For Martin Schrimpf, the promise of artificial intelligence is not in the tasks it can accomplish. It’s in what AI might reveal about human intelligence.

He is working to build a “digital twin” of the brain using artificial neural networks — AI models loosely inspired by how neurons communicate with one another.

That end goal sounds almost ludicrously grand, but his approach is straightforward. First, he and his colleagues test people on tasks related to language or vision. Then they compare the observed behavior or brain activity to results from AI models built to do the same things. Finally, they use the data to fine-tune their models to create increasingly humanlike AI.

An intuitive guide to reasoning in LLM and simple inference time techniques to improve reasoning with code implementations

AI avatars of dead people are teaching courses and testifying in court. Even with the best of intentions, the emerging practice of AI ‘reanimations’ is an ethical quagmire.

Not only can A.I. now make these assessments with remarkable, humanlike accuracy; it can make millions of them in an instant. A.I.’s superpower is its ability to recognize and interpret patterns: to sift through raw data and, by comparing it across vast data sets, to spot trends, relationships and irregularities.

As humans, we constantly generate patterns: in the sequence of our genes, the beating of our hearts, the repetitive motion of our muscles and joints. Everything about us, from the cellular level to the way our bodies move through space, is a source of grist for A.I. to mine. And so it’s no surprise that, as the power of the technology has grown, some of its most startling new abilities lie in its perception of us: our physical forms, our behavior and even our psyches.

Robotics goes mainstream. From humanoids to AVs, Physical AI is poised to reshape labor and create massive new investment opportunities.

Given the recent explosion of large language models (LLMs) that can make convincingly human-like statements, it makes sense that there’s been a deepened focus on developing the models to be able to explain how they make decisions. But how can we be sure that what they’re saying is the truth?

In a new paper, researchers from Microsoft and MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) propose a novel method for measuring LLM explanations with respect to their “faithfulness”—that is, how accurately an explanation represents the reasoning process behind the model’s answer.

As lead author and Ph.D. student Katie Matton explains, faithfulness is no minor concern: if an LLM produces explanations that are plausible but unfaithful, users might develop false confidence in its responses and fail to recognize when recommendations are misaligned with their own values, like avoiding bias in hiring.

The AI Bros, Bruce Burke and John Lawson III, dive into the latest advancements and trends including AI Video Gen, Anthropic’s growth, and OpenAI’s device ambitions.