Monthly Archives: July 2017

Toyota is the latest Fortune 500 company to launch an AI focused venture capital fund. The initial early-stage fund will deploy $100 million and operate as a subsidiary of theToyota Research Institute.The automaker has strategically positioned itself as an ROI rather than strategic-focused fund meaning that it aims to profit like any other VC firm.

Jim Adler will serve as managing director of the fund. He has been serving as vice president of Toyota Research and comes from a product background. Adler and the rest of the team at Toyota AI Ventures have made three investments to date. These include:

Nauto Developing driverless car technology

SLAMcore Buildingvisual tracking and mapping algorithms

Intuition Robotics Creating a robot companion for older adults

The team says their strongest value add is helping startups think about what business problems are worth solving. Of course, Toyota Research Institute also brings technical expertise to assist the AI fund with diligence and to help startups make improvements to core technology.

Most of the top founders I speak to tell me that they have little issue raising capital and tend to avoid corporate venture when they can. There is a general anxiety in the market that corporates are not genuine when they promise to be ROI rather than strategic investors. Many question whether even small IP and strategic risk warrants corporate involvement, particularly at the volatile seed stage.

We let startups lead these kinds of discussions, Adler said when asked about this tension. Were not here to extract IP from these investments.

Toyota has structured its fund as a separate company rather than an on-balance-sheet entity to minimize conflicts of interest. The firm expects to follow on and lead both seed and Series A deals.

Running effective corporate venture arms is difficult, and its even more difficult when dealing with AI startups. The capital-saturated AI startup ecosystem needs data, genuine corporate customers and advisors with product expertise. There are exactly four trillion corporate venture arms in the world, but shockingly few get this right fingers crossed Toyota knows what theyre getting themselves into.

Follow this link:

Toyota launches venture capital fund targeting artificial intelligence startups - TechCrunch

Tony Stark (Robert Downey Jr.) relies on the JARVIS artificial intelligence to help pilot his Iron Man suit. (Marvel Comics/Paramount Pictures)

Artificial intelligence, machine learning and autonomy are central to the future of American war. In particular, the Pentagon wants todevelop software that can absorb more information from more sources than a human can, analyzeit andeither advise the human how to respond or in high-speed situations like cyber warfare and missile defense act on its own with careful limits.Call it the War Algorithm,the holy grailof a single mathematical equation designed togive the US military near-perfect understanding of what is happening on the battlefield and help its human designers to react more quickly than our adversaries and thus win our wars. Our coverage of this issue attracted the attention ofCapt. Chris Telley, an Army information operations officer studying at the Naval Postgraduate School. In this op-ed, he offers something of a roadmap for the Pentagon to follow as it pursues this highly complex and challenging goal. Read on! The Editor.

If I had an hour to solve a problem Id spend 55 minutes thinking about the problem and five minutes thinking about solutions. Albert Einstein

Artificial intelligence is to be the crown jewel of the Defense Departmentsmuch-discussedThird Offset, the US militarys effort to prepare for the next 20 years. Unfortunately, joint collaborative human-machine battle networks are off to a slow, evenstumbling, start. Recognizing that todays AI is different from the robots that have come before, the Pentagon must seize what may be just a fleeting opportunity to get ahead on the adoption curve. Adapting the military to the coming radical change requires some simultaneous baby steps to learn first and buy second while growing leaders who can wield the tools of the fourth industrial revolution.

First and foremost, the US must be willing to stomach the cost to build cutting-edge systems. AI functions wired into free or discounted Internet services workbecause the companies profit byselling user data; the Pentagonis probablynot eligible for this discount. Also, some of our more stovepiped tactical networks may have difficulty providing the large numbers of training data points, up to 10,000,000 events, needed to teach a learning machine. Military AIs will go to school with crayons untilwe invest significant capital in open architecture data networks. Furthermore, the technicians needed to integrate military AI wont becheap either. According to data from Glassdoor,AI engineersearn a national average of 35 percent more thancybersecurity engineers, whom DoD is already jumping throughhoopsto recruit and those technical skills arent getting any less valuable.

Last year AI went from research concept toengineering application, one CEO said.Another thinks the next 10years may mean the dawn of anAge of Artificial Intelligence. This isnt just hype. In 2013 anOxford studyforecast that 47 percent of total US jobs were susceptible to computerization.Notably, white-collar workers are beginning to be replaced. It now seems that any job which involves routine manipulation of information on a computer is vulnerable to automation. J.P. Morgan is now using AI solutions to slice360,000 man hoursfrom loan reviews eachweek. This year,insurance claimsworkers began to be replaced by IBMs Watson Explorer. The crux of our human failing is that an AI is capable of analyzingintuitive solutionsout of millions of possible results and manipulating those answers far faster than we can. The fastesthuman gamerscan click a keyboard or mouse at a rate of several hundred actions per minute; a computer can do tens of thousands.

Planners DoDs white-collar workers will be replaced before riflemen. They are just as susceptible to automation as their civilian peers. Right now,synthesizing knowledge and producinga creative and flexible array of means to accomplish assigned missions belongs to staff planners. These service members and defense civilians usebasically the same tools PowerPoint, Excel, etc. as does a contemporary office worker. If a robot can buy stocks and turn a profit or satisfactorily answer 20,000,000 helpdesk queries, certainly it can understand the tactical terms and control measure graphics that compose the language of tactics. After all, field manuals and technique publications are just a voluminous trove of and, or, and not logic gates that can be algorithmically diagrammed.

Enemy contact front?Envelop! Need to plan field logistics?Lay thistemplateover semi-permissive terrain! If the product is an Excel workbook or a prefabricated PowerPoint slide, like intelligence preparation of the environment or battlefield calculus, an AI can probably do it better. The robots are coming for us all even the lowly staff officer.

According to Pedro Domingo, author of The Master Algorithm, the best way to not lose your job to a robot is to automate it yourself. The key to effectively and efficiently on-boarding these technologies, as well as the multi-domain battles they will effect, is human capital. We need a bench of service members and government civilians who at least understand the lexicon and how to ask the right questions of the application interface. These leaders will provide adoption capacity for eventually fielding unilaterally developed defense systems that will form the core of the Third Offset. They help us fight on new, cognitive, attack surfaces; Microsofts @TayTweets chatbot was hacked, not with code, but by Internet trolls slyly teaching it bad behaviors. Just as the Navy trains officers to use celestial navigation while still fighting with GPS, DoD needs leaders who can spar in both the twentieth and twenty-first centuries to enable graceful system degradation.

Overall, AI will be in everything but will not be everything, so the Department must create a career path for these people without creating acareer field. The machines will eventually write their own code so we need thinkers to operationalize automation rather than build software. Those skills can be acquired through intermixing funded massive openonline courses,broadening seminarswith academia, andtraining with industrytenures into standard professional timelines. The US is behind in computer science curriculum; if the DoD is to use AI to lighten the cognitive load by 2021, as the Armys Robotic and Autonomous Systems Strategy demands, they, and the rest of DoD, will need to nurture and retain people with skills in robotics, computational math, and computational art.These programs need selection criteria and retention incentives toproduce at least one AI literate leader for every battalion level command on that four-year timeline. This may seem fast, but leading AI experts expected a machine to beat humans at the game Go in2027;it happenedthis year.

Since the AI market space is accelerating quickly, there are many possibilities for dual-use applications for the Defense Department. Though the military, most notably DARPA, has dabbledwith AI in things like thecyberandself-driving cargrand challenges;fielding a variety of functional technologic solutions will provide proven ground before attempting unilateral projects.

There are many promising areas that would help defense planners get their toes in the water. The first is information operations.Predictiveandprogrammaticmarketing are incredibly lucrative algorithmically powered tools and they are already in use. Combined with AI systems forjournalistic contentcreation, perhaps DoD can overcome ahistorically slowinfluence apparatus to beat state and non-state adversary propaganda. (Editors note: We are VERY uneasy with this idea for moral and more provincial reasons.) Can Google Maps, or its competitors, tell us where traffic isnt, compared to where it was yesterday as a blend of HUMINT/SIGINT to identify roadside bombs (IEDs)? Similar questions should be asked of emerging applications to compete with humans in the strategy game StarCraft, to help combined arms planning at the tactical level. The tools being built to examine cancer genomes could also be developed to model the cell mutations of extremist networks.

Small, short timeline endeavors like Project Maven, recently created touse machine learning for wading through intelligence data, must provide the network integration experience needed for building larger programs of record. Many small successes will certainly be needed to garner senior leader buy-in if decisive AI tools are to survive the Valley of Death between lab experiments and the transition to a program of record.

Fortunately, the AImarket spaceis still coalescing. Unfortunately, it is an exponential technology so every success or failure is amplified by an order of magnitude. So far, Deputy Defense Secretary Bob Work wants$12 billion to $15 billionin 2017 for programs aimed at human-machine collaboration and combat teamingand has received11 recommendationsfrom the Defense Departments Innovation Advisory board toget started.If even half of those dollars go to AI research then the DoD will have matched theventure capitalspent last year on relevant startups. However, our adversaries will seek to gain advantage. China has already spent billions on AI research programs and they have state-owned investor companies, like ZGC Capitol, residing in Santa Clara, Calif.; their military leaders are aiming toward the leading edge of a military revolution of intelligentization. Its also worth noting that many resources, like Googles TensorFlow, are freely available online for whomever decides to use the technology.

So, the time is now for Artificial Intelligence; strategic surprise featuring things like data driven behavior change or A.I. modulated denial of the electromagnetic spectrum will pose difficult challenges from which to recover. If we are to ride the disruptive wave of what some call the Great Restructuring, existing AI applications should be re-purposed before attempting defense-only machine learning systems. Also, developing a cadre of AI-savvy leaders is essential for rapid application integration, as well as for planning to handle graceful system degradation. The right AI investment, in understanding, strategy, and leaders, should be our starting block for a race that will surely reshape thecharacter of warin ways we can only begin to imagine.

Capt. Chris Telley is an Army information operations officer assigned to the Naval Postgraduate School. He commanded in Afghanistan and served in Iraq as a United States Marine. He tweets at @chris_telleyThese are the opinions of the author and do not reflect the position of the Army or the United States Government.

See original here:

Info Ops Officer Offers Artificial Intelligence Roadmap - Breaking Defense

How successful Intels efforts prove to be will be crucial not only for the company but also for the long-term future of the computer chip industry.

Were seeing a lot more competition in the data-center market than weve seen in a long time, said Linley Gwennap, a semiconductor expert who leads a technology research firm in Mountain View, Calif.

Intel has long dominated the business for central processing chips that control industry-standard servers in data centers. Matthew Eastwood, an analyst at IDC, said the company controlled about 96 percent of such chips.

But others are making inroads into advanced data centers. Nvidia, a chip maker in Santa Clara, Calif., does not make Intel-style central processors. But its graphics-processing chips, used by gamers in turbocharged personal computers, have proved well suited for A.I. tasks. Nvidias data-center business is taking off, with the companys sales surging and its stock price nearly tripling in the last year.

Big Intel customers like Google, Microsoft and Amazon are also working on chip designs. AMD and ARM, which make central processing chips like Intel, are edging into the data-center market, too. IBM made its Power chip technology open source a few years ago, and Google and others are designing prototypes.

To counter some of these trends, Intel is expected on Tuesday to provide details about the performance and uses of its new chips and its plans for the future. The company is set to formally introduce the next generation of its Xeon data-center microprocessors, code-named Skylake. And there will be a range of Xeon offerings with different numbers of processing cores, speeds, amounts of attached memory, and prices.

Yet analysts said that would represent progress along Intels current path rather than an embrace of new models of computing.

Stacy Rasgon, a semiconductor analyst at Bernstein Research, said, Theyre late to artificial intelligence.

Intel disputes that characterization, saying that artificial intelligence is an emerging technology in which the company is making major investments. In a blog post last fall, Brian Krzanich, Intels chief executive, wrote that it was uniquely capable of enabling and accelerating the promise of A.I.

Intel has been working in several ways to respond to the competition in data-center chips. The company acquired Nervana Systems, an artificial intelligence start-up, for more than $400 million last year. In March, Intel created an A.I. group, headed by Naveen G. Rao, a founder and former chief executive of Nervana.

The Nervana technology, Intel has said, is being folded into its product road map. A chip code-named Lake Crest is being tested and will be available to some customers this year.

Lake Crest is tailored for A.I. programs called neural networks, which learn specific tasks by analyzing huge amounts of data. Feed millions of cat photos into a neural network and it can learn to recognize a cat and later pick out cats by color and breed. The principle is the same for speech recognition and language translation.

Intel has also said it is working to integrate Nervana technology into a future Xeon processor, code-named Knights Crest.

Intels challenge, analysts said, is a classic one of adapting an extraordinarily successful business to a fundamental shift in the marketplace.

As the dominant data-center chip maker, used by a wide array of customers with different needs, Intel has loaded more capabilities into its central processors. It has been an immensely profitable strategy: Intel had net income of $10.3 billion last year on revenue of $59.4 billion.

Yet key customers increasingly want computing designs that parcel out work to a collection of specialized chips rather than have that work flow through the central processor. A central processor can be thought of as part brain, doing the logic processing, and part traffic cop, orchestrating the flow of data through the computer.

The outlying, specialized chips are known in the industry as accelerators. They can do certain things, like data-driven A.I. tasks, faster than a central processor. Accelerators include graphics processors, application-specific integrated circuits (ASICs) and field-programmable gate arrays (F.P.G.A.s).

A more diverse set of chips does not mean the need for Intels central processor disappears. The processor just does less of the work, becoming more of a traffic cop and less of a brain. If this happens, Intels business becomes less profitable.

Intel is not standing still. In 2015, it paid $16.7 billion for Altera, a maker of field-programmable gate arrays, which make chips more flexible because they can be repeatedly reprogrammed with software.

Mr. Gwennap, the independent analyst, said, Intel has a very good read on data centers and what those customers want.

Still, the question remains whether knowing what the customers want translates into giving them what they want, if that path presents a threat to Intels business model and profit margins.

Follow Steve Lohr on Twitter @SteveLohr.

A version of this article appears in print on July 11, 2017, on Page B5 of the New York edition with the headline: Intel Protects Its Lead While Pivoting to A.I.

Excerpt from:

Intel, While Pivoting to Artificial Intelligence, Tries to Protect Lead - New York Times

I like Gartners concept of the technology hype cycle. It assumes that expectations of new technologies quickly ramp to an inflated peak, drop into a trough of disillusionment, then gradually ascend a slope of enlightenment until they plateau. Of course, not all technologies complete the cycle or transition through the stages at the same pace.

Artificial intelligence (AI) has arguably been in the trough for 60 years. I am thinking of Kubricks HAL and Roddenberrys computer that naturally interact with humans. Thats a long trough, and despite popular opinion, the end is nowhere in sight.

Theres so much excitement and specialized research taking place that AI has fragmented into several camps such as heuristic programming for game-playing AI, natural language processing for conversational AI, and machine learning for statistical problems. The hype is building again, and just about every major tech company and countless startups are racing toward another inflated peak and subsequent trough.

The reason expectations are so high is because of breakthroughs in three broad categories: compute, data and algorithms. The compute innovations refer to general cloud services and specific improvements in processing arrays and graphics processing units (GPUs).

The availability of huge data sets has also been important for machine learning. Large labeled and annotated data sets have enabled progress in computer vision, natural language and speech recognition. There are numerous public data sets available, plus many of the larger firms are also using their own private data.

The third ingredient is advanced algorithms that with compute power and data provide responses or predictions. For example, algorithms are used to recommend movies to watch, stocks to trade and updates to include on a timeline. The concept is as old as computing itself, but suddenly vastly improved.

Or is it? While a computer beat a human chess champion 21 years ago, it wasnt until two months ago that a different computer beat a human champion at Go. There was an impressive milestone on Jeopardy in 2011and more recently a breakthrough regarding Ms. Pac-Man.

AI will definitely change the world, but just dont hold your breath, at least not regarding general purpose AI. Specialized AI, such as self-driving cars, is progressing quickly. The general AI stuff is almost useless.

I have yet to find any general AI solution to be helpful. For example, Google Assistant often suggests to me the best time to leave for the airport. Its invariably wrong. It largely bases its recommendation on my current location and traffic conditions. My personal algorithm for determining the best time to leave for the airport involves relatively big data. I consider variables such as how I intend to get there (car, bus or shuttle). If by car, then I factor in where I intend to park. Then there's gate and concourse information; whether I have PRE on my boarding pass; and whether I intend to eat at the airport before departure.

Usually, I take the bus to the airport and query Google about the bus schedule. The famed Google Assistant cant recognize that pattern. Telling me when to leave to catch the airport bus could be more helpful.

But having the data to answer the question isnt Googles problem. The difficulty lies in understanding the question. Emmanuel Mogenet, head of Google Research Europe, recently highlighted the limitations of natural language processing with a similar example. Google Assistant cant answer will it be dark when I get home? Let me put that in context. Google cant answer this question even when it knows where the user is, where the user lives, and when the sun sets at that location.

This is not a question that has an answer Google can look up. It needs to pull all this information together, and doing that requires truly understanding the relationship between the question and the data. Thats a hard puzzle to solve. Now consider that Google Assistant is six times more likely to correctly answer a question than Amazons Alexa.

Alexa now boasts more than 15,000 skills. These skills are largely simple web queries. The AI part is using speech instead of a keyboard.

AI has a ways to go, but thats not even the whole problem. As with my airport example above, AI works best when it has access to contextual data. That often means exposing personal and confidential data to the service, which is a practice riddled with concerns and liabilities. Its not as if security breaches are rare.

Theres also the little issue that AI is very hard to test. Developing self-driving cars requires driving cars millions of miles. That just doesnt scale, so we keep discovering gaps with each new application. Even self-driving car behavior can be surprising. Volvo recently found that its self-driving cars cannot recognize kangaroos.Oops.

I think its important to reset expectations about AI. Its fantastic that some people find Siri, Google Assistant and Alexa helpful sometimes. We should briefly celebrate the tremendous progress in kitchen timer technologyand then get back to work.

Read the original post:

AI heading back to the trough - Network World

This article was originally published on The Conversation. Read the original article.

How do you stop a robot from hurting people? Many existing robots, such as those assembling cars in factories, shut down immediately when a human comes near. But this quick fix wouldnt work for something like a self-driving car that might have to move to avoid a collision, or a care robot that might need to catch an old person if they fall. With robots set to become our servants, companions and co-workers, we need to deal with the increasingly complex situations this will create and the ethical and safety questions this will raise.

Science fiction already envisioned this problem and has suggested various potential solutions. The most famous was author Isaac Asimovs Three Laws of Robotics, which are designed to prevent robots harming humans. But since 2005my colleagues and I at the University of Hertfordshire have been working on an idea that could be an alternative.

Daily Emails and Alerts - Get the best of Newsweek delivered to your inbox

Instead of laws to restrict robot behavior, we think robots should be empowered to maximize the possible ways they can act so they can pick the best solution for any given scenario. As we describe in a new paper in Frontiers, this principle could form the basis of a new set of universal guidelines for robots to keep humans as safe as possible.

Asimovs Three Laws are as follows:

While these laws sound plausible, numerous arguments have demonstrated why they are inadequate. Asimovs own stories are arguably a deconstruction of the laws, showing how they repeatedly fail in different situations. Most attempts to draft new guidelines follow a similar principle to create safe, compliant and robust robots.

One problem with any explicitly formulated robot guidelines is the need to translate them into a format that robots can work with. Understanding the full range of human language and the experience it represents is a very hard job for a robot. Broad behavioral goals, such as preventing harm to humans or protecting a robots existence, can mean different things in different contexts. Sticking to the rules might end up leaving a robot helpless to act as its creators might hope.

Our alternative concept, empowerment, stands for the opposite of helplessness. Being empowered means having the ability to affect a situation and being aware that you can. We have been developing ways to translate this social concept into a quantifiable and operational technical language. This would endow robots with the drive to keep their options open and act in a way that increases their influence on the world.

When we tried simulating how robots would use the empowerment principle in various scenarios, we found they would often act in surprisingly natural ways. It typically only requires them to model how the real world works but doesnt need any specialised artificial intelligence programming designed to deal with the particular scenario.

But to keep people safe, the robots need to try to maintain or improve human empowerment as well as their own. This essentially means being protective and supportive. Opening a locked door for someone would increase their empowerment. Restraining them would result in a short-term loss of empowerment. And significantly hurting them could remove their empowerment altogether. At the same time, the robot has to try to maintain its own empowerment, for example by ensuring it has enough power to operate and it does not get stuck or damaged.

Using this general principle rather than predefined rules of behavior would allow the robot to take account of the context and evaluate scenarios no one has previously envisaged. For example, instead of always following the rule dont push humans, a robot would generally avoid pushing them but still be able to push them out of the way of a falling object. The human might still be harmed but less so than if the robot didnt push them.

In the film I, Robot, based on several Asimov stories, robots create an oppressive state that is supposed to minimize the overall harm to humans by keeping them confined and protected. But our principle would avoid such a scenario because it would mean a loss of human empowerment.

While empowerment provides a new way of thinking about safe robot behavior, we still have much work to do on scaling up its efficiency so it can easily be deployed on any robot and translate to good and safe behaviour in all respects. This poses a very difficult challenge. But we firmly believe empowerment can lead us towards a practical solution to the ongoing and highly debated problem of how to rein in robots behavior, and how to keep robotsin the most naive senseethical.

Christoph Salgeis aMarie Curie Global Fellow at theUniversity of Hertfordshire.

Read more:

Artificial Intelligence and the Robot Apocalypse: Why We Need New Rules to Keep Humans Safe - Newsweek