Tag: IBM Watson supercomputer

Judement Day Update: IBM Watson’s Hard Questions

https://i0.wp.com/d.fastcompany.net/multisite_files/fastcompany/imagecache/1280/poster/2014/08/3034948-poster-p-1-now-ibms-watson-tackles-questions-that-have-no-answers.jpgWhen IBM’s artificial intelligence program – named Watson – won Jeopardy in 2011, it was a major boon for the industry. However, far from showing that it was Turing-compatible, Watson was merely processing information that it had been programmed to know. But now, IBM is pushing the software forward in the hopes that the machine will be able to answer the really tough questions – i.e. the ones that have no answer and require educated guesses.

This is part of IBMs attempts to turn Watson into a new line of business and make it useful in a wide range of industries that are dealing lately with an overwhelming amount of data. At an event in New York at the end of August, IBM showed off the ways some of its early customers are using the Watson “Discovery Advisor” in research, development, and innovation, especially in the realm of biotech and life sciences.

IBM_Watson1Watson’s aim is to speed up discoveries by teams of researchers by, for example, scanning and interpreting millions of scientific books, articles, and data points–far more than any person’s brain could analyze–and generating new hypothesis or leads that might be fruitful to investigate. As John Gordon, the vice-president of IBM’s Watson group, put it, it’s all about giving researchers “smarter hunches”:

It’s not giving answers that people know anymore, it’s pointing people in directions that they should investigate. We’re talking about a computing system that inspires people.

Scientists at the Baylor College of Medicine and IBM Research have already used Watson to discover new pathways to cancer therapies, which they reported in a study presented at an academic conference late last month. Watson looked closely at 70,000 scientific articles on a protein, called p53, that’s involved in more than half of all cancers, and picked out 6 different proteins that it felt were good candidates for new drugs and therapies.

IBM_watsonDrug companies, too, which are struggling today to develop new commercial drugs, are some of the earliest users of Watsons predictive capabilities. Sanofi is using Watson to look through the research literature and its own data to find new uses for its existing drugs on the market. And Johnson & Johnson has developed a system that analyzes clinical studies to compare the efficacy and safety of different treatments.

Soledad Cepeda, Johnson & Johnson’s director of epidemiology, used the example of back pain, for which there are 27 treatments studied in more than 3,000 clinical trials. As she explained:

[Analyzing this] is slow, it’s tedious, it’s expensive, and it’s prone to errors. Now imagine we can teach Watson to do that for us. So instead of six months, Watson can do it in minutes.

biomedicineJohnson & Johnson has been working to train Watson to read each study, put it in context, and pick out how many patients dropped out of the study or trial due to side effects or ineffective results. If Watson can give researchers all of this comparative data, rather than them combing through thousands of papers, it would allow researchers to come up with better questions to ask and directions to explore.

But of course, this is not always an easy things for Watson to do and requires setup and new skills for it to learn. For example, in Johnson & Johnson’s work, in the studies Watson was analyzing, authors often reported the key data in the form of flow diagrams. So Soledad and her team had to first teach Watson to correctly read flow diagrams of varying levels of complexity and design.

metadataSo far, Watson’s earliest discovery applications have been in the field of biomedical research. But the company hopes it will prove useful in a wide array of fields where the data available to analyze is growing faster than even the world’s top experts are capable of comprehending – such as law enforcement and finance. Whether it’s determining whodunnit, or predicting market trends in the next quarter, Watson could prove very helpful in tackling the task of big-data crunching.

To illustrate the potential for law enforcement, Roberto Villasenor, chief of police for the city of Tucson, Arizona, spoke at the event and detailed an open case of a young child who went missing from her home. Over two years of investigating, the police have generated 15,000 pages of lab reports, records, and warrants, 25,00 pages of interviews, 4,000 pages of transcribed wiretaps, and much other data.

artificial-intelligenceHis department has already worked with IBM on software that integrates different police databases to make it easier for investigators to make connections between disparate data sources. But he hopes systems like Watson will eventually go further and be capable of aiding investigators in combing through data, making subtle connections, and generating new leads in difficult cases.

However, the most public demonstration of Watson’s new abilities thus far has been in the form of “Chef Watson”. IBM has put Watson to the task of learning how to cook and then creating creative tasty and unexpected new recipes. It debuted this capability at a food truck at SxSW this year, but has also been working with the Institute of Culinary Education and Bon Appetit magazine to refine and stretch Watson’s cooking skills.

ibm-watsonBon Appetit is now beta testing a consumer app that allows readers to input an ingredient and desires and have Watson generate suggested recipes. It held a Watson recipe contest this summer–the winner of the “best use of Watson as a creative discovery tool” was a “Roasted tomato and mozzarella tart” recipe. Cooking isn’t like curing cancer or solving crimes, but to IBM it’s most about getting the public excited about its advances.

As IBM senior vice president Mike Rhodin said:

Much the same way that Jeopardy helped people understand systems that can answer questions using natural language, Chef Watson is a way for us to understand how these new systems can be used in our everyday lives.

And that’s the real aim here, long-term, which is adapting computer-learning into our daily lives. And given time, we may have access to AIs that can do the difficult and mind-numbing task of sifting through gigabytes, terabytes and even petabytes of information in order to find answers to tough questions, discern the patterns, and come up with solutions. One can only hope they don’t use that information for evil and try to destroy us!

Knowledge is power, after all…

Sources: fastcoexist.com, bcm.edu

The Future is Here: Cancer Drug Developed by AI

AI'sThe development of cancer drugs is a costly, expensive, time-consuming process that has a high probability rate of failure. On average, it takes 24 to 48 months to find a suitable candidate and costs upwards of $100 million. And in the end, roughly 95% of all potential drugs fail in clinical trials. Because of this, scientists are understandably looking for a way to speed up the discovery process.

That’s where the anti-cancer drug known as BPM 31510 comes in play. Unlike most pharmaceuticals, it was developed by artificial intelligence instead of a group of researchers toiling away in a lab. Created by biotech company Berg (named after real estate billionaire Carl Berg) the company seeks to use artificial intelligence to design cancer drugs that are cheaper, have fewer side effects, and can be developed in half the time it normally takes.

drugsTowards this end, they are looking to data-driven methods of drug discovery. Instead of generating cancer drugs based on chemical compounds identified in labs, the company compares tissue, urine, and blood samples from cancer patients and healthy patients, generating tens of trillions of data points that are fed into an artificial intelligence system. That system crunches all the data, looking for problems.

BPM 31510, which is the first of Berg’s drugs to get a real-world test, focuses on mitochondria – a framework within cells that’s responsible for programmed cell death. Normally, mitochondria triggers damaged cells to die. When cancer strikes, this process goes haywire, and the damaged cells spread. Berg’s drug, if successful, will be able to restore normal cell death processes by changing the metabolic environment within mitochondria.

MitochondriaSpeaking on the subject of the drug, which is now in human-clinical trials, Berg president and co-founder Niven Narain said:

BPM 31510 works by switching the fuel that cancer likes to operate on. Cancer cells prefer to operate in a less energy-efficient manner. Cancers with a high metabolic function, like triple negative breast cancer, glioblastoma, and colon cancer–that’s the sweet spot for this technology.

IBM is also leveraging artificial intelligence in the race to design better cancer treatments. In their case, this involves their much-heralded supercomputer Watson looking for better treatment options for patients. In a trial conducted with the New York Genome Center, Watson has been scanning mutations found in brain cancer patients, matching them with available treatments.

dna_cancerAll of these efforts are still in early days, and even on its accelerated timeline, BPM 31510 is still years away from winning an FDA approval. But, as Narain points out, the current drug discovery system desperately needs rethinking. With a success rate of 1 out of 20, their is definitely room for improvement. And a process that seeks to address cancer in a way that is more targeted, and more personalized is certainly in keeping with the most modern approaches to medicine.

Source: fastcoexist.com

Judgement Day Update: Bionic Computing!

big_blue1IBM has always been at the forefront of cutting-edge technology. Whether it was with the development computers that could guide ICBMs and rockets into space during the Cold War, or the creation of the Internet during the early 90’s, they have managed to stay on the vanguard by constantly looking ahead. So it comes as no surprise that they had plenty to say last month on the subject of the next of the next big leap.

During a media tour of their Zurich lab in late October, IBM presented some of the company’s latest concepts. According to the company, the key to creating supermachines that 10,000 faster and more efficient is to build bionic computers cooled and powered by electronic blood. The end result of this plan is what is known as “Big Blue”, a proposed biocomputer that they anticipate will take 10 years to make.

Human-Brain-project-Alp-ICTIntrinsic to the design is the merger of computing and biological forms, specifically the human brain. In terms of computing, IBM is relying the human brain as their template. Through this, they hope to be able to enable processing power that’s densely packed into 3D volumes rather than spread out across flat 2D circuit boards with slow communication links.

On the biological side of things, IBM is supplying computing equipment to the Human Brain Project (HBP) – a $1.3 billion European effort that uses computers to simulate the actual workings of an entire brain. Beginning with mice, but then working their way up to human beings, their simulations examine the inner workings of the mind all the way down to the biochemical level of the neuron.

brain_chip2It’s all part of what IBM calls “the cognitive systems era”, a future where computers aren’t just programmed, but also perceive what’s going on, make judgments, communicate with natural language, and learn from experience. As the description would suggest, it is closely related to artificial intelligence, and may very well prove to be the curtain raiser of the AI era.

One of the key challenge behind this work is matching the brain’s power consumption. The ability to process the subtleties of human language helped IBM’s Watson supercomputer win at “Jeopardy.” That was a high-profile step on the road to cognitive computing, but from a practical perspective, it also showed how much farther computing has to go. Whereas Watson uses 85 kilowatts of power, the human brain uses only 20 watts.

aquasar2Already, a shift has been occurring in computing, which is evident in the way engineers and technicians are now measuring computer progress. For the past few decades, the method of choice for gauging performance was operations per second, or the rate at which a machine could perform mathematical calculations.

But as a computers began to require prohibitive amounts of power to perform various functions and generated far too much waste heat, a new measurement was called for. The new measurement that emerged as a result was expressed in operations per joule of energy consumed. In short, progress has come to be measured in term’s of a computer’s energy efficiency.

IBM_Research_ZurichBut now, IBM is contemplating another method for measuring progress that is known as “operations per liter”. In accordance with this new paradigm, the success of a computer will be judged by how much data-processing can be squeezed into a given volume of space. This is where the brain really serves as a source of inspiration, being the most efficient computer in terms of performance per cubic centimeter.

As it stands, today’s computers consist of transistors and circuits laid out on flat boards that ensure plenty of contact with air that cools the chips. But as Bruno Michel – a biophysics professor and researcher in advanced thermal packaging for IBM Research – explains, this is a terribly inefficient use of space:

In a computer, processors occupy one-millionth of the volume. In a brain, it’s 40 percent. Our brain is a volumetric, dense, object.

IBM_stacked3dchipsIn short, communication links between processing elements can’t keep up with data-transfer demands, and they consume too much power as well. The proposed solution is to stack and link chips into dense 3D configurations, a process which is impossible today because stacking even two chips means crippling overheating problems. That’s where the “liquid blood” comes in, at least as far as cooling is concerned.

This process is demonstrated with the company’s prototype system called Aquasar. By branching chips into a network of liquid cooling channels that funnel fluid into ever-smaller tubes, the chips can be stacked together in large configurations without overheating. The liquid passes not next to the chip, but through it, drawing away heat in the thousandth of a second it takes to make the trip.

aquasarIn addition, IBM also is developing a system called a redox flow battery that uses liquid to distribute power instead of using wires. Two types of electrolyte fluid, each with oppositely charged electrical ions, circulate through the system to distribute power, much in the same way that the human body provides oxygen, nutrients and cooling to brain through the blood.

The electrolytes travel through ever-smaller tubes that are about 100 microns wide at their smallest – the width of a human hair – before handing off their power to conventional electrical wires. Flow batteries can produce between 0.5 and 3 volts, and that in turn means IBM can use the technology today to supply 1 watt of power for every square centimeter of a computer’s circuit board.

IBM_Blue_Gene_P_supercomputerAlready, the IBM Blue Gene supercomputer has been used for brain research by the Blue Brain Project at the Ecole Polytechnique Federale de Lausanne (EPFL) in Lausanne, Switzerland. Working with the HBP, their next step ill be to augment a Blue Gene/Q with additional flash memory at the Swiss National Supercomputing Center.

After that, they will begin simulating the inner workings of the mouse brain, which consists of 70 million neurons. By the time they will be conducting human brain simulations, they plan to be using an “exascale” machine – one that performs 1 exaflops, or quintillion floating-point operations per second. This will take place at the Juelich Supercomputing Center in northern Germany.

brain-activityThis is no easy challenge, mainly because the brain is so complex. In addition to 100 billion neurons and 100 trillionsynapses,  there are 55 different varieties of neuron, and 3,000 ways they can interconnect. That complexity is multiplied by differences that appear with 600 different diseases, genetic variation from one person to the next, and changes that go along with the age and sex of humans.

As Henry Markram, the co-director of EPFL who has worked on the Blue Brain project for years:

If you can’t experimentally map the brain, you have to predict it — the numbers of neurons, the types, where the proteins are located, how they’ll interact. We have to develop an entirely new science where we predict most of the stuff that cannot be measured.

child-ai-brainWith the Human Brain Project, researchers will use supercomputers to reproduce how brains form in an virtual vat. Then, they will see how they respond to input signals from simulated senses and nervous system. If it works, actual brain behavior should emerge from the fundamental framework inside the computer, and where it doesn’t work, scientists will know where their knowledge falls short.

The end result of all this will also be computers that are “neuromorphic” – capable of imitating human brains, thereby ushering in an age when machines will be able to truly think, reason, and make autonomous decisions. No more supercomputers that are tall on knowledge but short on understanding. The age of artificial intelligence will be upon us. And I think we all know what will follow, don’t we?

Evolution-of-the-Cylon_1024Yep, that’s what! And may God help us all!

Sources: news.cnet.com, extremetech.com

Judgement Day Update: Deep Learning and AIs

brain_chip2Moore’s Law, a sort of general law of computing that was coined by Intel co-founder Gordon E. Moore, states that the number of transistors on integrated circuits doubles approximately every two years. In accordance with this observation, which was made back in 1965, computers have evolved to the point where they are over 650 times faster than they were in the early 1970’s.

Given this trend, scientists and futurists have long suspected that it is only a matter of time before computers surpass human beings in terms of raw intelligence. Already, supercomputers like IBM’s Watson, University of Illinois’ “Blue Waters”, and MIT’s ConceptNet 4 have demonstrated that they are superior in terms of raw knowledge and retention. But when it comes to actual thinking – i.e. independent reasoning and common sense – they are seriously lacking.

IBM_Blue_Gene_P_supercomputerYes, despite all the progress made in the field of computing in the past 40 years, bridging that gap between computational power and artificial intelligence has remained elusive. And at this point, no one is quite sure what it will take to make that leap happen. However, a number of developers and designers are coming up with breakthroughs that may constitute a small jump.

One such breakthrough comes from Standford University, where computer engineers have come up with a new algorithm that could give computers the power to more reliably interpret language. Called Neural Analysis of Sentiment – or NaSent for short – the algorithm seeks to improve on current methods of written language analysis by drawing inspiration from the human brain.

AI_picNaSent is part of a movement in computer science known as deep learning, a new field that seeks to build programs that can process data in much the same way the brain does. According to Richard Socher, the Stanford University grad student who helped develop NaSent:

In the past, sentiment analysis has largely focused on models that ignore word order or rely on human experts. While this works for really simple examples, it will never reach human-level understanding because word meaning changes in context and even experts cannot accurately define all the subtleties of how sentiment works. Our deep learning model solves both problems.

Socher was joined by artificial-intelligence researchers Chris Manning, a Professor of Computer Science and Linguistics at Standford, and Andrew Ng, one of the engineers behind Google’s deep learning project and who is working on creating the “Google Brain”.  Their aim is to develop algorithms that can operate without continued help from humans.

word_cloudWhat sets this algorithm apart is its ability to identify the meaning of words in the context of phrases and sentences. This is a big change from previous methods of sentiment analysis which have been limited to parsing through a collection of words and ranking them as either positive or negative without taking word order into account.

To create NaSent, Socher and his team used 12,000 sentences taken from the movie reviews website Rotten Tomatoes. They split these sentences into roughly 214,000 phrases that were labeled as very negative, negative, neutral, positive, or very positive, and then they fed this labeled data into the system, which NaSent then used to predict whether sentences were positive, neutral or negative on its own.

metadataAccording to the researchers, NaSent is about 85 percent accurate, a 5% improvement over previous models. However, Socher and his team are working on increasing that rate of success by feeding the system more data from Twitter and the Internet Movie Database, and have also set up a live demo where people can type in their own sentences.

And whereas the field of deep learning began in the academic world, it has since spread to web giants such as Google and Facebook. In both cases, the companies have taken to hiring people familiar with the field to help them sort through the growing mountains of data they are indexing to help them improve their products, particularly those that already rely on machine learning.

In Google’s case, these include voice recognition technology, as well as their ongoing work in neural networks. In the case of Facebook, they are looking to deep learning to help them improve Graph Search, a search engine that allows people to search activity on their network. This presents a challenge, mainly because machines have a hard time interpreting the nuances of human language. Just ask Watson, IBM’s own supercomputer!

Sources: wired, (2), (3), engineering.stanford.edu