That's the message of a big Wall Street Journal post-mortem on Watson, the IBM project that was supposed to turn IBM's computing prowess into a scalable program that could deliver state-of-the-art personalized cancer treatment protocols to millions of patients around the world. Watson in general, and its oncology application in particular, has been receiving a lot of skeptical coverage of late; STAT published a major investigation last year, reporting that Watson was nowhere near being able to live up to IBM's promises. After that article came out, the IBM hype machine started toning things down a bit. But while a lot of the problems with Watson are medical or technical, they're deeply financial, too. IBM is shrinking: In 2011, when the company first introduced the idea that Watson might be able to one day cure cancer, its revenues were $107 billion. They've gotten smaller every year since, ending up at $79 billion in 2017.

Business leaders understand the advantage of using the power of artificial intelligence and machine learning to stay ahead of their competitors. However, understanding the power of AI is a lot different than actually successfully implementing it in companies. For example, in 2017, Gartner estimated that Big Data projects have a success rate of only 15%. While organizational factors may be a primary reason for this poor success rate, another reason for such a high failure rate could be due to a lack of AI / Machine Learning talent needed to successfully pursue these types of projects. Specifically, it's been shown that there is a lack of advanced machine learning talent among data professionals; less than 20% of surveyed data professionals said they were competent in such areas as Natural Language Processing (19%), Recommendation Engines (14%), Reinforcement Learning (6%), Adversarial Learning (4%) and Neural Networks – RNNs (15%).

We warmly invite you to read the presentation given to the INSEAD alumni à BCG Paris site to understand what is going on!.

I've been using Jupyter Notebooks with great delight for many years now, mostly with Python, and it's validating to see that their popularity keeps growing, both in academia and the industry. I do have a pet peeve though, which is the lack of a first-class visual debugger similar to these available in other IDEs like Eclipse, IntelliJ, or Visual Studio Code. Some would rightfully point out that Jupyter already supports pdb for simple debugging, where you can manually and sequentially enter commands to do things like inspect variables, set breakpoints, etc. -- and this is probably sufficient when it comes to debugging simple analytics. To raise the bar, the PixieDust team is happy to introduce the first (to the best of our knowledge) visual Python debugger for Jupyter Notebooks. As advertised, the PixieDebugger is a visual Python debugger built as a PixieApp, and includes a source editor, local variable inspector, console output, the ability to evaluate Python expressions in the current context, breakpoints management, and a toolbar for controlling code execution.

You've got to love the idea of IBM Watson. The super-computer using advanced AI to learn everything, faster and better than any human being could ever hope to do. The hope is it would help us solve some of our most pressing problems. One of IBM's (IBM) high-profile challenges was their desire to cure cancer. Unfortunately, it has not happened.

We study the problem of finite ontology mediated query answering (FOMQA), the variant of OMQA where the represented world is assumed to be finite, and thus only finite models of the ontology are considered. We adopt the most typical setting with unions of conjunctive queries and ontologies expressed in description logics (DLs). The study of FOMQA is relevant in settings that are not finitely controllable. This is the case not only for DLs without the finite model property, but also for those allowing transitive role declarations. When transitive roles are allowed, evaluating queries is challenging: FOMQA is undecidable for SHOIF and only known to be decidable for the Horn fragment of ALCIF. We show decidability of FOMQA for three proper fragments of SOIF: SOI, SOF, and SIF. Our approach is to characterise models relevant for deciding finite query entailment. Relying on a certain regularity of these models, we develop automata-based decision procedures with optimal complexity bounds.

It's no surprise by now that voice assistants like Google Home and Amazon Echo were some of the biggest sellers of the latest holiday season. In addition, people are using Siri on their Apple iPhones to ask questions and receive answers. And some people now prefer to ask Google questions directly with their voice while working on a laptop or desktop rather than type in a question. So how can you get out in front of this voice search trend? Your first step should be to think in terms of the long tail keywords that people are actually using in their voice queries.

Nothing kills a bad idea faster than good advertising. Yet, the diffusion of information into a system can be essential--especially in medicine. So the balance between the kind of stuff that "sticks to the roof of your customer's brain" and valuable information can be tricky and even contradictory. For most of us, the introduction of Watson's skill set wasn't as a peer-reviewed paper published in a top academic journal--it was a guy name Ken Jennings and the popular TV game show Jeopardy. After a winning streak of 74 shows, Jennings took on IBM Watson and the rest is history.

A key aspect of VQA models that are interpretable is their ability to ground their answers to relevant regions in the image. Current approaches with this capability rely on supervised learning and human annotated groundings to train attention mechanisms inside the VQA architecture. Unfortunately, obtaining human annotations specific for visual grounding is difficult and expensive. In this work, we demonstrate that we can effectively train a VQA architecture with grounding supervision that can be automatically obtained from available region descriptions and object annotations. We also show that our model trained with this mined supervision generates visual groundings that achieve a higher correlation with respect to manually-annotated groundings, meanwhile achieving state-of-the-art VQA accuracy.

Attention mechanisms in biological perception are thought to select subsets of perceptual information for more sophisticated processing which would be prohibitive to perform on all sensory inputs. In computer vision, however, there has been relatively little exploration of hard attention, where some information is selectively ignored, in spite of the success of soft attention, where information is re-weighted and aggregated, but never filtered out. Here, we introduce a new approach for hard attention and find it achieves very competitive performance on a recently-released visual question answering datasets, equalling and in some cases surpassing similar soft attention architectures while entirely ignoring some features. Even though the hard attention mechanism is thought to be non-differentiable, we found that the feature magnitudes correlate with semantic relevance, and provide a useful signal for our mechanism's attentional selection criterion. Because hard attention selects important features of the input information, it can also be more efficient than analogous soft attention mechanisms. This is especially important for recent approaches that use non-local pairwise operations, whereby computational and memory costs are quadratic in the size of the set of features.