Previous studies on continual knowledge learning (CKL) in large language models (LLMs) have predominantly focused on approaches such as regularization, architectural modifications, and rehearsal techniques to mitigate catastrophic forgetting. However, these methods naively inherit the inefficiencies of standard training procedures, indiscriminately applying uniform weight across all tokens, which can lead to unnecessary parameter updates and increased forgetting. To address these shortcomings, we propose a novel CKL approach termed Train-Attention-Augmented Language Model (TAALM), which enhances learning efficiency by dynamically predicting and applying weights to tokens based on their usefulness. This method employs a meta-learning framework that optimizes token importance predictions, facilitating targeted knowledge updates and minimizing forgetting. Also, we observe that existing benchmarks do not clearly exhibit the trade-off between learning and retaining, therefore we propose a new benchmark, LAMA-ckl, to address this issue.

An Artificial Intelligence Incubation Hub (AIIH), set up jointly by the Army and Bharat Electronics Limited (BEL) as part of a pact inked earlier this year, will be identifying artificial intelligence projects that may be critical for the Army and coordinate the evaluation of trials in those cases, News18 has learnt. As per defence sources, the AIIH will also be tasked with identifying partners from the industry, academia or start-ups to check the feasibility of ideas related to AI projects for implementation and will be responsible for hiring of resources and necessary hardware for validation of suggested AI projects for the force. Based on BEL's research and development expertise and areas where the Army could use AI, the new hub will be working in major domain areas of machine learning, auto-platforms, AI-based swarm technology, robotics, big data analytics, image processing as well as cyber security and AI-based response mechanism. In March this year, the Army and BEL had inked a pact to collaborate in AI for defence applications. The step was taken after the defence ministry had pushed for fast-tracking incorporation of the new tech for defence applications, as part of which the defence services were asked to collaborate with defence PSUs.

European software major SAP and startup accelerator Techstars will invest in at least 10 artificial intelligence (AI) startups every year through its SAP.iO Foundry in Berlin. SAP.iO is an initiative of the tech giant to create an ecosystem of products and software by working with entrepreneurs. In 2017, five Indian startups were among 10 companies selected by SAP.iO's AI accelerator's first batch. "We are looking at enterprise applications from startups that are founded around a year ago," Alexa Gorman, global vice-president of SAP.iO Fund, told VCCircle over a telephone call from Berlin. "By restricting our focus, we can add value in terms of marketing assistance and mentorship for these startups."

This straightforward order to display pictures of delicious fried confections, spoken into a Google Pixel 2 smartphone with the Google Assistant, is the type of command that users have been executing in Alphabet Inc.'s GOOGL, 1.13% GOOG, 1.00% search engine for years. Behind the scenes, however, the response to this type of query now leverages an enormous amount of machine-learning technology that Google has spent years and billions of dollars developing, in hopes of being a leader in artificial intelligence. For that command to function, software produced by Alphabet-owned Google needed to deploy image content analysis systems, voice recognition and a host of other technologies that revolve around machine learning and AI, mostly pumped through high-tech data centers the company has built. It also decided to make the hardware that runs it, with an eye on pushing the abilities of its services to new places in 2018 and beyond. Since 2013, Alphabet has ramped up its infrastructure spending, pouring $57.36 billion into capital expenditures--roughly $10 billion a year.

For several decades, there has been another face to the field, a technological one that provides tools for solving practical problems in various domains. Of these domains, engineering and medicine have had the closest interaction with AI. AI ideas in representation and reasoning have been especially relevant to diagnosis and corrective action planning in both domains, and in engineering, design has been another area of very fruitful collaboration. These disciplines have not been mere consumers of AI ideas and technology, however. They have had a deep effect on the theoretical side of AI. Complex real-world reasoning tasks, such as those in engineering and medicine, bring out the inadequacy of purely theoretical conceptions about the nature of intelligence. The enormous amounts and types of knowledge often required for carrying out practical reasoning, and the variety of inference techniques that are displayed by practitioners, and the need to arrive at conclusions rapidly--these ...

The central challenge in commonsense knowledge representation research is to develop content theories that achieve a high degree of both competency and coverage. We describe a new methodology for constructing formal theories in commonsense knowledge domains that complements traditional knowledge representation approaches by first addressing issues of coverage. We show how a close examination of a very general task (strategic planning) leads to a catalog of the concepts and facts that must be encoded for general commonsense reasoning. These concepts are sorted into a manageable number of coherent domains, one of which is the representational area of commonsense human memory. We then elaborate on these concepts using textual corpus-analysis techniques, where the conceptual distinctions made in natural language are used to improve the definitions of the concepts that should be expressible in our formal theories.

We describe an approach to intelligent user interfaces, based on the idea of making the computer a collaborator, and an application-independent technology for implementing such interfaces. For us, any interface that is called intelligent should at least be able to answer the six types of questions from users shown in figure 1. Being able to ask and answer these kinds of questions implies a flexible and adaptable division of labor between the human and the computer in the interaction process. Unlike most current interfaces, an intelligent user interface should be able to guide and support you when you make a mistake or if you don't know how to use the system well. What we are suggesting here is a paradigm shift. As an analogy, consider the introduction of the undo button.

One of the most important issues that Congress will face in 2018 is how and when to regulate our growing dependence on artificial intelligence (AI). During the U.S. National Governors Association summer meetings, Elon Musk urged the group to push forward with regulation "before it's too late," stating that AI was an "existential threat to humanity." Hyperbole aside, there are legitimate concerns about the technology and its use. But a rush to regulation could exacerbate current issues, or create new issues that we're not prepared to deal with along the way. To begin with, one of the biggest issues in the world of AI is the lack of clear definition for what the technology is -- and is not.

Primarily, Aboleth is much reduced in scope compared to frameworks like Edward and ZhuSuan; we really only concentrate on neural net-like structures (though you can construct a lot of non-neural net models with these too), and we have only implemented SGVB and "MAP" inference. We have slightly more of a "production" focus than a research focus too (though we would love people to use Aboleth for their research) - for instance we have efficient categorical feature embedding and imputation input layers, where all of the parameters of these layers are learned with the rest of the model. Because we have a limited scope it allows for a nice clean interface and minimal abstraction over pure TensorFlow. Really, all we have done in Aboleth is implement a writer monad, i.e. function composition (the neural net) with an accumulate (the complexity penalties of the layers). This structure is easy to use, extend, and test!

While most of the attention for how artificial intelligence (AI), machine learning and big data can impact companies is focused on the business to consumer (B2C) space, business to business (B2B) companies need to pay attention or they risk their future success. Customers have the same expectations for a simple and easy buying experience whether it's a B2C or B2B interaction. So, if you're in the B2B space, I hope your organization is beginning to explore and plan for, if not already implemented, big data and machine learning to your operations. Since the buying cycle for B2B is usually significantly longer and more complex than B2C (Gartner found there is an average of 5.4 people on corporate buying teams), I would argue that it's even more important for B2Bs to use machines in any way possible to quickly get to know customers and respond to decision-makers with the objective of closing sales as efficiently as possible. According to Salesforce's 2016 Connected Customer report, by 2020, 57% of business buyers will depend on companies to anticipate their needs and if they don't, business buyers will have no problem switching brands.