How LLMs Understand Human Thought Patterns

You know all those arguments that LLMs think like humans? Turns out it's not true 😱 In our new paper we put this to the test by checking if LLMs form concepts the same way humans do. Do LLMs truly grasp concepts and meaning analogously to humans, or is their success primarily rooted in sophisticated statistical pattern matching over vast datasets? We used classic cognitive experiments as benchmarks. What we found is surprising... 🧐 We used seminal datasets from cognitive psychology that mapped how humans actually categorize things like "birds" or "furniture" ('robin' as a typical bird). The nice thing about these datasets is that they are not crowdsourced, they're rigorous scientific benchmarks. We tested 30+ LLMs (BERT, Llama, Gemma, Qwen, etc.) using an information-theoretic framework that measures the trade-off between: - Compression (how efficiently you organize info) - Meaning preservation (how much semantic detail you keep) Finding #1: The Good News LLMs DO form broad conceptual categories that align with humans significantly above chance. Surprisingly (or not?), smaller encoder models like BERT outperformed much larger models. Scale isn't everything! Finding #2: But LLMs struggle with fine-grained semantic distinctions. They can't capture "typicality" - like knowing a robin is a more typical bird than a penguin. Their internal concept structure doesn't match human intuitions about category membership. Finding #3: The Big Difference Here's the kicker: LLMs and humans optimize for completely different things. - LLMs: Aggressive statistical compression (minimize redundancy) - Humans: Adaptive richness (preserve flexibility and context) This explains why LLMs can be simultaneously impressive AND miss obvious human-like reasoning. They're not broken - they're just optimized for pattern matching rather than the rich, contextual understanding humans use. What this means: - Current scaling might not lead to human-like understanding - We need architectures that balance compression with semantic richness - The path to AGI ( 😅 ) might require rethinking optimization objectives Our paper gives tools to measure this compression-meaning trade-off. This could guide future AI development toward more human-aligned conceptual representations. Cool to see cognitive psychology and AI research coming together! Thanks to Chen Shani, Ph.D., who did all the work and Yann LeCun and Dan Jurafsky for their guidance

🧠 LLMs and the Frontier of Understanding: Unveiling a Theory of Mind In a groundbreaking study led by Michal Kosinski at Stanford, the spotlight turns to an intriguing aspect of artificial intelligence: Can large language models (LLMs) understand that others may hold beliefs different from factual reality? This concept, known as the theory of mind, is a fundamental psychological construct that humans typically navigate with ease. The findings? The larger the model, the more adept it becomes at mirroring this uniquely human cognitive ability. 🔗 https://lnkd.in/e4YMMhgH Exploring the Theory of Mind in LLMs: The study meticulously evaluated LLMs (ranging from GPT-1 through GPT-4 and BLOOM) against 40 tasks designed to test human theory of mind capabilities. These tasks, split between "unexpected transfers" and "unexpected content," challenge the models to recognize that characters in the narratives may believe in factually incorrect information. Impressive Outcomes: The leap in performance is striking. While GPT-1, with its 117 million parameters, struggled to grasp these concepts, GPT-4—a behemoth rumored to exceed 1 trillion parameters—solved an impressive 90% of unexpected content tasks and 60% of unexpected transfer tasks. Astonishingly, this surpasses the understanding of 7-year-old children in similar tests. Why This Matters: This research doesn't just push the boundaries of what we expect from AI; it redefines them. By applying tests traditionally used to assess cognitive development in children, we now have a metric to compare aspects of intelligence between humans and deep learning models. This opens a fascinating dialogue on the evolving capabilities of AI and its potential to understand, predict, and interact with human mental states more effectively than ever before. A Thought to Ponder: As AI continues to blur the lines between computational processes and cognitive understanding, we're left to wonder: If an AI model can exhibit a theory of mind, how does that shape our interactions and trust in these systems? Are we ready to engage more deeply with entities that "understand" us in ways we're only beginning to comprehend? #AIResearch #TheoryOfMind #CognitiveAI #StanfordResearch #FutureOfAI

We need to update one of our most widely accepted truisms about #GenAI: “LLMs don’t think - they just predict the next word.” That framing helped ground us and reduce anxiety in the early days of generative AI. But research this year from Anthropic, academic groups, and interpretability labs shows that it’s incomplete. Yes, the training objective for LLMs is next-token prediction, but the way models achieve that is far more complex than we previously understood. Using new techniques, researchers can now trace how models like Claude build internal representations of concepts (from cities to emotions to legal ideas) and plan ahead before generating text. One example: when Claude is asked for “the opposite of small” in English, French, and Chinese, it first activates the same internal concept for “smallness” and its opposite, and only then translates the meaning into each language. In other words, it operates in a language-independent "meaning space" before moving to the translation task. In another example, when prompted to write rhyming poetry, the LLM was revealed to pick the rhyming end-word of a line before generating the sentence that leads to it - it works backwards from the rhyme. This doesn’t mean AI “thinks” like humans. But it definitely means it’s more complex than autocomplete. This is another reminder that our understanding of AI must keep evolving. For the legal industry, it may mean that various policies and risk models need to be adapted to reflect this complexity. The output of LLMs only tells part of the story - understanding how vendors control internal risk and bias in the context of this kind of conceptual modeling requires deeper evaluation. See full article linked in the comments to learn more, including insights into how recent findings on the internal workings of LLM shed light on why certain hallucinations happen. #legaltech #AI #LLMs #law

Did we just get closer to understanding how the brain works? Two groundbreaking papers explore how AI models and the human brain process language, with some interesting implications for text-to-speech. 📌 Google’s Research: Deciphering the human brain with LLM representations The most striking takeaway from Google’s study is that LLMs may process language in ways surprisingly similar to the human brain. By comparing fMRI scans of neural responses to LLM representations, researchers found a fascinating alignment between how the brain’s cortical regions handle language and how LLMs decompose linguistic information. Why does this matter for text-to-speech? Today’s state-of-the-art voice models, which primarily rely on Transformer architectures, excel at producing coherent, fluent speech. However, they often fall short when it comes to replicating the natural prosody, emotional nuance, and contextual awareness that human speech embodies. If model architectures can be refined to reflect the brain’s approach to semantic understanding — particularly how meaning is encoded and represented over time — it could vastly improve the naturalness and expressiveness of AI-generated speech. 📌 Anthropic’s Research: Tracing thoughts in language models Anthropic’s work emphasizes how LLMs break down complex tasks through structured chains of reasoning. The key takeaway? LLMs aren’t just retrieving information; they’re simulating cognitive processes that resemble human-like problem-solving. This process, along with chain-of-thought prompting, allows models to handle intricate tasks by breaking them into manageable steps. The implications for voice AI are profound. Incorporating structured reasoning architectures could allow text-to-speech systems to dynamically adjust prosody, tone, and pacing based on context. For instance, if a model can determine from the conversational structure that a user is expressing frustration or joy, it can modulate the generated speech to mirror that emotional state. It’s about creating models that don’t just speak, but speak with understanding. It's fascinating is that engineering is almost evolving into a science as we build increasingly complex systems that we only partially understand. I think we'll see more use of empirical methods to understand how these systems work, in addition to just building them. At Rime, we’re deeply excited about these findings and are closely monitoring updates. Bridging the gap between neural processes and machine learning architectures will be the key to building voice systems that feel truly human. 🧠💡 Links below... So what’s your take? Are LLMs closer to mimicking the brain than we previously thought?

This is a weekend must-read for those interested in the philosophical nature of LLMs. Anthropic's research team deep dives into the thought process behind LLM answers. What they discover is fascinating: * these models have learned their own method for performing mental math (ex: 36 + 59). It uses parallel threads to estimate the approximate answer (~88-97) and supplements it with what it knows to the correct last digit (5) to arrive at the correct answer. However, in offering its reasoning to a human, it types out the traditionally taught method of carrying the one. Its offered reasoning is different from its actual mental model. * there are times that it offers *motivated reasoning*. When faced with a question it does not know how to answer, there are scenarios where it will come up with an answer then work backwards to provide (incorrect) evidence for that (incorrect) answer. * competing priorities result in conflicting answers. These models have a very strong drive to create complete, grammatically correct answers. They also have strong training to avoid talking about dangerous subjects. When tricked into talking about a bomb, these two directives complete, which result in a complete first sentence about the ingredients required to make a bomb, followed by the model catching itself and refusing to discuss further. Absolutely worth the read. These comics have been making the rounds as some clever users asked ChatGPT to create comics with itself as the protagonist. We're quick to ascribe human elements to models, but this article shows what's actually happening under the surface. https://lnkd.in/gyASKPRg