Daily Archives: August 8, 2017

In an effort to combat bioterrorism, scientists in South Korea have trained artificial intelligence to speedily spot anthrax. The new technique is not 100 percent accurate yet, but its orders of magnitude faster than our current testing methods. And it could revolutionize how we screen mysterious white powders for the deadly bioweapon.

Researchers at the Korea Advanced Institute of Science and Technology combined a detailed imaging technique called holographic microscopy with artificial intelligence. The algorithm they created can analyze images of bacterial spores to identify whether theyre anthrax in less than a second. Its accurate about 96 percent of the time, according to a paper published last week in the journal Science Advances.

Anthrax can kill quickly, if left untreated

Anthrax is an infection caused by the bacteria Bacillus anthracis, which lives in soil. (Both the infection and the bacteria are often referred to as anthrax.) People can accidentally get anthrax infections when they handle the skin or meat of infected animals. But anthrax can also be a dangerous bioweapon: in 2001, anthrax spores sent in the mail infected 22 people and killed five of them.

Once the spores enter the body, they germinate and multiply, causing a flu-like illness that poisons the blood. At least 85 percent of people infected by inhaling the spores die if left untreated, sometimes within just one to two days after symptoms appear. (Anthrax infections of the skin, by contrast, tend to be less fatal.) For people especially at risk of contracting anthrax, like lab workers or people who work with animal hair, theres a vaccine. For the rest of us, there are antibiotics but these work best when theyre started as soon as possible after exposure.

Its important to detect anthrax fast

So its important to detect anthrax fast. Right now, one of the most common methods is to analyze the genetic material of the spores or, once someone is infected, of the bacteria found in infected tissue. But that typically requires giving the spores a little time to multiply in order to yield enough genetic material to analyze. Its still going to take the better part of a day with the most rapid approaches to get a result, says bacteriologist George Stewart at the University of Missouri, who has also developed an anthrax detector and was not involved in this study.

In search of a quicker screening technique, the studys lead author, physicist YongKeun Park, teamed up with South Koreas Agency for Defense Development. The goal is to be prepared in case North Korea is developing anthrax as a bioweapon, he says.

Park turned to an imaging technique called holographic microscopy: unlike conventional microscopes, which can only capture the intensity of the light scattering off an object, a holographic microscope can also capture the direction that light is traveling. Since the structure and makeup of a cell can change how light bounces off of it, the researchers suspected that the holographic microscope might capture key, but subtle, differences between spores produced by anthrax and those produced by closely related, but less toxic species.

The AI could ID the anthrax spores within seconds

Park and his team then trained a deep learning algorithm to spot these key differences in more than 400 individual spores from five different species of bacteria. One species was Bacillus anthracis, which causes anthrax, and four were closely related doppelgngers. The researchers didnt tell the neural network exactly how to spot the different species the AI figured that out on its own. After some training, it could distinguish the anthrax spores from the non-anthrax doppelgnger species about 96 percent of the time.

The technique isnt perfect, and as a tool intended to detect bioweapons, it has to be. The drawback is that the accuracy is lower than conventional methods, Park says. There are also multiple strains of each of the bacteria species analyzed but the machine was trained on only one strain per species. Subtle differences between the strains might be able to throw off the algorithm, Stewart says. Still, the new technique is so rapid that it could come in handy. It doesnt require culturing organisms, it doesnt require extracting DNA, it doesnt require much of anything other than being able to visualize the spores themselves, Stewart says.

It could enhance our preparation for this kind of biological threat.

Next, Park wants to feed the neural network more spore images, in order to boost accuracy. In the meantime, the method could be used as a pre-screening tool to rapidly determine whether a white powder that people have been exposed to is anthrax, and if they should start antibiotics. A slower, more accurate method could then confirm the results.

This paper will not change everything, Park says, but its one step toward a method that can quickly detect anthrax. It could enhance our preparation for this kind of biological threat.

Go here to see the original:

AI can now detect anthrax which could help the fight against ... - The Verge

As an artificial intelligence researcher, I often come across the idea that many people are afraid of what AI might bring. Its perhaps unsurprising, given both history and the entertainment industry, that we might be afraid of a cybernetic takeover that forces us to live locked away, Matrix-like, as some sort of human battery.

And yet it is hard for me to look up from the evolutionary computer models I use to develop AI, to think about how the innocent virtual creatures on my screen might become the monsters of the future. Might I become the destroyer of worlds, as Oppenheimer lamented after spearheading the construction of the first nuclear bomb?

I would take the fame, I suppose, but perhaps the critics are right. Maybe I shouldnt avoid asking: As an AI expert, what do I fear about artificial intelligence?

The HAL 9000 computer, dreamed up by science fiction author Arthur C. Clarke and brought to life by movie director Stanley Kubrick in 2001: A Space Odyssey, is a good example of a system that fails because of unintended consequences. In many complex systemsthe RMS Titanic, NASAs space shuttle, the Chernobyl nuclear power plantengineers layer many different components together. The designers may have known well how each element worked individually, but didnt know enough about how they all worked together.

That resulted in systems that could never be completely understood, and could fail in unpredictable ways. In each disastersinking a ship, blowing up two shuttles, and spreading radioactive contamination across Europe and Asiaa set of relatively small failures combined together to create a catastrophe.

I can see how we could fall into the same trap in AI research. We look at the latest research from cognitive science, translate that into an algorithm, and add it to an existing system. We try to engineer AI without understanding intelligence or cognition first.

Systems like IBMs Watson and Googles Alpha equip artificial neural networks with enormous computing power, and accomplish impressive feats. But if these machines make mistakes, they lose on Jeopardy! or dont defeat a Go master; these are not world-changing consequences. Indeed, the worst that might happen to a regular person as a result is losing some money betting on their success.

But as AI designs get even more complex and computer processors even faster, their skills will improve. That will lead us to give them more responsibility, even as the risk of unintended consequences rises. We know that to err is human, so it is likely impossible for us to create a truly safe system.

Im not very concerned about unintended consequences in the types of AI I am developing, using an approach called neuroevolution. I create virtual environments and evolve digital creatures and their brains to solve increasingly complex tasks. The creatures performance is evaluated; those that perform the best are selected to reproduce, making the next generation. Over many generations these machine-creatures evolve cognitive abilities.

Right now we are taking baby steps to evolve machines that can do simple navigation tasks, make simple decisions, or remember a couple of bits. But soon we will evolve machines that can execute more complex tasks and have much better general intelligence. Ultimately we hope to create human-level intelligence.

Along the way, we will find and eliminate errors and problems through the process of evolution. With each generation, the machines get better at handling the errors that occurred in previous generations. That increases the chances that well find unintended consequences in simulation, which can be eliminated before they ever enter the real world.

Another possibility thats farther down the line is using evolution to influence the ethics of artificial intelligence systems. Its likely that human ethics and morals, such as trustworthiness and altruism, are a result of our evolutionand factor in its continuation. We could set up our virtual environments to give evolutionary advantages to machines that demonstrate kindness, honesty, and empathy. This might be a way to ensure that we develop more obedient servants or trustworthy companions and fewer ruthless killer robots.

While neuroevolution might reduce the likelihood of unintended consequences, it doesnt prevent misuse. But that is a moral question, not a scientific one. As a scientist, I must follow my obligation to the truth, reporting what I find in my experiments, whether I like the results or not. My focus is not on determining whether I like or approve of something; it matters only that I can unveil it.

Being a scientist doesnt absolve me of my humanity, though. I must, at some level, reconnect with my hopes and fears. As a moral and political being, I have to consider the potential implications of my work and its potential effects on society.

As researchers, and as a society, we have not yet come up with a clear idea of what we want AI to do or become. In part, of course, this is because we dont yet know what its capable of. But we do need to decide what the desired outcome of advanced AI is.

One big area people are paying attention to is employment. Robots are already doing physical work like welding car parts together. One day soon they may also do cognitive tasks we once thought were uniquely human. Self-driving cars could replace taxi drivers; self-flying planes could replace pilots.

Instead of getting medical aid in an emergency room staffed by potentially overtired doctors, patients could get an examination and diagnosis from an expert system with instant access to all medical knowledge ever collectedand get surgery performed by a tireless robot with a perfectly steady hand. Legal advice could come from an all-knowing legal database; investment advice could come from a market-prediction system.

Perhaps one day, all human jobs will be done by machines. Even my own job could be done faster, by a large number of machines tirelessly researching how to make even smarter machines.

In our current society, automation pushes people out of jobs, making the people who own the machines richer and everyone else poorer. That is not a scientific issue; it is a political and socioeconomic problem that we as a society must solve. My research will not change that, though my political selftogether with the rest of humanitymay be able to create circumstances in which AI becomes broadly beneficial instead of increasing the discrepancy between the one percent and the rest of us.

There is one last fear, embodied by HAL 9000, the Terminator, and any number of other fictional superintelligences: If AI keeps improving until it surpasses human intelligence, will a superintelligence system (or more than one of them) find it no longer needs humans? How will we justify our existence in the face of a superintelligence that can do things humans could never do? Can we avoid being wiped off the face of the Earth by machines we helped create?

The key question in this scenario is: Why should a superintelligence keep us around?

I would argue that I am a good person who might have even helped to bring about the superintelligence itself. I would appeal to the compassion and empathy that the superintelligence has to keep me, a compassionate and empathetic person, alive. I would also argue that diversity has a value all in itself, and that the universe is so ridiculously large that humankinds existence in it probably doesnt matter at all.

But I do not speak for all humankind, and I find it hard to make a compelling argument for all of us. When I take a sharp look at us all together, there is a lot wrong: We hate each other. We wage war on each other. We do not distribute food, knowledge, or medical aid equally. We pollute the planet. There are many good things in the world, but all the bad weakens our argument for being allowed to exist.

Fortunately, we need not justify our existence quite yet. We have some timesomewhere between 50 and 250 years, depending on how fast AI develops. As a species we can come together and come up with a good answer for why a superintelligence shouldnt just wipe us out. But that will be hard: Saying we embrace diversity and actually doing it are two different thingsas are saying we want to save the planet and successfully doing so.

We all, individually and as a society, need to prepare for that nightmare scenario, using the time we have left to demonstrate why our creations should let us continue to exist. Or we can decide to believe that it will never happen, and stop worrying altogether. But regardless of the physical threats superintelligences may present, they also pose a political and economic danger. If we dont find a way to distribute our wealth better, we will have fueled capitalism with artificial intelligence laborers serving only very few who possess all the means of production.

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

Read more:

An artificial intelligence researcher reveals his greatest fears about the future of AI - Quartz

Since its founding in 2000 as an alternative provider of legal services to large corporations, Axiom has grown to have more than 1,500 lawyers and 2,000 employees across three continents, serving half the Fortune 100. With a reputation for innovation, it describes itself as a provider of tech-enabled legal services.

Given that description, it would seem inevitable that Axiom would bring artificial intelligence into the mix of the services it offers. Now it has. This week, it announced the launch of AxiomAI, a program that aims to leverage AI to improve the efficiency and quality of Axioms contract work.

AxiomAI has two components, Paul Carr, senior executive overseeing Axioms AI efforts, told me during an interview Friday. One is research, development and testing of AI tools for legal services. The other is deploying AI tools within Axioms standard workflows as testing proves them ready.

The first such deployment will come later this month, as Axiom embeds Kira Systems machine-learning contract analysis technology into its M&A diligence and integration offering. In an M&A deal, which can require review of thousands of corporate contracts, Kira automates the process of identifying key provisions such as change of control and assignment.

In the context of M&As, the AI will be invisible to our clients, Carr said. They know they have to understand the risks that may be in those agreements. They need someone to sort that out which agreements apply, whats in them in a very accurate way. And they need actionable recommendations, very specific recommendations. Thats what we deliver today, but well deliver it better and faster using AI behind the scenes.

Beyond this immediate deployment, AxiomAI will encompass a program of ongoing research and testing of AIs applicability to the delivery of legal services. In fact, it turns out that Axiom has quietly been performing this research for four years, including partnering with leading experts and vendors in the field of machine learning.

Weve been watching this space for a while, Carr said. Weve been testing really actively, running proofs of concept, of various AI tools over the last four years. At a fundamental level, we do believe that for a lot of legal work, AI will have really important applications and will change legal workflows into the future.

The focus of Axioms AI research is, as Carr put it, all things contracting, from creating single contracts to applying analytics to collections of contracts. And the type of AI on which it is focused is machine learning. We think the area that is most interesting is machine learning and, specifically, the whole area of deep learning within machine learning.

In the case of Kira, Axioms testing had demonstrated that the product was ready for deployment. We felt that the maturity of the technology which is really code for the ability of the technology to perform at a level that makes economic sense was such that it makes sense to move it, in a sense, from the lab to production, in a business-as-usual context.

Going forward, Axiom plans to keep testing other AI tools in partnership with leading practitioners in the field. A key benefit Axiom brings to the equation is an enormous collection of contractual data that can be used to train the AI technology.

We analyze over 10 million pieces of contractual information every year, Carr said. We have a very powerful data set that we plan to use to train AI technology. What we will certainly do is train and improve that technology with our training data.

The training that is performed using Axioms data will remain proprietary, and Carr believes that will add greater value for Axioms customers in the use of these AI tools.

The roadmap for Axioms research has two tracks, Carr said. One is to explore how to go deeper and further into the M&A offering its launching this month, in order to train AI tools to do even more of the work. The second is to consider the other use cases to focus on next.

One use case under consideration involves regulatory remediation for banks. Another would assist pharmaceutical companies in the negotiation and execution of clinical trial agreements.

Carr came to Axiom in 2008 from American Express, where he had run its International Insurance Services division and was its global head of strategy. He started his career working on systems integration design. He believes that technological integration takes much longer to achieve than technological innovation.

You need to put in place the surrounding capabilities that allow you to take advantage of that technology and, not immaterially, you need to go through the process of change management and behavioral change, he said. In the legal industry, thats a big deal. Theres a lot that has to happen for technical innovations to be consumed.

Driving that adoption curve is the heart of Axioms business, Carr suggests. The best way to do that, the company believes, is to combine people, process and technology in ways that allow the value of the technology to be realized. That is what Axiom now plans to do for AI.

AI today is like the internet in the late 90s, Carr said. I have no doubt that in a couple of decades, AI will be embedded in everything that impacts corporate America. But how it unfolds and takes shape is the stage were in now.

Robert Ambrogiis aMassachusetts lawyerand journalist who has been covering legal technology and the web for more than 20 years, primarily through his blogLawSites.com. Former editor-in-chief of several legal newspapers, he is a fellow of theCollege of Law Practice Managementand an inauguralFastcase 50honoree. He can be reached by email atambrogi@gmail.com, and you can follow him onTwitter(@BobAmbrogi).

Link:

Legal Services Innovator Axiom Unveils Its First AI Offering - Above the Law

Courtesy of SK Telecom.

SK Telecom has launched an outdoor version of its AI speaker in South Korea.

The NUGU Mini is a portable version of NUGU, the mobile carrier's AI speaker for the home that has sold over 150,000 units since launching last year.

NUGU, the name of both the speaker and the AI platform, uses a cloud-based deep-learning framework that increases its speech recognition accuracy over time.

The mobile carrier said over 130 million conversations have been stored in its cloud server. Voice recognition rate for both children and regional Korean dialects significantly increased since launching, it said.

NUGU Mini weighs 219 grams -- almost a fifth of NUGU's 1,030 grams -- and can connect to external speakers and a battery that lasts over four hours.

Users can turn on Melon, Korea's number one music streamer, use smart home features offered by SK Telecom, and control their set-top boxes.

They can also order food, manage schedules, check for weather and traffic information, enquire about currency exchange rates, and register to visit local bank branches.

NUGU Mini will cost 99,000 won ($87) but SK Telecom will offer it for 49,900 won for three months when sales begin August 11.

AI speakers and speech recognition have been a big trend in the Korean tech scene this year. Kakao, which owns the country's biggest chat app KakaoTalk, launched the Kakao Mini, its own AI speaker, and is also cooperating with compatriot car maker Hyundai for in-car speech recognition.

South Korea's largest search giant Naver and its chat app subsidiary Line also partnered with Qualcomm to make AI-based Internet of Things devices.

Read more:

SK Telecom launches portable version of AI speaker - ZDNet

The artificial intelligence oftodayhas almost nothing in common with the AI of science fiction.In Star Wars, Star Trek and Battlestar Galactica, were introduced to robots who behave like we do they are aware of their surroundings, understand the context of their surroundings and can move around and interact with people just as I can with you. These characters and scenarios are postulated by writers and filmmakers as entertainment, and while one day humanity will inevitably develop an AI like this, it wont happen in the lifetime of anyone reading this article.

Because we can rapidly feed vast amounts of data to them, machines appear to be learning and mimicking us, but in fact they are still at the mercy of the algorithms we provide. The way for us to think of modern artificial intelligence is to understand two concepts:

To illustrate this in grossly simplified terms, imagine a computer system in an autonomous car. Data comes from cameras placed around the vehicle, from road signs, from pictures that can be identified as hazards and so on. Rules are then written for the computer system to learn about all the data points and make calculations based on the rules of the road. The successful result is the vehicle driving from point A to B without making mistakes (hopefully).

The important thing to understand is that these systems dont think like you and me.People are ridiculously good at pattern recognition, even to the point where we prefer forcing ourselves to see patterns when there are none.We use this skill to ingest less information and make quick decisions about what to do.

Computers have no such luxury; they have to ingest everything, and if youll forgive the pun, they cant think outside the box. If a modern AI were to be programmed to understand a room (or any other volume) it would have to measure all of it.

Think of the little Roomba robot that can automatically vacuum your house.It runs randomly around until it hits every part of your room.An AI would do this (very fast) and then would be able to know how big the room is.A person could just open the door, glance at the room and say (based on prior experience), Oh, its about 20 ft. long and 12 ft. wide. Theyd be wrong, but it would be close enough.

Over the past two decades, weve delved into data science and developed vast analytical capabilities.Data is put into systems, people look it, manipulate it, identify trends and make decisions based on it.

Broadly speaking, any job like this can be automated.Computer systems are programmed with machine learning algorithms and continuously learn to look at more data more quickly than any human would be able to.Any rule or pattern that a person is looking for, a computer can be programmed to understand and will be more effective than a person at executing.

We see examples of this while running digital advertising campaigns. Before, a person would log into a system, choose which data provider to use, choose which segments to run (auto intenders, fashionistas, moms and so on), run the campaign, and then check in on it periodically to optimize.

Now, all the data is available to an AI the computer system decides how to run the campaign based on given goals (CTR, CPA, site visits and so on) and tells you during and after the campaign about the decisions it made and why.Put this AI up against the best human opponent, and the computer should win unless a new and hitherto unknown variable is introduced or required data is unavailable.

There are still lots of things computers cannot do for us. For example, look at the United Airlines fiasco last April, when a man was dragged out of his seat after the flight was overbooked. Uniteds tagline is Fly the friendly skies. The incident was anything but friendly, and any current ad campaign touting so would be balked at.

To a human, the negative sentiment is obvious. The ad campaign would be pulled and a different strategy would be attempted in this case, a major PR push. But a computer would just notice that the ads arent performing as they once were but would continue to look for ways to optimize the campaign. It might even notice lots of interactions when Fly the Friendly Skies ads are placed next to images of a person being brutally pulled off the plane and place more ads there!

The way that artificial intelligence will affect us as consumers is more subtle than we think.Were unlikely to have a relationship with Siri or Alexa (see the movie Her), and although self-driving cars will become real in our lifetime, its unlikely that traffic will improve dramatically, since not everyone will use them, and ride-sharing or service-oriented vehicles will still infiltrate our roads, contributing to traffic.

The difference will be that cars, roads and signals may all be connected with an AI running the system based on our rules. We could expect the same amount of traffic, but the flow of traffic will be much better because AI will follow the rules, meaning no slow drivers in the fast lane! And we can do whatever we want while stuck in traffic rather than being wedded to the steering wheel.

Artificial intelligence, machine learning and self-aware systems are real.They will affect us and the way we do our jobs. All of us have opportunities in our current work to embrace these new tools and effect change in our lives that will make us more efficient.

While these systems may not be R2-D2, they are still revolutionary. If you invest in and take advantage of what AI can do for your business, good things are likely to happen to you. And if you dont, youll still discover that the revolution is real but you might not be on the right side of history.

Some opinions expressed in this article may be those of a guest author and not necessarily MarTech Today. Staff authors are listed here.

Go here to read the rest:

Demystifying AI: Understanding the human-machine relationship - MarTech Today