W hen Lunos, an AI startup in New York City, was gearing up for launch, its founder and chief executive, Duncan Barrigan, and his team wanted to make a splash. So they shelled out $3,500 to do the unconventional: hire a horse and a cowboy to lasso the bull of Wall Street. Wearing ranch gear and a western hat stamped with the Lunos logo, he lassoed the bull's horns as invitees and curious passersby watched. He and the horse then circled the statue, handing out cowboy hats and branded stress balls. The goal was simple: deliver Lunos's pitch of "taming the wild west" of accounts receivables in the most literal, public way possible.

Bayesian optimisation in the latent space of a Variational AutoEncoder (VAE) is a powerful framework for optimisation tasks over complex structured domains, such as the space of scientifically interesting molecules. However, existing approaches tightly couple the surrogate and generative models, which can lead to suboptimal performance when the latent space is not tailored to specific tasks, which in turn has led to the proposal of increasingly sophisticated algorithms. In this work, we explore a new direction, instead proposing a decoupled approach that trains a generative model and a Gaussian Process (GP) surrogate separately, then combines them via a simple yet principled Bayesian update rule. This separation allows each component to focus on its strengths -- structure generation from the VAE and predictive modelling by the GP. We show that our decoupled approach improves our ability to identify high-potential candidates in molecular optimisation problems under constrained evaluation budgets.

In our opinion the exuberance surrounding the relative success of datadriven large language models (LLMs) is slightly misguided and for several reasons (i) LLMs cannot be relied upon for factual information since for LLMs all ingested text (factual or non-factual) was created equal; (ii) due to their subsymbolic nature, whatever'knowledge' these models acquire about language will always be buried in billions of microfeatures (weights), none of which is meaningful on its own; and (iii) LLMs will often fail to make the correct inferences in several linguistic contexts (e.g., nominal compounds, copredication, quantifier scope ambiguities, intensional contexts. Since we believe the relative success of data-driven large language models (LLMs) is not a reflection on the symbolic vs. subsymbolic debate but a reflection on applying the successful strategy of a bottom-up reverse engineering of language at scale, we suggest in this paper applying the effective bottom-up strategy in a symbolic setting resulting in symbolic, explainable, and ontologically grounded language models. Keywords: Bottom-up reverse engineering of language, Symbolic large language models, Language Agnostic Ontology.

A number of recent benchmarks seek to assess how well models handle natural language negation. However, these benchmarks lack the controlled example paradigms that would allow us to infer whether a model had learned how negation morphemes semantically scope. To fill these analytical gaps, we present the Scoped Negation NLI (ScoNe-NLI) benchmark, which contains contrast sets of six examples with up to two negations where either zero, one, or both negative morphemes affect the NLI label. We use ScoNe-NLI to assess fine-tuning and in-context learning strategies. We find that RoBERTa and DeBERTa models solve ScoNe-NLI after many shot fine-tuning. For in-context learning, we test InstructGPT models and find that most prompt strategies are not successful, including those using step-by-step reasoning. To better understand this result, we extend ScoNe with ScoNe-NLG, a sentence completion test set that embeds negation reasoning in short narratives. Here, InstructGPT is successful, which reveals the model can correctly reason about negation, but struggles to do so on prompt-adapted NLI examples outside of its core pretraining regime.

Are you familiar with Herschel Walker's incredible resume? If not, let me tell you, it is truly impressive. From his college football career at the University of Georgia to his professional career in the NFL and USFL, Herschel Walker has proven time and time again that he is a force to be reckoned with on the field. One of his most notable achievements was winning the 1982 Heisman Trophy while playing at the University of Georgia. He set multiple records during his college career and led the Bulldogs to a national championship in 1980. This alone is enough to solidify Herschel's place in college football history.

I was one of those who said it could never be done: that a computer wouldn't ever manage to beat the best at the game of poker. I was romantic and wide-eyed at 18, when my heroes were the cowboys from Texas who ruled the felt. They were uneducated and coarse, yet chock full of the human qualities needed to excel at poker. With nicknames like Amarillo Slim and Texas Dolly, these larger-than-life characters had fearlessness, aptitude, and a deep understanding of what makes people tick. The higher the stakes, the better they played.

"This story line will make Hieronymus Bosch look like he was doodling kittens," Lee Sizemore brags. He's the head of the "narrative department" at Westworld, a frontier-themed vacation park where customers act out their darkest fantasies. A special little something I call the ourobouros." Self-cannibalism and the snake that eats its own tail: that's a fair description of "Westworld," a come-hither drama that introduces itself as a science-fiction thriller about cyborgs who become self-aware, then reveals its true identity as what happens when an HBO drama struggles to do the same. Created by the husband-and-wife team of Jonathan Nolan and Lisa Joy, "Westworld" is explicitly, and often wittily, an exploitation series about exploitation, full of naked bodies that are meant to make us think about nudity and violence that comments on violence.

It has a simple design--a silver box set atop four long legs with wheels for feet. It is battery operated and has a camera for remote operation. For now, it has two main jobs: handling animals and hauling things around. SwagBot has a hook on the back that allows it to pull around other equipment, which means it can be used to move feed from the barn to the field, for example. But it is clear that the primary purpose of the robot is to monitor and work animals such as horses and cows in a manner similar to a herding dog.

Mezzo is a computer program designed that procedurally writes Romantic-Era style music in real-time to accompany computer games. Leitmotivs are associated with game characters and elements, and mapped into various musical forms.  These forms are distinguished by different amounts of harmonic tension and formal regularity, which lets them musically convey various states of markedness which correspond to states in the game story. Because the program is not currently attached to any game or game engine, “virtual” gameplays were been used to explore the capabilities of the program; that is, videos of various game traces were used as proxy examples.  For each game trace, Leitmotivs were input to be associated with characters and game elements, and a set of ‘cues’ was written, consisting of a set of time points at which a new set of game data would be passed to Mezzo to reflect the action of the game trace.  Examples of music composed for one such game trace, a scene from Red Dead Redemption , are given to illustrate the various ways the program maps Leitmotivs into different levels of musical markedness that correspond with the game state.