Amazon Alexa: "Sorry to hear that. Talking to a friend, listening to music or taking a walk might help. I hope you feel better soon." Alexa's artificial intelligence-infused heart may be in the right place, but there's only so far it or any AI can go to comfort someone who is alone. All the same, Alexa's response raises questions about just what kind of role an AI can play to "cure" loneliness, especially among the elderly.

Increasing interest among students in artificial intelligence has prompted administrators in the Computer Science department and the School of Engineering and Applied Sciences more broadly to grow its AI program. Over the past 10 years, enrollment in introductory artificial intelligence course COMPSCI 181: "Machine Learning" has more than quadrupled from 35 students in 2009 to more than 150 in recent years. Computer Science Area Co-Chair Edward W. Kohler said the growing student demand has led administrators to focus not only on supporting existing faculty members who teach courses in artificial intelligence, but also on recruiting additional faculty members. "I would say that we come close, but we are not able to meet the demand," Kohler said. Roughly 20 faculty members across SEAS and related departments in the Faculty of Arts and Sciences currently specialize in artificial intelligence -- a number that continues to grow with recent and upcoming faculty appointments.This fall, SEAS welcomed Computer Science Professor Milind Tambe, whose work aims to tackle societal problems -- such as wildlife conservation -- using artificial intelligence.

Are the Liberal Arts Degrees Useless?

Design and create video games using Construct 2. No prior experience is required.

Google two years ago launched Teachable Machine, a web experiment intended to elucidate machine learning concepts. It let any user with a webcam train an AI model to output specific media -- an image, sound, speech, or GIF -- corresponding with a hand gesture, object, or activity. Now Teachable Machine is expanding to incorporate inputs beyond those it initially supported, including audio. Additionally, it will allow folks to export their trained models to websites, apps, devices, and more. Google says it worked with people across industries with different needs -- like architect Steve Saling, who has amyotrophic lateral sclerosis (ALS) -- to test and shape the new Teachable Machine.

This is most apparent when auto-encoders are trained, where a network is trained to map the input data upon itself but is forced to project them into a lower-dimensional embedding space on the way (Vincent et al., 2010). In addition to the conventional fully connected layers, there are various special types of network connections. For example, in computer vision, convolu-tional layers are commonly used, which train multiple sliding windows that move over the image data and process just a part of the image at a time, thereby learning to recognize local features. These layers are subsequently abstracted into more and more complex visual patterns (Krizhevsky et al., 2017). For temporal data, one can use recurrent neural networks, which do not make predictions for individual input vectors, but for a sequence of input vectors. To do so, they allow feeding abstracted information from previous data points forward to the next layers.

This paper proposes a step toward obtaining general models of knowledge for facial analysis, by addressing the question of multi-source transfer learning. More precisely, the proposed approach consists in two successive training steps: the first one consists in applying a combination operator to define a common embedding for the multiple sources materialized by different existing trained models. The proposed operator relies on an auto-encoder, trained on a large dataset, efficient both in terms of compression ratio and transfer learning performance. In a second step we exploit a distillation approach to obtain a lightweight student model mimicking the collection of the fused existing models. This model outperforms its teacher on novel tasks, achieving results on par with state-of-the-art methods on 15 facial analysis tasks (and domains), at an affordable training cost. Moreover, this student has 75 times less parameters than the original teacher and can be applied to a variety of novel face-related tasks.

Autonomous reinforcement learning agents, like children, do not have access to predefined goals and reward functions. They must discover potential goals, learn their own reward functions and engage in their own learning trajectory. Children, however, benefit from exposure to language, helping to organize and mediate their thought. We propose LE2 (Language Enhanced Exploration), a learning algorithm leveraging intrinsic motivations and natural language (NL) interactions with a descriptive social partner (SP). Using NL descriptions from the SP, it can learn an NL-conditioned reward function to formulate goals for intrinsically motivated goal exploration and learn a goal-conditioned policy. By exploring, collecting descriptions from the SP and jointly learning the reward function and the policy, the agent grounds NL descriptions into real behavioral goals. From simple goals discovered early to more complex goals discovered by experimenting on simpler ones, our agent autonomously builds its own behavioral repertoire. This naturally occurring curriculum is supplemented by an active learning curriculum resulting from the agent's intrinsic motivations. Experiments are presented with a simulated robotic arm that interacts with several objects including tools.

We consider the fundamental learning problem of estimating properties of distributions over large domains. Using a novel piecewise-polynomial approximation technique, we derive the first unified methodology for constructing sample- and time-efficient estimators for all sufficiently smooth, symmetric and non-symmetric, additive properties. This technique yields near-linear-time computable estimators whose approximation values are asymptotically optimal and highly-concentrated, resulting in the first: 1) estimators achieving the $\mathcal{O}(k/(\varepsilon^2\log k))$ min-max $\varepsilon$-error sample complexity for all $k$-symbol Lipschitz properties; 2) unified near-optimal differentially private estimators for a variety of properties; 3) unified estimator achieving optimal bias and near-optimal variance for five important properties; 4) near-optimal sample-complexity estimators for several important symmetric properties over both domain sizes and confidence levels. In addition, we establish a McDiarmid's inequality under Poisson sampling, which is of independent interest.

Today, part of the continuing education required of marketers and others in business careers is learning about artificial intelligence. After all, it's now nearly ubiquitous in software, smartphones, and online, and business tech without built-in AI has serious shortcomings compared to that with it. One powerful type of AI is natural language processing, or NLP. This process teaches computers to "use language like people" and is built into a variety of business tech, including chatbots, CRM platforms, and social media monitoring tools. And while it has a number of uses, some of the best-known reasons companies should invest in NLP-powered tech include improved customer service, automated writing corrections, and analysis of customer intent and sentiment.