AI Expert Discusses Early Stages of Intelligence Understanding

✦ Sunday Reads | The Reading Room From the Thinking About Thinking Fellowship Reading Room. This week’s Sunday Read is What Sparseness Means for Brains, AI, and Learning by Prof Daniel Graham. The essay explores the idea of sparseness across neuroscience and artificial intelligence, tracing its influence from early theories of sensory coding through to modern deep learning systems and mixture-of-experts models. Drawing on work by researchers including Horace Barlow, David Field, and Bruno Olshausen, the piece examines why efficient systems, biological or artificial, tend toward sparse representations, and why this may be fundamental to learning itself. Members of the Fellowship can read the full piece inside The Reading Room in our private members hub. To join the Fellowship and access future Sunday Reads, seminars, and discussions, apply below. ↴ https://lnkd.in/eMD7_5Rd Edited by Dr Yukti Chopra #ThinkingAboutThinking #ThAT #SundayReads #AI #Neuroscience

7/8 Megan Peters says science fiction often conflates superintelligence with sentience. The really frightening prospect isn't a conscious AI that hates us. It's a super-capable system, with no experience of its own, and no care for ours. https://lnkd.in/e7aRyJJZ

7/8 Megan Peters says science fiction often conflates superintelligence with sentience. The really frightening prospect isn't a conscious AI that hates us. It's a super-capable system, with no experience of its own, and no care for ours. https://lnkd.in/eAqgm5Gm

Have you ever wondered if we could truly understand thoughts directly from the brain? Imagine a world where AI deciphers what you're thinking, not through your words, but through your neural activity. This isn't science fiction anymore! 🤯 New research is demonstrating a groundbreaking end-to-end NeuroAI pipeline that decodes linguistic features straight from brain signals. Using MEG data and advanced deep learning with the NeuralSet library, we can now transform raw neural activity into meaningful predictions, like estimating word length from brain responses. This is a monumental step in understanding cognitive processes! ✨ This sophisticated framework bridges the gap between raw neuroscience data and powerful machine learning, offering a modular and reusable approach for complex brain decoding tasks. It's building a robust foundation for truly interpretable NeuroAI systems and pushing the boundaries of what's possible in understanding the human mind. 🧠 **Comment "Brain" to get the full article** Learn more about decoding linguistic features from brain signals https://lnkd.in/gQQmtBnF 𝗥𝗲𝗮𝗱𝘆 𝘁𝗼 𝘀𝗲𝗲 𝘄𝗵𝗲𝗿𝗲 𝘆𝗼𝘂𝗿 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝘀𝘁𝗮𝗻𝗱𝘀 𝗶𝗻 𝘁𝗵𝗲 𝗿𝗮𝗽𝗶𝗱𝗹𝘆 𝗲𝘃𝗼𝗹𝘃𝗶𝗻𝗴 𝘄𝗼𝗿𝗹𝗱 𝗼𝗳 𝗔𝗜? 𝗧𝗮𝗸𝗲 𝗼𝘂𝗿 𝗾𝘂𝗶𝗰𝗸 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 𝘁𝗼 𝗯𝗲𝗻𝗰𝗵𝗺𝗮𝗿𝗸 𝘆𝗼𝘂𝗿 𝗔𝗜 𝗿𝗲𝗮𝗱𝗶𝗻𝗲𝗲𝘀𝘀 𝗮𝗻𝗱 𝘂𝗻𝗹𝗼𝗰𝗸 𝘆𝗼𝘂𝗿 𝗽𝗼𝘁𝗲𝗻𝘁𝗶𝗮𝗹! https://lnkd.in/g_dbMPqx #NeuroAI #BrainDecoding #DeepLearning #Neuroscience #HumanMind #SaizenAcuity

The human brain is one of the most efficient learning systems ever known. Billions of neurons continuously exchange electrical and chemical signals, forming patterns that allow us to learn, adapt, recognize faces, process language, and make decisions — often within milliseconds. Artificial Neural Networks were inspired by this biological structure. While today’s AI is still vastly simpler than the human brain, the core idea is similar: interconnected units process information, strengthen important connections, and improve through experience. The fascinating part is not that machines think like humans — they don’t. It’s that studying the brain has helped us build systems capable of learning from data, recognizing patterns, and solving increasingly complex problems. Neuroscience and AI are more connected than many people realize. #ArtificialIntelligence #NeuralNetworks #Neuroscience #MachineLearning #DeepLearning #AI #BrainScience #Technology #DataScience #Innovation

A very interesting article about designing explainable cognitive systems. "Designing explainable cognitive machines with prescribed features requires precise technical definitions for cognition and its constituents, such as memories, behavioural code, learning and thinking, and for explainability itself. To describe cognition, we introduce a cohesive mathematical framework comprising a general model and technical definitions. Our definition of cognition is an interpretation in terms of dynamical systems theory of the existing definition from computational neuroscience, which builds upon an earlier idea of cognition as self-organisation of a vector field. This vector field, serving as both the behavioural code and the substrate for memory imprinting, evolves obeying some learning rules." Link in comment. . . . . . #AI #explainableAI

The most interesting question in AI right now may not be what models can do, but whether doing implies experiencing. This article highlights a remarkable neuroscience study showing that under general anesthesia, the human brain still performs predictive language processing. In other words: computation can continue without conscious awareness. That challenges a convenient assumption in AI discourse, that sophisticated prediction or language behavior signals consciousness. Worth reading if you care about the intersection of neuroscience, philosophy, and AI. https://lnkd.in/eVKjGwRs

What do snow avalanches, earthquakes, forest fires, and your cerebral cortex have in common? They all operate near a critical state, a boundary between rigid order and runaway chaos. In neuroscience, this boundary is captured by a deceptively simple metric: the branching ratio (σ). When σ ≈ 1, each neuron's spike generates, on average, one more spike. Activity reverberates rather than dying out or exploding. And at that exact point, biological neural networks achieve maximal dynamic range, peak information transmission, optimal memory retention, and maximum computational complexity. The latest volume of the Neuraxon Intelligence Academy (Vol. 8) explores the science behind brain criticality and why this same principle, reframed as the "edge of chaos" in AI research, predicts maximal computational capacity in artificial networks too. We dig into the original experiments by Beggs and Plenz, the connection to reservoir computing and spectral radius in recurrent networks, evidence that the brain actually operates slightly below true criticality (and why that matters), and how Neuraxon uses the branching ratio as a real-time operational invariant and self-regulation target. If you're interested in the intersection of neuroscience, physics, and AI architecture, this one goes deep. 📖 Read the full article:

Human Brain vs Convolutional Neural Network (CNN): Beyond Computation The human brain and AI share structural inspiration — neurons and layers — yet they diverge profoundly in meaning. Human Brain: - Learns through emotion, experience, and meaning. - Adapts with remarkable efficiency (~20 W). - Creates knowledge through intuition and imagination. AI (CNN): - Learns through data and optimization. - Requires massive computational power. - Recognizes patterns but lacks understanding. Philosophically, the brain is embodied intelligence, while CNNs are disembodied computation. AI can simulate perception, but it cannot invite meaning. The frontier lies in neuromorphic computing and hybrid AI models — bridging biological cognition and artificial perception. Yet, genuine research remains a human invitation — connecting science to ethics, creativity, and life itself. AI #Neuroscience #Philosophy #AppliedSciences #HumanInvitation #DeepLearning #Research

Headlines claiming scientists can now "read your thoughts" with 90% accuracy are going viral again. The neuroscience research behind this is real and genuinely impressive — but what it actually does versus what the headlines say are two very different things. Worth understanding if you follow AI, neuroscience, or just want to cut through the hype: https://lnkd.in/dwNfTPjJ