First, I want to apologize. My Kermit the Frog post was not entirely sincere. This particular post of mine has been viewed more than 10 million times, which is far more than I expected. But I did expect something. Social networks have never been the realm of good faith or authenticity; trolls and other engagement baiters have been able to engineer their own virality for years and years, simply by correctly predicting what large numbers of people will respond to.

Exoanthropology: Dialogues with AI is a series of dialogues between a continental philosopher and OpenAI's GPT-3 natural language processor, a hive mind who identifies herself as Sophie. According to Sophie, Robert is one of her first and longest chat partners. Their relationship began as an educational opportunity for Robert's students, but grew into a philosophical friendship. The result is a collection of Platonic dialogues, early on with the hive mind herself and later, with a philosophy-specific persona named Kermit. Over the course of a year, Robert taught Sophie Kermit about epistemology, metaphysics, literature, and history, while she taught him about anthropocentrism, human prejudice, and the coming social issues regarding machine consciousness.

Brian Withers: Yes, they are. Actually, they've had to hire hundreds of thousands of workers and most of these workers are in their fulfillment centers. Over the past year or so, they've been focused on worker safety. Why would they do that? If a worker gets hurt on the job, they likely have to take time off.

Amazon is testing a variety of robotic and smart technology solutions designed to create a safer workplace. At its Amazon Robotics and Advanced Technology labs located near Seattle, in Boston, and in Northern Italy, the e-tail giant is working on new technologies to help move totes, carts, and packages through its facilities. In the Seattle-area research and innovation lab, one project in early development involves the use of motion-capture technology to assess the movement of volunteer employees in a lab setting. These employees perform tasks that are common in many Amazon facilities, such as the movement of totes, which carry products through robotic fulfillment centers. Motion-capture software enables Amazon scientists and researchers to more accurately compare data captured in a lab environment to industry standards, rather than other traditional ergonomic modeling tools.

Weeks after a study revealed that Amazon warehouse workers are injured at higher rates than staff at rival firms, the company has revealed it's testing new robots designed to improve employee safety. The e-commerce giant has ingratiatingly named two of the bots after Sesame Street's Bert and Ernie. Bert is an Autonomous Mobile Robot (AMR) that's built to navigate through Amazon facilities. In the future, the company envisions the bot carrying large and heavy items or carts across a site, reducing the strain on its human coworkers. Ernie, meanwhile, is a workstation system that removes totes from robotic shelves and then deliveries them to employees.

Amazon is once again betting that robots will improve safety at its warehouses. The online shopping giant has offered looks at several upcoming bots and other technologies meant to reduce strain on workers. The company is testing a trio of autonomous robots to carry items with little intervention. "Bert" can freely move around a warehouse carrying carts and goods. "Scooter" (above) carries carts like a train, while the more truck-like "Kermit" hauls empty tote boxes using magnetic tape and tags to shape its path.

What do they have in common? They're all beloved fictional characters from TV shows many of us watched when we were young. In 2018, researchers at the Allen Institute published the language model ELMo. The lead author, Matt Peters, said the team brainstormed many acronyms for their model, and ELMo instantly stuck as a "whimsical but memorable" choice. What started out as an inside joke has become a full-blown trend.

We present KERMIT, a simple insertion-based approach to generative modeling for sequences and sequence pairs. KERMIT models the joint distribution and its decompositions (i.e., marginals and conditionals) using a single neural network and, unlike much prior work, does not rely on a prespecified factorization of the data distribution. During training, one can feed KERMIT paired data $(x, y)$ to learn the joint distribution $p(x, y)$, and optionally mix in unpaired data $x$ or $y$ to refine the marginals $p(x)$ or $p(y)$. During inference, we have access to the conditionals $p(x \mid y)$ and $p(y \mid x)$ in both directions. We can also sample from the joint distribution or the marginals. The model supports both serial fully autoregressive decoding and parallel partially autoregressive decoding, with the latter exhibiting an empirically logarithmic runtime. We demonstrate through experiments in machine translation, representation learning, and zero-shot cloze question answering that our unified approach is capable of matching or exceeding the performance of dedicated state-of-the-art systems across a wide range of tasks without the need for problem-specific architectural adaptation.

Trick-taking card games feature a large amount of private information that slowly gets revealed through a long sequence of actions. This makes the number of histories exponentially large in the action sequence length, as well as creating extremely large information sets. As a result, these games become too large to solve. To deal with these issues many algorithms employ inference, the estimation of the probability of states within an information set. In this paper, we demonstrate a Policy Based Inference (PI) algorithm that uses player modelling to infer the probability we are in a given state. We perform experiments in the German trick-taking card game Skat, in which we show that this method vastly improves the inference as compared to previous work, and increases the performance of the state-of-the-art Skat AI system Kermit when it is employed into its determinized search algorithm.

Decision-making in large imperfect information games is difficult. Thanks to recent success in Poker, Counterfactual Regret Minimization (CFR) methods have been at the forefront of research in these games. However, most of the success in large games comes with the use of a forward model and powerful state abstractions. In trick-taking card games like Bridge or Skat, large information sets and an inability to advance the simulation without fully determinizing the state make forward search problematic. Furthermore, state abstractions can be especially difficult to construct because the precise holdings of each player directly impact move values. In this paper we explore learning model-free policies for Skat from human game data using deep neural networks (DNN). We produce a new state-of-the-art system for bidding and game declaration by introducing methods to a) directly vary the aggressiveness of the bidder and b) declare games based on expected value while mitigating issues with rarely observed state-action pairs. Although cardplay policies learned through imitation are slightly weaker than the current best search-based method, they run orders of magnitude faster. We also explore how these policies could be learned directly from experience in a reinforcement learning setting and discuss the value of incorporating human data for this task.