This article was originally published by Quanta Magazine. A picture may be worth a thousand words, but how many numbers is a word worth? The question may sound silly, but it happens to be the foundation that underlies large language models, or LLMs--and through them, many modern applications of artificial intelligence. Every LLM has its own answer. In Meta's open-source Llama 3 model, words are split into tokens represented by 4,096 numbers; for one version of GPT-3, it's 12,288.

This paper examines the question of whether Large Language Models (LLMs) like ChatGPT possess minds, focusing specifically on whether they have a genuine folk psychology encompassing beliefs, desires, and intentions. We approach this question by investigating two key aspects: internal representations and dispositions to act. First, we survey various philosophical theories of representation, including informational, causal, structural, and teleosemantic accounts, arguing that LLMs satisfy key conditions proposed by each. We draw on recent interpretability research in machine learning to support these claims. Second, we explore whether LLMs exhibit robust dispositions to perform actions, a necessary component of folk psychology. We consider two prominent philosophical traditions, interpretationism and representationalism, to assess LLM action dispositions. While we find evidence suggesting LLMs may satisfy some criteria for having a mind, particularly in game-theoretic environments, we conclude that the data remains inconclusive. Additionally, we reply to several skeptical challenges to LLM folk psychology, including issues of sensory grounding, the "stochastic parrots" argument, and concerns about memorization. Our paper has three main upshots. First, LLMs do have robust internal representations. Second, there is an open question to answer about whether LLMs have robust action dispositions. Third, existing skeptical challenges to LLM representation do not survive philosophical scrutiny.

This work explores whether language models encode meaningfully grounded representations of sounds of objects. We learn a linear probe that retrieves the correct text representation of an object given a snippet of audio related to that object, where the sound representation is given by a pretrained audio model. This probe is trained via a contrastive loss that pushes the language representations and sound representations of an object to be close to one another. After training, the probe is tested on its ability to generalize to objects that were not seen during training. Across different language models and audio models, we find that the probe generalization is above chance in many cases, indicating that despite being trained only on raw text, language models encode grounded knowledge of sounds for some objects.

Cutting-edge NASA imaging technology can detect early signs of a plant virus that, if unaddressed, often proves devastating for wineries and grape growers, new research has found. While the breakthrough is good news for the wine and grape industry, which loses billions of dollars a year to the crop-ruining disease, it could eventually help global agriculture as a whole. Using intricate infrared images captured by airplane over California's Central Valley, researchers were able to distinguish Cabernet Sauvignon grape vines that were infected but not showing symptoms -- before the point at which growers can spot the disease and respond. The technology, coupled with machine learning and on-the-ground analysis, successfully identified infected plants with almost 90% accuracy in some cases, according to two new research papers. "This is the first time we've ever shown the ability to do viral disease detection on the airborne scale," said Katie Gold, an assistant professor of grape pathology at Cornell University and a lead researcher on the project.

The widespread public deployment of large language models (LLMs) in recent months has prompted a wave of new attention and engagement from advocates, policymakers, and scholars from many fields. This attention is a timely response to the many urgent questions that this technology raises, but it can sometimes miss important considerations. This paper surveys the evidence for eight potentially surprising such points: 1. LLMs predictably get more capable with increasing investment, even without targeted innovation. 2. Many important LLM behaviors emerge unpredictably as a byproduct of increasing investment. 3. LLMs often appear to learn and use representations of the outside world. 4. There are no reliable techniques for steering the behavior of LLMs. 5. Experts are not yet able to interpret the inner workings of LLMs. 6. Human performance on a task isn't an upper bound on LLM performance. 7. LLMs need not express the values of their creators nor the values encoded in web text. 8. Brief interactions with LLMs are often misleading.

Shootings are an epidemic in the US, but federal funding for research into gun violence has been in a deep freeze since 1996, thanks in part to the NRA-backed Dickey Amendment, which prevents the Center for Disease Control from pursuing research "to advocate or promote gun control." Basically, humans can't get money to research the problem of gun violence in the US. To get around this, some scientists want machines to do the job. On September 25, University of Pennsylvania computer scientists Ellie Pavlick and Chris Callison-Burch unveiled a new, human-annotated database of gun violence incidents in the US at the Bloomberg Data for Good Exchange Conference in New York. The database was created by workers on Amazon's Mechanical Turk platform, and carefully highlights information from thousands of news articles over the course of several years, Pavlick told me in an interview.