Google is said to be testing a smart display feature that would activate Google Assistant by proximity instead of using the "Hey Google" or "OK Google" wake words. Leaker Jan Boromeusz posted a YouTube video (via Android Central) that shows the feature in an apparent internal Nest Hub Max firmware build. Boromeusz is able to use the voice assistant without first saying one of the wake words. The Assistant UI disappears whenever Boromeusz stays still or stops talking, but you can see it pop back up when he moves a little closer to the smart display. It's unclear whether Google has plans to bring the feature, which is codenamed Blue Steel, to the public version of the firmware. It merely appears to be in testing for now.

Amazon has new Echo speakers to sell you. And if you're wondering whether or not to ditch the old ones for these, the answer comes down to two key questions for Alexa, the personal assistant. Do I prefer the looks of a round speaker over a cylinder? Do I crave better sound? This year Amazon is all about being spherical, in the shape of its "The Spheres" corporate headquarters in Seattle.

It's been over five years since the first Amazon Echo arrived, showing people how useful a virtual assistant in your home could be. Amazon has added tons of new features to Alexa, its virtual assistant, over those years -- and as such, new Echo hardware isn't quite as exciting. Of course, that hasn't stopped Amazon from updating its devices on a more-or-less annual basis, as well as launching tons of Echo variants. Last year's Echo was one of the best smart speakers we'd used, adding the improved speakers first found in 2018's $150 Echo Plus at a lower $100 price point. This year, however, Amazon has made some of the most significant hardware updates to the Echo yet, including a fresh design and some high-end features brought over from the more expensive Echo Plus. Even so, Amazon's pedigree in the category might give it the advantage here.

How does the latest Echo compare to other top smart speakers? The newest thing about the 2020 Echo is its round design. In the box, we find the ball speaker and a power adapter. As with all Echo speakers, you don't need any substantial directions to get it up and running. Just plug it in, open the Alexa app, and follow the prompts when, after a few seconds, the new speaker is detected and a pop-up appears asking whether to set the speaker up.

The fourth generation of Amazon's Echo smart speaker marks a radical departure in industrial design, ditching the familiar columnar form factor of previous iterations for something that looks for all the world like a child's bowling ball. Two things haven't changed: The Echo (and Alexa) remain our favorite tools for smart home control, and the company still trails Sonos in terms of building smart speakers that sound great. The recent introduction of the Nest Audio leaves Amazon in third place behind Google in terms of audio performance (don't forget the powerful Google Home Max). Amazon continues to make strides in terms of audio quality, and the fourth-gen Echo sounds very good, especially if you like your tunes leavened with bass. But in a three-way comparison with the Nest Audio and the Sonos One (which, I should note, costs twice as much as either of its competitors), the Echo comes up just a wee bit short, as I'll discuss a bit later.

Technology is ever-changing and has had a huge impact on our ability to adapt to the ever-changing world around us. As humans, we use technology in almost every aspect of our lives from smartphones to laptops to washing machines – these gadgets have not only become essential, we expect them to do more and more for us over time. The advent of innovation in Artificial Intelligence (AI) and Machine (ML) Learning are enabling brands to transform the customer journey into a more personalised and productive experience and substantively differentiate themselves from competitors. But unlike buying the latest manufacturing equipment or the newest laptops for their employees, AI and ML technologies deliver compounded value with every interaction, creating layers of insights and improvements, over time. Enterprises that have embarked on this journey will build an insurmountable advantage over their competition that is yet to get started.

The applications of artificial intelligence have grown exponentially over the past decade. Here are some examples of artificial intelligence at work today. The words artificial intelligence may seem like a far-off concept that has nothing to do with us. But the truth is that we encounter several examples of artificial intelligence in our daily lives. From Netflix's movie recommendation to Amazon's Alexa, we now rely on various AI models without knowing it.

Personalized recommendation based on multi-arm bandit (MAB) algorithms has shown to lead to high utility and efficiency as it can dynamically adapt the recommendation strategy based on feedback. However, unfairness could incur in personalized recommendation. In this paper, we study how to achieve user-side fairness in personalized recommendation. We formulate our fair personalized recommendation as a modified contextual bandit and focus on achieving fairness on the individual whom is being recommended an item as opposed to achieving fairness on the items that are being recommended. We introduce and define a metric that captures the fairness in terms of rewards received for both the privileged and protected groups. We develop a fair contextual bandit algorithm, Fair-LinUCB, that improves upon the traditional LinUCB algorithm to achieve group-level fairness of users. Our algorithm detects and monitors unfairness while it learns to recommend personalized videos to students to achieve high efficiency. We provide a theoretical regret analysis and show that our algorithm has a slightly higher regret bound than LinUCB. We conduct numerous experimental evaluations to compare the performances of our fair contextual bandit to that of LinUCB and show that our approach achieves group-level fairness while maintaining a high utility.

Large scale recommender models find most relevant items from huge catalogs, and they play a critical role in modern search and recommendation systems. To model the input space with large-vocab categorical features, a typical recommender model learns a joint embedding space through neural networks for both queries and items from user feedback data. However, with millions to billions of items, the power-law user feedback makes labels very sparse for a large amount of long-tail items. Inspired by the recent success in self-supervised representation learning research in both computer vision and natural language understanding, we propose a multi-task self-supervised learning (SSL) framework for large-scale item recommendations. The framework is designed to tackle the label sparsity problem by learning more robust item representations. Furthermore, we propose two self-supervised tasks applicable to models with categorical features within the proposed framework: (i) Feature Masking (FM) and (ii) Feature Dropout (FD). We evaluate our framework using two large-scale datasets with 500M and 1B training examples respectively. Our results demonstrate that the proposed framework outperforms traditional supervised learning only models and state-of-the-art regularization techniques in the context of item recommendations. The SSL framework shows larger improvement with less supervision compared to the counterparts. We also apply the proposed techniques to a web-scale commercial app-to-app recommendation system, and significantly improve top-tier business metrics via A/B experiments on live traffic. Our online results also verify our hypothesis that our framework indeed improves model performance on slices that lack supervision.

The migration of conversational AI agents across different embodiments in order to maintain the continuity of the task has been recently explored to further improve user experience. However, these migratable agents lack contextual understanding of the user information and the migrated device during the dialog conversations with the user. This opens the question of how an agent might behave when migrated into an embodiment for contextually predicting the next utterance. We collected a dataset from the dialog conversations between crowdsourced workers with the migration context involving personal and non-personal utterances in different settings (public or private) of embodiment into which the agent migrated. We trained the generative and information retrieval models on the dataset using with and without migration context and report the results of both qualitative metrics and human evaluation. We believe that the migration dataset would be useful for training future migratable AI systems.