for its search results rather than Bing, but that, as an interface, voice is decontextualized.

You have to wonder how brands will manage discovery in a world increasingly dominated by voice. If this seems hyperbolic, consider that, by 2020, 30% of Web browsing will happen without a screen (according to Gartner) and 50% of searches will be voice (according to comScore) and that besides Alexa we have Google Assistant, Cortana, Siri, and Ozlo, to name a few.

These new ecosystems will change how we connect buyers and sellers of media, as well as the fundamental role of publishers.

The main thing a brand can do today to prepare is to get its data strategy in order. A data strategy starts with a DMP and extends from that foundation into all of the new, AI-enabled contexts its customers are going to be, whether voice or messaging, cars, homes or AR, and whatever comes next.

The time to prepare for this is now. Fundamental goals like reach, engagement, discovery and sales will be the same in these new contexts, but only the brands that have adopted a sound data strategy, informed by ongoing advances in machine learning, computer vision and natural language understanding, will benefit.

In closing, I should note that Alexa did answer one of my queries with the perfect response: Alexa, what is the meaning of life? The traditional answer is 42.

Original post:

Exploding Old Contexts With AI - MediaPost Communications

Humans will adapt to automation and benefit from it the same way they adapted to the industrial revolution, Alphabet Executive Chairman Eric Schmidt said Friday at a forum at Columbia University.

What I think is that the technology thats being built is being built in the open and for the benefit of everyone, he said.

The comments were made at Columbia University School of International and Public Affairs (SIPA) Global Digital Futures Policy Forum, where Columbia professor Merit Janow, who is dean of SIPA, was interviewing him about globalization, automation and artificial intelligence.

Schmidt said fears that advances in technology will kill certain jobs havent necessarily played out historically. For example, 20 years ago, people thought ATMs would make bank tellers unnecessary by automating their jobs.

There are more bank tellers now than ever because banks are more efficient, he said, also mentioning that despite automation in recent decades, unemployment is at its lowest in 10 years.

The statement echoed comments he made Wednesday that he is a job elimination denier.

He argued that those who think automation and other advances in technology will lead to negative effects overall miss clear economic benefits of the efficiencies achieved with technological advances.

Youd have to convince yourself that a declining workforce and an ever-increasing idle force, the sum of that wont generate more demand, he said. Thats roughly the argument that you have to make. Thats never been true.

Since the Luddites of the 19th century destroyed weaving machinery, the argument against technological advances has been that a new technology comes along and that there is materially significant displacement and there are local concerns that are very serious, said Schmidt.

But people have adapted, and internet technology in particular has made people smarter and more efficient, he said.

While the future impact of new technology could play out differently than in past technological revolutions, Schmidt said that in order to believe its different now, you have to believe that humans are not adaptable, that theyre not creative.

Go here to read the rest:

Eric Schmidt makes the case that AI will be good for society - Recode

Researchers tested the AI by having a pianist play a variety of music -- baroque, classical, ragtime and jazz -- in a live demonstration. The AI then assessed the likely genre in real time, vastly outperforming conventional software hand-coded by humans.

"I think the deep learning system performs better because it's had a dispassionate look at quite a lot of audio material," says Monty Barlow, director of Machine Learning at Cambridge Consultants. "It's found the best way to detect one genre from another without any prejudice or bias. It's strangely more human-like in its capabilities than our programmers were in the classical engineering approach."

Cambridge Consultants says its algorithm could lead to more sophisticated methods of organizing and searching music databases. But, it could also be an important advancement in the medical industry too. It could potentially be used to quickly evaluate a patient's health using sensor waveforms (although Cambridge Consultants hasn't exactly said how this will work). And it's not the only recent machine learning-driven medical breakthrough. Last month, researchers at the University of Nottingham in the UK developed an algorithm that can predict your chances of having a heart attack or stroke with over 70 percent accuracy. In the future, all of our doctors are going to be AI music snobs.

See the original post here:

An AI can recognize musical genres better than humans - Engadget - Engadget

Artificial intelligence, machine learning, and other cognitive disciplines keep finding secure toeholds in risk management, but experts doubt that the industry will see a big bang adoption of the technology.

Were going to see a continual shift, and that is what weve seen for the past few years, said Josh Sutton, vice president and global head of data and artificial intelligence at Sapient.

The technologys presence will make itself apparent across the front-, middle-, and back offices in fashions as different as the organizations themselves.

In the near term, Sutton foresees most of AIs advancement in the middle- and back offices wrapped in the cloak of greaterrobotic process automation adoption.

Josh Sutton, Sapient

RPA effectively hit a wall over the past decade in that it could only handle the automation of activities that were 100% rules-based, said Sutton. The intersection of AI technologies with a lot of legacy automation work has enabled the ability to handle processing of things with a little bit more ambiguity, such as detecting and pro-actively addressing potential illegal trading behavior before the regulators do.

One firm, which Sutton declined to name, is using RAP/AI to allocate its compliance teams investigation resources.

The organizations compliance system couldtag the transactions of a trader who made a series of out of the ordinary trades on a day when the trader received a significant number of external phone calls for further investigation.

The AI could look at the incident and use its common sense to understand it was the traders birthday and that they were reasonable trades, explained Sutton. Im not going to investigate, and Im not going to prioritize that for an investigation. If that continues, Ill raise it back up to the top of the stack and let our team work on it. But initially, I would not waste my teams time on that.

Similarly, the technology also handles the ambiguity of correlation within credit risk when firm wants to determinethe knock-on effects of a company melting down or defaulting.

Thats always been the realm of economists and statisticians to come up with answers to that, he said. Thats an area where I think machine learning can leverage a lot of insight from historical data to come up with improved models.

Although Sutton has not seen a lot of firms do that yet, but he hasnt been involved in that type of project personally, he said.

Where Suttondoesnt see AI making that much of a difference would be in areas that already highly automated like risk management in the front office.

This has been highly automated for a while, whether it is calculating value-at-risk or Comprehensive Capital Analysis and Review reports, said Sutton.

A more disruptive technology likely will be the real-world deployment of quantum computing, he added. What used to take hours of computational processing from our existing systems can be done in seconds with quantum computing. I think that will be a bigger game changer in calculating market risk than AI.

(Visited 1 times, 1 visits today)

Excerpt from:

Wall Street Saunters into AI - Markets Media (press release) (registration) (blog)

University of Nottingham researchers created an AI system that scanned routine medical data to predict which patients would have strokes or heart attacks within 10 years. The AI system beat the standard method of prediction, correctly making calls in 355 more cases than traditional means. Predicting cardiovascular events like strokes and heart attacks is a notoriously challenging task. In fact, the researchers note in their recent paper that around half of all strokes and heart attacks occur in patients who were never identified as being at risk.

The records included a decade of health outcomes, lab data, drug information, hospital records, and demographic information. The team identified the distinguishing characteristics of patients who experienced strokes and heart attacks using 75 percent of the records. They then tested their models against the standard guidelines using the remaining 25 percent of the records. The standard guidelines scored 0.728 out of 1.0, with the latter signifying 100 percent accuracy. The machine models scored between 0.745 to 0.764, with the neural network making 355 more accurate predictions than the standard guidelines, therefore earning the best score. Had those predictions been made in real time, the patients could have been provided with preventative care.

According to lead researcher researcher Stephen Weng, within five years the AI medical tools they are testing in labs will be improving the accuracy of clinicians diagnoses and the prognoses of patients. In practice, Weng envisions busy doctors making the most of their time with AI tools that are essentially masters of pattern recognition. [T]he algorithm can look through the entire patient list, flag this up, and bring this to the attention of the doctor, he said to IEEE Spectrum. This could be done with the patient sitting in front of them during a routine appointment, or in a systematic screen of the entire list. Although there is already clinical decision support software available, none of it uses AI pattern recognition which is at the crux of these more accurate results.

Major regulatory hurdles remain before youll be seeing Dr. AI, however: The key barrier to implementation will be managing privacy and patient confidentiality issues, with computer algorithms trawling through vast amounts of patient data which contain confidential and sensitive medical information, Weng told IEEE Spectrum. Wariness of decision-making capabilities in medical machines on the part of regulators will also be a hurdle for AI technology to clear. For all of these reasons, when, exactly, the tech will be put into practice remains uncertain.

Visit link:

Confirmed: AI Can Predict Heart Attacks and Strokes More Accurately Than Doctors - Futurism