An Automatic Observer for Neural Nets
August 25, 2017
We are making progress in training AI systems through the neural net approach, but exactly how those systems make their decisions remains difficult to discern. Now, Tech Crunch reveals, “MIT CSAIL Research Offers a Fully Automated Way to Peer Inside Neural Nets.” Writer Darrell Etherington recalls that, a couple years ago, the same team of researchers described a way to understand these decisions using human reviewers. A fully automated process will be much more efficient and lead to greater understanding of what works and what doesn’t. Etherington explains:
Current deep learning techniques leave a lot of questions around how systems actually arrive at their results – the networks employ successive layers of signal processing to classify objects, translate text, or perform other functions, but we have very little means of gaining insight into how each layer of the network is doing its actual decision-making. The MIT CSAIL team’s system uses doctored neural nets that report back the strength with which every individual node responds to a given input image, and those images that generate the strongest response are then analyzed. This analysis was originally performed by Mechanical Turk workers, who would catalogue each based on specific visual concepts found in the images, but now that work has been automated, so that the classification is machine-generated. Already, the research is providing interesting insight into how neural nets operate, for example showing that a network trained to add color to black and white images ends up concentrating a significant portion of its nodes to identifying textures in the pictures.
The write-up points us to MIT’s own article on the subject for more information. We’re reminded that, because the human thought process is still largely a mystery to us, AI neural nets are based on hypothetical models that attempt to mimic ourselves. Perhaps, the piece suggests, a better understanding of such systems could inform the field of neuroscience. Sounds fair.
Cynthia Murrell, August 25, 2017
Lucidworks: The Future of Search Which Has Already Arrived
August 24, 2017
I am pushing 74, but I am interested in the future of search. The reason is that with each passing day I find it more and more difficult to locate the information I need as my routine research for my books and other work. I was anticipating a juicy read when I requested a copy of “Enterprise Search in 2025.” The “book” is a nine page PDF. After two years of effort and much research, my team and I were able to squeeze the basics of Dark Web investigative techniques into about 200 pages. I assumed that a nine-page book would deliver a high-impact payload comparable to one of the chapters in one of my books like CyberOSINT or Dark Web Notebook.
I was surprised that a nine-page document was described as a “book.” I was quite surprised by the Lucidworks’ description of the future. For me, Lucidworks is describing information access already available to me and most companies from established vendors.
The book’s main idea in my opinion is as understandable as this unlabeled, data-free graphic which introduces the text content assembled by Lucidworks.
However, the pamphlet’s text does not make this diagram understandable to me. I noted these points as I worked through the basic argument that client server search is on the downturn. Okay. I think I understand, but the assertion “Solr killed the client-server stars” was interesting. I read this statement and highlighted it:
Other solutions developed, but the Solr ecosystem became the unmatched winner of the search market. Search 1.0 was over and Solr won.
In the world of open source search, Lucene and Solr have gained adherents. Based on the information my team gathered when we were working on an IDC open source search project, the dominant open source search system was Lucene. If our data were accurate when we did the research, Elastic’s Elasticsearch had emerged as the go-to open source search system. The alternatives like Solr and Flaxsearch have their users and supporters, but Elastic, founded by Shay Branon, was a definite step up from his earlier search service called Compass.
In the span of two and a half years, Elastic had garnered more than a $100 million in funding by 2014and expanded into a number adjacent information access market sectors. Reports I have received from those attending Elastic meetings was that Elastic was putting considerable pressure on proprietary search systems and a bit of a squeeze on Lucidworks. Google’s withdrawing its odd duck Google Search Appliance may have been, in small part, due to the rise of Elasticsearch and the changes made by organizations trying to figure out how to make sense of the digital information to which their staff had access.
But enough about the Lucene-Solr and open source versus proprietary search yin and yang tension.
Attack Planes Soon to Be Equipped with Lasers
August 24, 2017
The US Air Force soon will be equipping its attack planes with laser weapons to fight UAVs that terrorist organizations may use for launching attacks.
According to an op-ed published by Defense One and titled The Future of the Air Force, the author says:
We are currently investing in the hardware to ensure space superiority; in the near future we will need to grow the number of space airmen and the accompanying infrastructure much like we did for the combat Air Force 40 years ago.
Wars in the future will be fought on multiple fronts, including space. As per the op-ed, the US Air Force needs to be equipped sufficiently to fight these battles without putting people on the front line.
The op-ed also says about the acquisition of an Israeli company that enables attack planes using lasers to fend off drones that are used for dropping bombs and other weapons. The acquisition does not come as a surprise as Pentagon had been researching use of lasers as tactical weapons since long. It seems the days of Star Wars are very near.
Vishal Ingole, August 24, 2017
Demanding AI Labels
August 16, 2017
Artificial intelligence has become a standard staple in technology driven societies. It still feels like that statement should still only be in science-fiction, but artificial intelligence is a daily occurrence in developed nations. We just do not notice it. When something becomes standard practice, one thing we like to do is give it labels. Guess what Francesco Corea did over at Medium in his article, “Artificial Intelligence Classification Matrix”? He created terminology to identify companies that specialize in machine intelligence.
Before we delve into his taxonomy, he stated that if the framework for labeling machine intelligence companies is too narrow it is counterproductive to the sector’s purpose of maintaining flexibility. Corea came up with four ways to classify machine intelligence companies :
i) Academic spin-offs: these are the more long-term research-oriented companies, which tackle problems hard to break. The teams are usually really experienced, and they are the real innovators who make breakthroughs that advance the field.
- ii) Data-as-a-service (DaaS): in this group are included companies which collect specific huge datasets, or create new data sources connecting unrelated silos.
iii) Model-as-a-service (MaaS): this seems to be the most widespread class of companies, and it is made of those firms that are commoditizing their models as a stream of revenues.
-
iv) Robot-as-a-service (RaaS): this class is made by virtual and physical agents that people can interact with. Virtual agents and chatbots cover the low-cost side of the group, while physical world systems (e.g., self-driving cars, sensors, etc.), drones, and actual robots are the capital and talent-intensive side of the coin.
There is also a chart included in the article that explains the differences between high vs. low STM and high vs. low defensibility. Machine learning companies obviously cannot be categorized into one specific niche. Artificial intelligence can be applied to nearly any field and situation.
Whitney Grace, August 16, 2017
Doctors Fearful of Technology? Too Bad for Them and Maybe the Patients?
July 27, 2017
IBM, Google, and other outfits want doctors to get with the technology program. Sure, docs use mobile phones, but email and such wonderful innovations as selfies from the operating theatre have not yet caught on. Watching my doc fumble with the required online medical record system is interesting. Try it sometime. Puzzled expressions, eye squinting, and sloooow keyboarding are part of the show. One of my docs expressed interest in my Dark Web Notebook. I sent him a link so he could download a comp copy. Guess what? He couldn’t figure out how to download the book. Amazing expertise.
I read a Thomson Reuters’ article which seems to stray dangerously close to my view of technology in the medical profession. Mind you, here in Louisville sales people are in the operating room to provide information to a doc who may not be familiar with a new gadget. Get enough gadgets and peddlers in the facility and the patients may have to rest on gurneys in the hall.
But I digress. The write up i noticed was “Doctors View Technology as Largely Problematic.” I highlighted this “real” news statement:
69 percent of the 100 doctors in the audience said increased reliance on technology and electronic health records only served to separate them from their patients….But the biggest problem stemming from technology for the doctors, and the bane of many doctors’ existence, is the electronic health record, also known as an EHR.
Now think about the over the top marketing from IBM about Watson’s ability in a narrow field like bladder cancer. Put that Anderson affair out of your main. Google continues to push forward with an even more interesting approach. I recall the phrase was “solving death.” And there are other outfits which believe that their technologists can make life so much better for doctors.
Seems like the revolution may take a bit more time. The good news is that since Google has not solved death, the doubting docs will die. Their replacements may be more into the IBM, Google, et al approach to health care.
No worries in Harrod’s Creek. We just use a mixture of black powder and bourbon to cure all manner of ills.
Stephen E Arnold, July 27, 2017
Shopping List of Technologies That Will Make Life Really Good… Well, Some Lives’ Lives
July 26, 2017
I read or rather scanned an infographic called “Things to Come: A Timeline of Technology.” Sci fi? Nope, the stuff that makes VCs’ ice cold blood flow just a little bit faster. In terms of search, two entries seem to suggest that finding information will be a semi big thing:
- 2036, The next evolution of AI. Okay, but won’t today’s whiz bang systems get better really rapidly? According to the “Things to Come”, AI has its booster rocker fired in 19 years. Make your hotel reservation the IBM Watson B&B today.
- 2040. Genetic computing. Okay, more than a GIF in a protein chain and less than a quantum computer maybe?
- 2045, Algorithmic advances. Hmm. Algorithms are “evolving” if one is into the genetic algorithms. For the rule-based approaches, that seems to move less rapidly. But advances down the road in 28 years. Seems like a made up target date. But that’s just the view of an old guy in rural Kentucky
What I did like about the VC blood enhancer was this list of buzzwords. How many can you define?
- Eye controlled technology
- Paper diagnostics
- Designer antibiotics
- Ingestible robots
- Smart clothing
- Photonics in space
- Volcanic mining
- Spintronics revolution
- Carbon breathing batteries
- Super antivirals
- Diamond batteries
- Optogenetics
- Nano feasibility
- Unhackable quantum Internet
- Cheap solar power
- Biomimetic materials
- 3D printing in every home
- Fully immersive computer interface
- Self sufficient energy ecosystem
- Germ line genetic modification
- Holograpic pets
- Rapid genetic screening
- Microwave rockets
- Space based solar energy
- Fusion power
- Evolutionary enhancement
- Carbon sequestration
- Geoengineering
- Wavetop and undersea cities
- Geoneutrino satellites.
This list reminds me of my eavesdropping on a group of PC Magazine editors cooking up a list of the Top 10 trends in personal computers or the brainstorming sessions at Booz, Allen when a Top Dog said, “We need to identify hot business trends.”
Anyway, you can use these buzzwords in your LinkedIn résumé or in your next tweet. Better yet, print out the list and look for mid tier consulting firms at the next conference you attend. You can pitch your expertise in wave top and undersea cities and hope the individual with whom you are speaking has never watched the film Waterworld.
Stephen E Arnold, July 26, 2017
Machine Learning Does Not Have the Mad Skills
July 25, 2017
Machine learning and artificial intelligence are computer algorithms that will revolutionize the industry, but The Register explains there is a problem with launching it: “Time To Rethink Machine Learning: The Big Data Gobble Is OFF The Menu.” The technology industry is spouting that 50 percent of organizations plan to transform themselves with machine learning, but the real truth is that it is less than 15 percent.
The machine learning revolution has supposedly started, but in reality, the cannon has only be fired and the technology has not been implemented. The problem is that while companies want to use machine learning, they are barely getting off the ground with big data and machine learning is much harder. Organizations do not have workers with the skills to launch machine learning and the tech industry as a whole has a huge demand for skilled workers.
Part of this inaction comes down to the massive gap between ML (and AI) myth and reality. As David Beyer of Amplify Partners puts it: ‘Too many businesses now are pitching AI almost as though it’s batteries included.’ This is dangerous because it leads companies to either over-invest (and then face a tremendous trough of disillusionment), or to steer clear when the slightest bit of real research reveals that ML is very hard and not something the average Python engineer is going to spin up in her spare time.
Organizations also do not have the necessary amount of data to make machine learning feasible and they also lack the corporate culture to do the required experimentation for machine learning to succeed.
This article shares a story that we have read many times before. The tech industry gets excited about the newest shiny object, it explodes in popularity, then they realize that the business world is not ready for implementing the technology.
Whitney Grace, July 25, 2017
ArnoldIT Publishes Technical Analysis of the Bitext Deep Linguistic Analysis Platform
July 19, 2017
ArnoldIT has published “Bitext: Breakthrough Technology for Multi-Language Content Analysis.” The analysis provides the first comprehensive review of the Madrid-based company’s Deep Linguistic Analysis Platform or DLAP. Unlike most next-generation multi-language text processing methods, Bitext has crafted a platform. The document can be downloaded from the Bitext Web site via this link.
Based on information gathered by the study team, the Bitext DLAP system outputs metadata with an accuracy in the 90 percent to 95 percent range.
Most content processing systems today typically deliver metadata and rich indexing with accuracy in the 70 to 85 percent range.
According to Stephen E Arnold, publisher of Beyond Search and Managing Director of Arnold Information Technology:
“Bitext’s output accuracy establish a new benchmark for companies offering multi-language content processing system.”
The system performs in near real time, more than 15 discrete analytic processes. The system can output enhanced metadata for more than 50 languages. The structured stream provides machine learning systems with a low cost, highly accurate way to learn. Bitext’s DLAP platform integrates more than 30 separate syntactic functions. These include segmentation, tokenization (word segmentation, frequency, and disambiguation, among others. The DLAP platform analyzes more than 15 linguistic features of content in any of the more than 50 supported languages. The system extracts entities and generates high-value data about documents, emails, social media posts, Web pages, and structured and semi-structured data.
DLAP Applications range from fraud detection to identifying nuances in streams of data; for example, the sentiment or emotion expressed in a document. Bitext’s system can output metadata and other information about processed content as a feed stream to specialized systems such as Palantir Technologies’ Gotham or IBM’s Analyst’s Notebook. Machine learning systems such as those operated by such companies as Amazon, Apple, Google, and Microsoft can “snap in” the Bitext DLAP platform.
Copies of the report are available directly from Bitext at https://info.bitext.com/multi-language-content-analysis Information about Bitext is available at www.bitext.com.
Kenny Toth, July 19, 2017
The New York Times Pairs up with Spotify for Subscription Gains
July 18, 2017
The article on Quartz Media titled The New York Times Thinks People Will Still Pay for News—
If Given Free Music examines the package deal with Spotify currently being offered by the Times. While subscriptions to the news publication have been on the rise thanks in large part to Donald Trump, they are still hurting. The article points out that if the news and music industries have one thing in common, it is trying to get people to pay for their services.
The two companies announced an offer… giving a free year of Spotify Premium to anyone in the US who signs up for an all-access subscription to the news publication. Premium normally costs $120 a year, and the offer slashes the price of an all-access Times subscription too—from $6.25 a week to $5 a week… While it may seem like both companies will take a hit from these discounts, the boost in new subscribers/readers will likely more than make up for it.
It is a match made on Tinder, a coupling for the new world order. Will this couple get along? As millennials seek new outlets for activism, purchasing a subscription to the Times is a few steps above posting a rant on Facebook. Throw a year of Spotify into the mix and this deal is really appealing to anyone who doesn’t consider the Times a “liberal rag.” So maybe the Donald won’t be interested, but the rest of us sure might consider paying $5/month for legitimate news and music.
Chelsea Kerwin, July 18, 2017
Hope for Improvement in Predictive Modeling
July 18, 2017
A fresh approach to predictive modeling may just improve the process exponentially. Phys.org reports, “Molecular Dynamics, Machine Learning Create ‘Hyper-Predictive Computer Models.” The insight arose, and is being tested, at North Carolina State University.
The article begins by describing the incredibly complex and costly process of drug development, including computer models that predict the effects of certain chemical compounds. Such models traditionally rely on QSAR modeling and molecular docking. We learn:
Denis Fourches, assistant professor of computational chemistry, wanted to improve upon the accuracy of these QSAR models. … Fourches and Jeremy Ash, a graduate student in bioinformatics, decided to incorporate the results of molecular dynamics calculations – all-atom simulations of how a particular compound moves in the binding pocket of a protein – into prediction models based on machine learning. ‘Most models only use the two-dimensional structures of molecules,’ Fourches says. ‘But in reality, chemicals are complex three-dimensional objects that move, vibrate and have dynamic intermolecular interactions with the protein once docked in its binding site. You cannot see that if you just look at the 2-D or 3-D structure of a given molecule.’
See the article for some details about the team’s proof-of-concept study. Fourches asserts the breakthrough delivers a simulation that would previously have been built over six months in a mere three hours. That is quite an improvement! If this technique pans out, we could soon see more rapid prediction not only in pharmaceuticals but many other areas as well. Stay tuned.
Cynthia Murrell, July 18, 2017
-
- Subscribe to Beyond Search
Feature archive
News archive
-
