SAP SE (NYSE: SAP) today said it will offer a new course on the openSAP platform that focuses on the ethical implications when developing and interacting with artificial intelligence (AI). Creating Trustworthy and Ethical Artificial Intelligence, offered June 25 through July 24, is geared toward all leaders, professionals, developers and general users of AI technology. "The potential for AI and machine learning is great, and the technology has already had significant impact in automating tasks and efficiently analyzing data sets," said Bernd Welz, chief knowledge officer, SAP. "As this technology continues to evolve and becomes further engrained in our society, it's important that we take the necessary steps to ensure that its development and continued application are carried out in an ethical way. Through this course, we're showing learners how they can keep ethics in the forefront when developing AI- and machine learning–enabled technologies."

What goes into making plants taste good? For scientists in MIT's Media Lab, it takes a combination of botany, machine-learning algorithms, and some good old-fashioned chemistry. Using all of the above, researchers in the Media Lab's Open Agriculture Initiative report that they have created basil plants that are likely more delicious than any you have ever tasted. No genetic modification is involved: The researchers used computer algorithms to determine the optimal growing conditions to maximize the concentration of flavorful molecules known as volatile compounds. But that is just the beginning for the new field of "cyber agriculture," says Caleb Harper, a principal research scientist in MIT's Media Lab and director of the OpenAg group.

If you want to keep up with the latest e-learning trends, then you must be aware that Artificial Intelligence has a great potential in the LMS world and can help to add never before seen value for the e-learning team. It is true that online training and learning needs a lot of management and proper system to deliver what the learners actually need and also in the form they need. So, the need for the moment is a new type of LMS that supports AI, which will help to put the learners in the center and help them better understand and also manage their courses. AI can bring a great change in the learning system and make it more effective and aligned with the needs. "Learning is a two-way process- A dialogue. With latest technology and social media, learners expect fast and quick response to their questions."

Nisha Rataria remembers the moment that she understood the power of technology to significantly improve a child's learning and comprehension. As a teacher at the public Vidhya Nagar Primary School in Ahmedabad, Gujarat, India, Rataria teaches students from across the spectrum – bright, struggling, poor and middle class. A few years ago, her school implemented an artificial-intelligence based education program called EnglishHelper that provides a suite of tools to help children learn to speak, read and write English. Many of her students, who she says could not even recognize the alphabet, could now read English with some confidence. By the end of the 2019-2020 school year, EnglishHelper and ReadToMe could be used by nearly 20 million students worldwide.

It's in our nature to be addicted to things that distract us

From this academic year (2019-20), aspiring engineering candidates in Karnataka will have the opportunity to study the most in-demand courses - Artificial Intelligence (AI) and Machine Learning (ML). The Visvesvaraya Technological University (VTU) in its recent Executive Council meeting on May 30 resolved to introduce Bachelor of Engineering (BE) in Artificial Intelligence (AI) and Machine Learning with effect from the academic year 2019-20. The eligibility for admission to this course remains the same as BE and B.Tech programs in VTU. As of now, several colleges have modules in Machine Vision, Robot Programming and Artificial Intelligence in the third semester of the Instrumentation engineering course. But, offering it as degree course in itself is a first in VTU.

Automatic program correction is an active topic of research, which holds the potential of dramatically improving productivity of programmers during the software development process and correctness of software in general. Recent advances in machine learning, deep learning and NLP have rekindled the hope to eventually fully automate the process of repairing programs. A key challenges is ambiguity, as multiple codes -- or fixes -- can implement the same functionality. In addition, dataset by nature fail to capture the variance introduced by such ambiguities. Therefore, we propose a deep generative model to automatically correct programming errors by learning a distribution of potential fixes. Our model is formulated as a deep conditional variational autoencoder that samples diverse fixes for the given erroneous programs. In order to account for ambiguity and inherent lack of representative datasets, we propose a novel regularizer to encourage the model to generate diverse fixes. Our evaluations on common programming errors show for the first time the generation of diverse fixes and strong improvements over the state-of-the-art approaches by fixing up to $61\%$ of the mistakes.

An increasing number of well-trained deep networks have been released online by researchers and developers, enabling the community to reuse them in a plug-and-play way without accessing the training annotations. However, due to the large number of network variants, such public-available trained models are often of different architectures, each of which being tailored for a specific task or dataset. In this paper, we study a deep-model reusing task, where we are given as input pre-trained networks of heterogeneous architectures specializing in distinct tasks, as teacher models. We aim to learn a multitalented and light-weight student model that is able to grasp the integrated knowledge from all such heterogeneous-structure teachers, again without accessing any human annotation. To this end, we propose a common feature learning scheme, in which the features of all teachers are transformed into a common space and the student is enforced to imitate them all so as to amalgamate the intact knowledge. We test the proposed approach on a list of benchmarks and demonstrate that the learned student is able to achieve very promising performance, superior to those of the teachers in their specialized tasks.

We present a deep neural-network model for lifelong learning inspired by several forms of neuroplasticity. The neural network develops continuously in response to signals from the environment. In the beginning, the network is a blank slate with no nodes at all. It develops according to four rules: (i) expansion, which adds new nodes to memorize new input combinations; (ii) generalization, which adds new nodes that generalize from existing ones; (iii) forgetting, which removes nodes that are of relatively little use; and (iv) backpropagation, which fine-tunes the network parameters. We analyze the model from the perspective of accuracy, energy efficiency, and versatility and compare it to other network models, finding better performance in several cases.

In classical Q-learning, the objective is to maximize the sum of discounted rewards through iteratively using the Bellman equation as an update, in an attempt to estimate the action value function of the optimal policy. Conventionally, the loss function is defined as the temporal difference between the action value and the expected (discounted) reward, however it focuses solely on the future, leading to overestimation errors. We extend the well-established Q-learning techniques by introducing the hindsight factor, an additional loss term that takes into account how the model progresses, by integrating the historic temporal difference as part of the reward. The effect of this modification is examined in a deterministic continuous-state space function estimation problem, where the overestimation phenomenon is significantly reduced and results in improved stability. The underlying effect of the hindsight factor is modeled as an adaptive learning rate, which unlike existing adaptive optimizers, takes into account the previously estimated action value. The proposed method outperforms variations of Q-learning, with an overall higher average reward and lower action values, which supports the deterministic evaluation, and proves that the hindsight factor contributes to lower overestimation errors. The mean average score of 100 episodes obtained after training for 10 million frames shows that the hindsight factor outperforms deep Q-networks, double deep Q-networks and dueling networks for a variety of ATARI games.