Despite widespread use of LLMs as conversational agents, evaluations of performance fail to capture a crucial aspect of communication: interpreting language in context---incorporating its pragmatics. Humans interpret language using beliefs and prior knowledge about the world. For example, we intuitively understand the response I wore gloves to the question Did you leave fingerprints?

Despite widespread use of LLMs as conversational agents, evaluations of performance fail to capture a crucial aspect of communication: interpreting language in context---incorporating its pragmatics. Humans interpret language using beliefs and prior knowledge about the world. For example, we intuitively understand the response "I wore gloves" to the question "Did you leave fingerprints?" as meaning "No". To investigate whether LLMs have the ability to make this type of inference, known as an implicature, we design a simple task and evaluate four categories of widely used state-of-the-art models. We find that, despite only evaluating on utterances that require a binary inference (yes or no), models in three of these categories perform close to random.

Catastrophic forgetting poses a significant challenge in continual learning, where models often forget previous tasks when trained on new data. Our empirical analysis reveals a strong correlation between catastrophic forgetting and the learning speed of examples: examples learned early are rarely forgotten, while those learned later are more susceptible to forgetting. We demonstrate that replay-based continual learning methods can leverage this phenomenon by focusing on mid-learned examples for rehearsal. We introduce Goldilocks, a novel replay buffer sampling method that filters out examples learned too quickly or too slowly, keeping those learned at an intermediate speed. Goldilocks improves existing continual learning algorithms, leading to state-of-the-art performance across several image classification tasks.

Training deep neural networks is an important problem which is still far from solved. At the core of the problem is our still relatively poor understanding of what happens under the hood of a deep neural network. Practically, this translates to a wide variety of deep network architectures and activation functions used in them. They all, however, suffer from the same problem when it comes to interpretability. It is next to impossible to understand how and why even a single layer network performs a simple classification task, and this probelm only increases with the size and the depth of the network. Activation functions stem from Cybenko's seminal 1989 paper [1], which proved that sigmoidal functions are universal approximators. This gave rise to a number of sigmoidal activation functions, including the sigmoid, tanh, arctan, binary step, Elliott sign [2], SoftSign [3] [4], SQNL [5], soft clipping [6] and many others. Sigmoidal activations were useful in the early days of neural networks, but the most serious problem that they suffered from was vanishing gradients.

Someone said to an editor: "why not have a regular lighthearted column on AI topics?" The editor said: "what an excellent idea, and when will we get the first manuscript?" We go to press next week." While looking for something to give him, we stumbled on this old manuscript, written years ago (with our esteemed colleague Neil Agnew, the Duke of York). Ever had an old sock that you try to throw away, but keep finding in the bottom of a drawer?

Wars over office temperature may be coming to a thaw. Thanks to advances in workplace architecture and new sensor and app technologies, individual workers are getting more control over the climate around them, which has long been a battleground for office workers. Some of the new technologies seem straight out of science fiction. One building under renovation in Italy is going to provide workers with their own "thermal bubbles" that can follow them around the building, so workers will each have their own climate-controlled zone. Elsewhere, smartphone apps such as Comfy let workers order a 10-minute blast of hot or cold air.

This column, like many strange things in the modern world, was conceived in an email exchange. Someone said to an editor: "why not have a regular lighthearted column on AI topics?" The editor said: "what an excellent idea, and when will we get the first manuscript?" and the first person said: "oh but I didn't volunteer;" and the editor said: "listen, buddy, I can make your life very uncomfortable if I don't get some cooperation. We go to press next week." While looking for something to give him, we stumbled on this old manuscript, written years ago (with our esteemed colleague Neil Agnew, the Duke of York). Ever had an old sock that you try to throw away, but keep finding in the bottom of a drawer? This is a bit like that. Come to think of it, so is the frame problem. Anyway, you can't make an omelette without breaking eggs, so here is our first reflection. It's a variation on an old, old story ....