Daily Archives: May 2, 2017

Even the most natural-sounding computerized voiceswhether its Apples Siri or Amazons Alexastill sound like, well, computers. Montreal-based start-up Lyrebird is looking to change that with an artificially intelligent system that learns to mimic a persons voice by analyzing speech recordings and the corresponding text transcripts as well as identifying the relationships between them. Introduced last week, Lyrebirds speech synthesis can generate thousands of sentences per secondsignificantly faster than existing methodsand mimic just about any voice, an advancement that raises ethical questions about how the technology might be used and misused.

The ability to generate natural-sounding speech has long been a core challenge for computer programs that transform text into spoken words. Artificial intelligence (AI) personal assistants such as Siri, Alexa, Microsofts Cortana and the Google Assistant all use text-to-speech software to create a more convenient interface with their users. Those systems work by cobbling together words and phrases from prerecorded files of one particular voice. Switching to a different voicesuch as having Alexa sound like a manrequires a new audio file containing every possible word the device might need to communicate with users.

Lyrebirds system can learn the pronunciations of characters, phonemes and words in any voice by listening to hours of spoken audio. From there it can extrapolate to generate completely new sentences and even add different intonations and emotions. Key to Lyrebirds approach are artificial neural networkswhich use algorithms designed to help them function like a human brainthat rely on deep-learning techniques to transform bits of sound into speech. A neural network takes in data and learns patterns by strengthening connections between layered neuronlike units.

After learning how to generate speech the system can then adapt to any voice based on only a one-minute sample of someones speech. Different voices share a lot of information, says Lyrebird co-founder Alexandre de Brbisson, a PhD student at the Montreal Institute for Learning Algorithms laboratory at the University of Montreal. After having learned several speakers voices, learning a whole new speaker's voice is much faster. Thats why we dont need so much data to learn a completely new voice. More data will still definitely help, yet one minute is enough to capture a lot of the voice DNA.

Lyrebird showcased its system using the voices of U.S. political figures Donald Trump, Barack Obama and Hillary Clinton in a synthesized conversation about the start-up itself. The company plans to sell the system to developers for use in a wide range of applications, including personal AI assistants, audio book narration and speech synthesis for people with disabilities.

Last year Google-owned company DeepMind revealed its own speech-synthesis system, called WaveNet, which learns from listening to hours of raw audio to generate sound waves similar to a human voice. It then can read a text out loud with a humanlike voice. Both Lyrebird and WaveNet use deep learning, but the underlying models are different, de Brbisson says. Lyrebird is significantly faster than WaveNet at generation time, he says. We can generate thousands of sentences in one second, which is crucial for real-time applications. Lyrebird also adds the possibility of copying a voice very fast and is language-agnostic. Scientific American reached out to DeepMind but was told WaveNet team members were not available for comment.

Lyrebirds speed comes with a trade-off, however. Timo Baumann, a researcher who works on speech processing at the Language Technologies Institute at Carnegie Mellon University and is not involved in the start-up, noted Lyrebirds generated voice carries a buzzing noise and a faint but noticeable robotic sheen. Moreover, it does not generate breathing or mouth movement sounds, which are common in natural speaking. Sounds like lip smack and inbreathe are important in conversation. They actually carry meaning and are observable to the listener, Baumann says. These flaws make it possible to distinguish the computer-generated speech from genuine speech, he adds. We still have a few years before technology can get to a point that it could copy a voice convincingly in real-time, he adds.

Still, to untrained ears and unsuspecting minds, an AI-generated audio clip could seem genuine, creating ethical and security concerns about impersonation. Such a technology might also confuse and undermine voice-based verification systems. Another concern is that it could render unusable voice and video recordings used as evidence in court. A technology that can be used to quickly manipulate audio will even call into question the veracity of real-time video in live streams. And in an era of fake news it can only compound existing problems with identifying sources of information. It will probably be still possible to find out when audio has been tampered with, Baumann says, but Im not saying that everybody will check.

Systems equipped with a humanlike voice may also pose less obvious but equally problematic risks. For example, users may trust these systems more than they should, giving out personal information or accepting purchasing advice from a device, treating it like a friend rather than a product that belongs to a company and serves its interests. Compared to text, voice is just much more natural and intimate to us, Baumann says.

Lyrebird acknowledges these concerns and essentially issues a warning in the brief ethics statement on the companys Web site. Lyrebird cautions the public that the software could be used to manipulate audio recordings used as evidence in court or to assume someone elses identity. We hope that everyone will soon be aware that such technology exists and that copying the voice of someone else is possible, according to the site.

Just as people have learned photographs cannot be fully trusted in the age of Photoshop, they may need to get used to the idea that speech can be faked. There is currently no way to prevent the technology from being used to make fraudulent audio, says Bruce Schneier, a security technologist and lecturer in public policy at the Kennedy School of Government at Harvard University. The risk of encountering a fake audio clip has now become the new reality, he says.

View original post here:

New AI Tech Can Mimic Any Voice - Scientific American

At SXSW, we caught up with Diane Bryant, EVP and General Manager of Intel's Data Center Group, who has a very optimistic outlook on tech like AI.

At SXSW Interactive this year, I had the chance to sit down with a number of tech industry execs for my interview series Fast Forward, including Chris Becherer, VP of Product at Pandora; Thad Starner, Professor of Computing at Georgia Tech; and Ron Howard, director and producer of the new NatGeo series Genius.

In this edition of Fast Forward, we're talking with Diane Bryant, EVP and General Manager of Intel's Data Center Group, about artificial intelligence. Read and watch our discussion below.

Dan Costa: Most people think data centers are kind of boring, but you can do incredible things with them. It's also the most profitable division inside of Intel. People think of Intel as a chip company, but the data center business has exploded in recent years, and part of that is driving this AI revolution.

Diane Bryant: Right. Absolutely. The artificial intelligence...discipline was founded in 1956, so we're talking a long time ago, and so it's crazy to now think about how that area has simply exploded and is transforming literally all businesses, and it'll transform the way you and I engage with the world. This has all just happened really in the 2010s, it's just really taken off.

Artificial intelligence can be difficult to define. How do you define it?

Artificial intelligence is a computer system with human-like capabilities, so the ability to think and predict, to learn and predict. That's the definition of it. You say, "Well it sounds pretty simple, so why did it take from 1956 to now?" The issue is in order for a computer system to be able to learn and to demonstrate some of those human attributes of learning and predicting an event, you have to feed it massive, massive amounts of data, and compute on these very very large models. It's gonna take a lot of information for a computer system to draw a conclusion. Historically, there just has not been the affordable compute capacity and storage capacity and network bandwidth capacity to actually process that magnitude of information.

We'll take credit at Intel for Moore's Law, and this ever increasing beat rate of deliverymore and more technology, more capability at lower cost. Because of Moore's Law, you get to the point where you literally can compute on those massive data sets, and actually have a computer system predict an event for you.

There's a processing component to it, and it's also you need the data sets to work with?

Absolutely. There's the compute and storage technology that my organization's responsible for delivering, the fundamental technology into those systems. Then you need the algorithms, those predictive algorithms which is a rapidly evolving space, really a fun space, lots and lots of data scientists needed in order to keep that beat rate going. Then, you're right, there's the software...and service on top of it. It's an entire solution set that needs to come together.

Intel plays in this solution set in multiple places?

We do. At our heart, we're a technology company. As you noted, in the old days we were the PC company, that's our roots, our legacy.

In the old days, we were the PC Magazine.

There you go.

Now we're much more.

Now we're much more, we've all evolved to be bigger and better. We were the PC company, and now with the fact that every company is increasingly dependent upon server storage and network infrastructure, not just to run their business like the good old-fashioned IT organization was there to help businesses run their business. Now [it's] using IT services as revenue opportunities, so cloud services to augment your existing revenue stream.

See, you hear about every company on the planet is going through the digitization of their company. What does it mean to be living in the world of mobile computing and cloud computing? It is definitely evolved. We've moved from the PC company to the data center company and all of the billions of things that connect into the data center.

Every business is becoming a technology business. You may be a retailer, but you need to not just use these tools, but use them to innovate.

Yes it is. At Intel then our responsibility is obviously to continue to serve those enterprises, their IT organizations that have the traditional IT operations, but also our responsibility is to help those companies evolve themselves to digitize, to look for new revenue streams, and new revenue opportunities based on cloud computing. An obvious easy example of artificial intelligence applied to an existing industry is autonomous driving or assisted driving moving to highly autonomous driving moving to fully autonomous driving out in 2035.

That happens because you have a car manufacturer and they have now evolved to where they're going to deploy cloud services to their consumers, so I sell you the car, but then I continue to engage with you and deploy services to that car, whether it's navigation services or entertainment services or maintenance services. Those are additional services that I deploy with you, and then over time you end up being an autonomous car manufacturer, and you're now really a technology company with cars.

In terms of AI, it's early days in a lot of ways, but there are AI systems available today. Can you just describe some of the best examples?

Yeah, there's a lot of them. Assisted driving today, you have cars that will give you a three-point parking solution, or keep you in your lane. That's all AI solutions today. You look at the healthcare industry. The healthcare industry is rapidly deploying AI solutions. We worked with some of the biggest hospitals in China recently on an AI solution that allows the AI system to actually read your medical images to determine whether or not you have a malignant tumor. They have a significant shortage of radiologists to do that manually, when you have a population of 1.3 billion people. Out of demand, out of a true need, they've deployed an AI solution to do that first pass, and they've proven now that the AI solution is actually more accurate than the human-based solution.

You have an opportunity here through AI to unleash a constraint. You're delivering an AI solution either to deliver you something you never had before like assisted driving, or to unleash a physical, fundamental constraint in the system like a shortage of a certain skillset or opportunity.

It seems like a lot of the promise of AI is to break through a lot of those constraints in all sorts of different industries.

In all sorts of industries. We talked this morning with a company called FarmLogs, and it was founded by two kids that grew up on farms and they looked at the world of technology and said, "How come technology hasn't come to our world let alone the farming community?" There's a huge demand, pent-up demand. The world banks will say that by 2050, the world needs to increase food production by 50 percent in order just to serve the population at that time, 9 billion people projected and to do so while the amount of agricultural land total is declining every single year.

It's a huge challenge. You have a huge constraint, a fundamental constraint around food production, there's nothing more basic. You take artificial intelligence, you apply it to the field. Now you can aggregate data about weather and soil content and fertilization and output, yield of that acreage of land, and you can now improve upon the production of the land and help farmers get a bigger return out of their fixed capacity.

It's really something that's happening globally, too. In the US, we have a certain set of problems, certain sets of issues, but when you start to scale this technology out globally, that's where it really starts to make a difference.

Well, it is. You think about China as a still-developing nation, and yet they are the first to embrace next-generation technology. They don't have the legacy in many instances. They're not carrying around the way we used to do things and having to go through that change factor. They're actually bringing up solutions as computer-based AI solutions.

Smart cities is another great example. They're bringing up new cities. They're bringing them up as smart cities. Let's get efficient in the way we deploy solutions to the residents in our area.

Yeah, they're leapfrogging all those intermediate steps.

Exactly. You see it in India as well, as they're building out infrastructure, they're going straight to wireless. They're skipping that whole generation. Once you have wireless connectivity and pervasive connectivity, there's things you can do that you couldn't do with a legacy environment.

A big part of AI as you've mentioned is the ability to predict human behavior. First of all, that's where a lot of utility gets generated, but it's also where a lot users are like, "Is artificial intelligence going to know me better than I know myself? Am I going to be marketed to in ways that I can't control or predict?" Is it a good thing that AIs can predict human behavior so efficiently?

I will obviously say of course it's a good thing. Again, you're stuck with a certain situation, and if you can unleash a given constraint through technology, then I would say it's a good thing. You think back really when the internet became pervasive, late 90s, early 2000s, there was a lot of concern about privacy of data, data privacy was a big big topic. You've noticed now over time that it's become less of a topic.

Every time there is a revolution, and I will say AI is a revolution, just like the industrial revolution and then the digital revolution, and then the information revolution with the internet, this is the next big revolution. Any time you have a big revolution, and rightly so, people stand back and say, "What are all the unintended consequences? Should I be that excited about this?" As you see with each of those waves, nobody's looking back and saying, "Oh we should have never launched that internet thing, that was crazy." But at the time, there was a lot of concern of everyone being connected, everyone having access to data, and you can talk yourself into a situation where you're concerned. But of course there's so many positive benefits of it that just dwarf the concerns that the revolution moves on and people move with it.

Today, you have conversations around AI and what could it mean from a negative perspective, if you have a rogue computer who's unsupervised, who doesn't adhere to the social norms, what will happen? You hear these concerns but it is good to have the conversations, to talk it through, and to then work through those concerns and get a positive end state, because the positive impact of artificial intelligence, the opportunities far outweigh the negatives.

I think there are also individuals who feel a little bit powerless. They feel like AI is something for big companies.

I definitely wouldn't agree with that, as you would anticipate I would say. That's the beauty of the cloud, as we started I said, "Why now?" One of the key technology innovations that has unleashed AI is cloud computing. Just like now, you have your phone and you have all kinds of access to apps and services on your phone, those services are delivered from the cloud. You will have AI services delivered from the cloud and it's completely democratized. Everyone has access.

We were talking today to a company called Picasso, and they take these artificial intelligence to analyze an artistic style of a Matisse or a Monet or a cubist artist, and then take that style and help you develop your own art with that style. It's the merging of the two images through artificial intelligence.

The democratization of AI comes with the reduction in the cost of access. Again, that's back to Moore's Law, and that continuous the decline in the cost while you're increasing the capacity of technology, it becomes democratized.

The other concern people have is automation. We interviewed Vivienne Ming at CES. She's an AI expert and an entrepreneur and she builds AI engines to solve problems for people. The quote she gave me was that if you're doing the same job you were doing a year ago in the same way, she's gonna write an AI engine that will replace you. Are you worried that there's a fundamental risk to jobs when AI really comes to its own?

The risk to a worker of their job being displaced by an AI or a computer system is one of those worries that comes up, absolutely. There will be some jobs certainly that are displaced though continued automation and improving the intelligence of those automated systems. But if you also were to go out and ask any company in the US what their biggest constraint is, they would say workforce. They don't have sufficient trained workforce.

To your point, or to Vivienne's point I guess, your job may not be the same, but we still need you to do a job. It's just a different skilled job. The key I think for all of us in enterprise, is to continue to [train] the workforce as technology continues is deployed into enterprises.

Andrew Ng, who's a data scientist at Baidu, said, "Just like the industrial revolution eliminated much of the physical drudgery for you and me, the artificial intelligence revolution will eliminate much of the mental drudgery for me." We're pulling off the low-hanging fruit of the work that can be, to your point, can be itemized, can be displaced so that we're all free to move up the stack and do the higher-value, higher-meaning work. It does mean a re-skilling.

Microsoft Excel didn't put all accountants out of business; they're doing higher level work?

And much more efficiently. They're more efficient. They can apply their true brain power and skills to solving the gnarly problems instead of the manual calculations. You've made them more efficient, you've made their impact to the world greater.

What are those skills? If you're a 21-year-old coming out of college and getting your first job, what skills do you need? If you're a 50-year-old, your job has been eliminated through automation and you need to re-task and re-skill for the next chapter in your career.

It is a technology-based skillset. We talk about the fact that you still have liberal arts colleges, kids getting liberal arts degrees, but whatever industry they go in, that industry is connected to technology in some way.

It used to be that technology had permeated all industries. Now it's pervasive in all industries. That fundamental knowledge of technology, the application of technology, it's kind of like math or applied math, there's technologists, the people that are really inventing the next generation of AI solution, and then there is applied math, applied technology. You don't have to be the deep technologist, but you've got to be able to operate in an environment where there is applied technology.

I want to get to my closing questions I ask all my guests. What technology trends concern you the most going forward?

I have no fear of technology, are you kidding?

No fears at all?

Maybe I live in the Pollyanna world of everything is bright and rosy in the world of technology, but I've been at Intel for 32 years. I have seen innovation left and right. I am always amazed at what gets invented tomorrow. I live amongst 40,000 engineers who wake up the same way, thinking, "What are we gonna do tomorrow?" Nothing about technology scares me.

Let's flip it and say what are you most optimistic about? What really inspires you?

I am truly inspired by the application of technology and in particular artificial intelligence to healthcare. The healthcare industry has been ripe for disruption and innovation for a very long time, and now we have very tangible solutions, the impact that can be had. In my group in 2015, we kicked off an effort called All In One Day By 2020.

In 2015, we said, by 2020, if you have cancer, your doctor should be able to fully sequence your genome and compare that genome sequence and all the imaging data with data from around the world. Take those results, find the matches. Determine what your disease actually is, what treatments have been applied to that, what the impact of that treatment was. Did the patient survive or not? Through that, deliver to you a personalized treatment plan, and do that all in one day.

There's no reason why that can't be done. We can have a huge impact on something that's as pervasive as cancer. Half of all men and a third of all women will have cancer in their lifetime. I'm sure you know someone. My mom died of cancer. I'm so inspired by the tremendous impact we can have in curing some fundamental diseases through technology and AI.

2020 is only three years away.

I'm not scared.

Just letting you know.

See I'm not scared of technology.

I'm detecting some optimism.

I am very optimistic cause you can see AI solutions getting deployed piecemeal into the healthcare industry as they're awakening. You bring that together, the partnerships we've formed with major cancer research institutes around the world, the environment is ripe for disruption.

In terms of a product, a service, a gadget that you use every day, is there anything you use that every time you use it you're just like, "Wow, this is fantastic. I'm so glad somebody invented this"?

The latest thing I found was my online doggy daycare. You can literally track your dog via GPS as he leaves your house and goes to daycare. A lot of information that I probably don't need...

It's important to know who your dog is hanging out with. You don't want him to fall in with the wrong crowd.

The wrong crowd, and before you know it, you've got a behavioral problem. I just subscribed. I'm like, "Wow. This makes my life so easy." As a busy working mom, my online doggy daycare is fabulous.

For more Fast Forward with Dan Costa, subscribe to the podcast. On iOS, download Apple's Podcasts app, search for "Fast Forward" and subscribe. On Android, download the Stitcher Radio for Podcasts app via Google Play.

Dan Costa is the Editor-in-Chief of PCMag.com and the Senior Vice President of Content for Ziff-Davis. He oversees the editorial operations for PCMag.com, Geek.com, ExtremeTech.com as well as PCMag's network of blogs, including AppScout and SecurityWatch. Dan makes frequent appearances on local, national, and international news programs, including CNN, MSNBC, FOX, ABC, and NBC where he shares his perspective on a variety of technology trends. Dan began working at PC Magazine in 2005 as a senior editor, covering consumer electronics, blogging on Gearlog.com, and serving as... More

See original here:

Fast Forward: Why Intel's Diane Bryant Does Not Fear AI - PCMag

For all the potential scientific and cultural advances brought about by artificial intelligence (AI), it is the humble silicon chip that makes it all possible. Sure, there was the confluence of big data, cloud computing, and the right algorithms, but without the underlying CPUs and GPUs, none of this would have occurred.

With artificial intelligence still in its infancy, there are opportunities for investors to benefit from this paradigm shift that melds science and computing. We don't know who the ultimate winner in AI will be, and there will likely be more than one. What we do know, though, is that chips will be there.

Market intelligence firm Tractica estimates that chipset shipments for AI will grow from 863,000 units in 2016 to 41.2 million units annually by 2025. Two companies are currently positioned to benefit the most from the rapid technological innovations brought on by AI, with one more to watch: NVIDIA Corporation (NASDAQ:NVDA), Intel Corporation (NASDAQ:INTC), and Advanced Micro Devices, Inc. (NASDAQ:AMD).

NVIDIA DGX-1 AI supercomputer in a box. Image source: NVIDIA.

NVIDIA leveraged its position in the AI revolution by having the right tool for the job when AI came calling. Its graphics processing units (GPUs) can handle large numbers of basic mathematical calculations simultaneously. The way these chips processed graphics and AI math calculations was strikingly similar and gave NVIDIA a lead in the space. The computationally intense training of AI systems is still primarily the domain of GPUs, and NVIDIA sits firmly in the lead.

In 2016, NVIDIA grew its revenue 38% over the prior year to a record $6.91 billion and increased earnings per share by 138%. In its most recent quarter, data center revenue, where its AI chip business is housed, more than tripled over the prior-year quarter to 14% of the company's revenue. The stock price mirrored its financial performance with the stock up 230% for 2016.

Intel held its "AI Day: Unleashing the Next Wave" to lay out its vision for AI. Image source: Intel.

Intel has designs on the AI space, and the company has made numerous acquisitions that place it squarely at the intersection of several emerging trends.

Altera developed a field programmable gate array (FPGAs), a chip that can be customized or configured by a customer after manufacturing. Microsoft Corporation (NASDAQ:MSFT) deploys these across its Azure cloud server system for the inferencing stage of AI -- performing a function once the AI system has already been trained -- for applications such as image recognition.

Machine learning start-up Movidius produced systems on a chip (SoC) designed for the computer vision systems used by drones and virtual reality headsets that reduced the power consumption of these data-intensive systems. Deep learning start-up Nervana developed an application specific integrated circuit (ASIC) that is central to Intel's Lake Crest chip (aka, Nervana Engine), which is custom-designed and optimized for AI.

Most recently, Intel acquired machine learning and computer vision company Mobileye N.V. (NYSE:MBLY), which developed cameras used in autonomous driving and software that could detect hazards or obstructions that will be central to self-driving car technology.

Each of the chips in Intel's arsenal serves a distinct function, and these acquisitions give the company a stake in a variety of applications that cut a broad path through the AI marketplace. It would be hard to gage the impact of these latest developments on future performance, but given the breadth of applications, Intel will likely thrive.

AMD introduces Radeon Instinct for AI. Image source: AMD.

AMD is playing catch up in the field and recently released an entire line of chips aimed squarely at deep learning AI applications. While it has produced GPUs for years, it only recently developed processors specifically for AI. AMD has a dedicated following among those looking for a "quality on a budget," and it appears to be pursuing the same strategy in AI. There are also indications that its new line of dedicated chips may seek to match or improve upon existing performance across a broad range of AI applications.

AMD gained market share and rode the coattails of its larger rival with its stock price quadrupling over 2016, while its revenue grew 7% to $4.27 billion. It remains to be seen if AMD can affect a market that is currently dominated by NVIDIA. It may be able to create a niche similar to the one it carved out for itself in the gaming market.

While each of these companies offers a different way to play the AI market, it is important to note that the field is still in its infancy, and no one solution works best for every application. There is still the potential that someone will build a better mousetrap that will make any or all of the choices described above obsolete. Such is the nature of investing in emerging technology -- fortunes are made and lost, sometimes overnight.

Teresa Kersten is an employee of LinkedIn and is a member of The Motley Fool's board of directors. LinkedIn is owned by Microsoft. Danny Vena has the following options: long January 2018 $25 calls on Intel. The Motley Fool owns shares of and recommends Nvidia. The Motley Fool recommends Intel. The Motley Fool has a disclosure policy.

Read the rest here:

3 Top Chip Stocks Benefiting From AI - Motley Fool