Artificial Intelligence in Cognitive Neuroscience

Collaborative innovation combining AI with neuropsychology is proving to be transformative. Six research clusters show specific value and potential: 🌱 Neuroscience and Mental Health: Understanding mental health through neuroimaging and machine learning enables earlier, more precise interventions for conditions like ADHD and depression. By examining correlations in brain function, this research helps identify key markers for cognitive impairments, aiding in early diagnosis and personalized treatment plans. 🔍 Computational Modeling: Computational models simulate decision-making and cognitive markers, which are crucial for neurological conditions like epilepsy. Machine learning applied to seizure detection, for instance, offers a potential breakthrough in predicting and managing epilepsy, helping patients gain better control and care. 🧠 Cognitive Neuroscience: Studies of cognitive decline and neurodegenerative diseases, such as Alzheimer’s, benefit from reinforcement learning models that reveal patterns in brain degeneration. These insights are essential for developing strategies to slow disease progression, offering hope for more effective interventions. 💡 Cognitive Neurology and Neuropsychology: Examining cognitive functions through neuroimaging and machine learning provides deeper insights into disorders like aphasia and neurocognitive deficits. By mapping brain functions and assessing structural changes, these studies advance our understanding of how specific neurological impairments affect behavior and cognition. 💗 Neuropsychological Features: Machine learning models predict mental health outcomes and cognitive declines by analyzing attention and processing speed. This focus on prediction and prevention, especially for conditions like cardiovascular disease impacting cognition, enables proactive care and lifestyle adjustments to mitigate risks. ⚙️ Neurodegenerative Conditions: AI-based predictive models for neurodegenerative diseases like Parkinson’s allow for early, more accurate diagnoses. By analyzing markers in social cognition and emotional processing, this cluster supports personalized interventions, helping to maintain patient quality of life and reduce care burdens. This is only the beginning. This field is absolutely ripe for rapid advance and massive real-world value.

AI is getting closer to accessing the one thing we’ve always considered private: your thoughts. Recent advances in neuro-AI can now identify whether a person recognizes specific information using EEG signals. A 2025 study using deep-learning reached 86.7% accuracy in detecting recognition through the P300 brain wave: a response triggered before conscious awareness. Meanwhile, some jurisdictions are already experimenting with this technology. 🇮🇳 India has used brain-mapping techniques in hundreds of criminal investigations, showing just how quickly neuroscience can enter real-world decision systems. But the implications go beyond law enforcement. AI models can now (fMRI + diffusion models): Reconstruct visual experiences directly from brain activity ✔️ Models that reconstruct what you’re seeing — in near real-time — based solely on your brain activity (Think: AI generating the images your eyes are looking at.) Decode unspoken language in early experimental settings ✔️ Models that reconstruct the words you’re thinking, even if you never speak A 2023–2024 wave of studies using fMRI + LLMs demonstrated the ability to decode semantic meaning of inner speech—turning thoughts into text-like outputs. This raises critical questions for business leaders, policymakers, and innovators: How do we prepare for a world where cognitive data becomes a new category of sensitive information? What safeguards, standards, and governance frameworks will protect mental privacy as neuro-AI scales? The technology is advancing faster than the regulations around it and the organisations that understand this early will be better positioned to navigate what comes next. #AI #Neuroscience #Innovation #Leadership #Ethics #FutureOfWork Reference: Kim, S., Cheon, J., Kim, T., Kim, S. C., & Im, C.-H. (2025). Improving electroencephalogram-based deception detection in concealed information test under low stimulus heterogeneity. arXiv. https://lnkd.in/dyVqBbG3 Takagi & Nishimoto (2022). High-resolution image reconstruction with latent diffusion models from human brain activity. BioRxiv. https://lnkd.in/dfc32mS7 Tang, J., LeBel, A., Jain, S. et al. Semantic reconstruction of continuous language from non-invasive brain recordings. Nat Neurosci 26, 858–866 (2023). https://lnkd.in/dnQxcS_d

Some research projects are just cool. Meta’s AI lab has created a system that can predict how your brain will respond when you watch a movie, not in a vague way, but down to the second, across different regions of your cortex. It is a leap forward for neuroscience and it just earned Meta first place in one of the most competitive brain modelling contests in the world. The winning system is called TRIBE, short for TRImodal Brain Encoder. It was built to compete in the Algonauts 2025 Challenge, an annual event where teams from around the globe race to design the most accurate computer models of the human brain. The challenge is run by the Algonauts Project, launched in 2019 to bring neuroscientists and AI researchers together in open competition. The aim is to push forward our understanding of how biological and artificial intelligence work, using shared datasets and transparent methods. The Algonauts Project is inspired by the idea that comparing brains and AI models could reveal what makes intelligent systems efficient, robust and trustworthy. It promotes faster innovation by using algorithms to test theories about the brain, fosters collaborative science by encouraging open sharing of methods, and is designed to expand across disciplines as the community grows. Meta’s team trained TRIBE on one of the richest brain datasets ever assembled, more than 80 hours of high-resolution fMRI brain scans per person, recorded while participants watched TV shows and films ranging from Friends to The Bourne Supremacy. For each moment of footage, TRIBE analysed three streams of information. These streams were synchronised in time and fed into a transformer model that learned how the brain integrates them. TRIBE won by a clear margin over more than 260 competing teams. It explained more than half of the explainable variation in brain responses, a high benchmark in neuroscience. The biggest performance gains came from using all three modalities, text, sound and vision, mirroring the way human perception combines sensory inputs. In association areas of the brain, the multimodal approach delivered up to 30 percent better predictions than any single modality model. For people, the immediate impact is not that Meta can read your mind, because it cannot. But it shows that AI is getting better at modelling how the brain processes complex, real-world information. In time, tools like TRIBE could support better diagnostics for conditions that affect language or sensory processing, create more engaging and accessible entertainment, or adapt education to the way an individual’s brain responds to material. TRIBE offers a way to study how language, vision and sound are combined in real time across the cortex, and it could help unify previously separate strands of neuroscience research. By showing that a single, non-linear, multimodal model can predict whole-brain responses, it also points to the possibility of more integrated models of cognition in the years ahead.

🧠 𝐅𝐫𝐨𝐦 𝐍𝐞𝐮𝐫𝐨𝐧𝐬 𝐭𝐨 𝐍𝐞𝐭𝐰𝐨𝐫𝐤𝐬: 𝐇𝐨𝐰 𝐍𝐞𝐮𝐫𝐨𝐬𝐜𝐢𝐞𝐧𝐜𝐞 𝐢𝐬 𝐒𝐡𝐚𝐩𝐢𝐧𝐠 𝐭𝐡𝐞 𝐅𝐮𝐭𝐮𝐫𝐞 𝐨𝐟 𝐀𝐈 The brain is the most efficient computational system we know—20 watts of power, trillions of connections, and a lifetime of learning. For decades, artificial intelligence has borrowed its metaphors from neuroscience, but we’re only now beginning to integrate its principles. Neuroscience teaches us that intelligence emerges from prediction, attention, memory, and modulation. Each of these maps beautifully to AI: convolutional networks mirror visual cortex processing; attention mechanisms reflect how our brains route focus; reinforcement learning follows dopamine’s reward prediction; and replay systems echo the hippocampus consolidating experience during sleep. What’s next is not more data—it’s better cognition. The next generation of AI will adopt the brain’s hybrid strategy: combining fast episodic recall with slow semantic learning, training models that dream, replay, and adapt continually. As we refine spiking neural networks, neuromodulatory systems, and local learning rules, we edge closer to biologically plausible intelligence—machines that reason, remember, and evolve with us. AI is becoming less of a tool and more of an organism—one that learns as we do, forgets as we do, and one day, perhaps, understands as we do. Singularity Systems, the AI research arm of Cybersecurity Insiders is studying these biological parallels to architect systems that bridge the gap between human cognition and machine autonomy. What if our next AI doesn’t just calculate—but thinks?

🧠 Are we rediscovering the brain through AI? I’m excited to share our new paper just published in Computers in Biology and Medicine (Volume 204, March 2026, 111533) introducing ELSA (Emergent Language Symbolic Autoencoder): a weakly supervised framework for hierarchical modeling and labeling of intrinsic functional brain networks https://lnkd.in/gSK73-RY. 🤖 The bigger question for NeuroAI If CNNs learn through hierarchical local features, and Transformers through dynamic routing and long-range interactions… what is the brain doing when it coordinates distributed functional networks across scales? Are we uncovering shared computational principles (hierarchy, routing, compression) or are modern architectures simply powerful metaphors for biology? And if there is a real parallel, is it closer to message passing, attention-like gating, predictive coding, energy minimization, or something we haven’t named yet? 🧬 Why this matters We still don’t fully understand how the brain organizes and reconfigures functional systems. These networks are hierarchical, overlapping, and dynamic, not flat categories. In neuroscience, models matter not just for prediction but for mechanistic insight. For clinical translation, interpretability is essential: models should help us reason about brain organization, not only predict outcomes. 🧩 What this work does 🧠 Uses resting-state fMRI data to study the hierarchical organization of intrinsic functional brain networks 🤖 Trains ELSA, a weakly supervised symbolic autoencoder, to learn multilevel network structure from neuroimaging signals 🔤 Links learned clusters to symbolic “sentences” to support interpretability 📐 Introduces a generalized hierarchical loss to encourage consistency across levels of organization 📊 Evaluates performance using a hierarchical consistency metric (>97% in the best-performing configuration) 💻 Code & resources: https://lnkd.in/g-mTGG3q 🙏 Grateful to Dr. Haris Sair for his leadership and collaboration throughout this work. Deep thanks as well to Ammar Ahmed, Craig Jones, and to the teams at The Johns Hopkins University School of Medicine and Johns Hopkins Malone Center for Engineering in Healthcare. Collaborating with such an exceptional group, bringing together clinical insight, engineering rigor, and scientific curiosity, has been truly inspiring. #NeuroAI #ComputationalNeuroscience #fMRI #BrainNetworks #Interpretability #DeepLearning #MedicalImaging #MedicalImaging #MicrosoftAI #AzureAI #AIforHealthcare #JohnsHopkins

Excited to release what we’ve been working on at Amaranth Foundation, our latest whitepaper, NeuroAI for AI safety! A detailed, ambitious roadmap for how neuroscience research can help build safer AI systems while accelerating both virtual neuroscience and neurotech. Building safe AI is more pressing as AI becomes more powerful. The human brain remains our only example of safe general intelligence. Despite our flaws, we have robust senses, cooperate with others, & recover from mistakes. Understanding the brain is a promising route toward AI safety. We adopt Deepmind’s framework and evaluate 7 concrete proposals to enhance safety without dramatically increasing capabilities. For each approach, we identify specific opportunities for how neuroscientists can contribute to building safer AI systems. The proposals we evaluate span the whole gamut of NeuroAI: biophysically detailed models, embodied and sensory digital twins, cognitive architectures, fine-tuning conventional AI models with brain data, reverse-engineering the loss functions of the brain and mech interp. Traditional neuroscience moves far too slowly to impact AI development on relevant timescales. To meaningfully contribute to AI safety, we need to dramatically accelerate our ability to record, analyze, simulate, and understand neural systems. Nothing about safer AI is inevitable. We think now is the time to pursue this differential path, because scalable neuroscience tools and technology make collecting high-res datasets possible, accelerated by ML progress and decreased compute costs. Beyond AI safety, we’ve synthesized a lot of literature (90 pages and 700+ references). Intermediate milestones on this path to safe AI will shorten the time to translate new neurotechnologies and quicken the scientific feedback loop with virtual neuroscience experiments. Last but not least, a huge thanks to James Fickel and to our collaborators. Without them, this work would not have been possible! Read online: https://neuroaisafety.com ArXiV PDF: https://lnkd.in/eRrHreTT

Brain-computer interfaces now let paralyzed patients control devices with thoughts. The technology is advancing faster than expected. Current breakthrough applications: Paralyzed patients typing with brain signals ↳ Speech restoration for ALS patients ↳ Robotic arms controlled by thoughts ↳ Depression treatment through targeted stimulation ↳ Memory enhancement research beginning How it works: Electrodes record individual neuron activity ↳ AI decodes intended movements or words ↳ Computer translates signals to actions ↳ Real-time feedback improves accuracy ↳ Learning happens on both sides The medical revolution: Deep brain stimulation for Parkinson's ↳ Responsive neurostimulation for epilepsy ↳ Transcranial magnetic stimulation for depression ↳ Cochlear implants restore hearing ↳ Visual prosthetics in early trials What patients tell me: Brain stimulation changes lives completely ↳ Parkinson's tremor disappears instantly ↳ Seizures stop after years of suffering ↳ Depression lifts when medications failed ↳ Feel like they got their identity back The safety evolution: Early devices required open brain surgery ↳ Now using ultrasound and magnetic fields ↳ Temporary effects tested before permanent ↳ Complication rates very low ↳ Safer than many common medications Consumer applications emerging: Enhanced meditation through neurofeedback ↳ Sleep optimization via brain monitoring ↳ Attention training for focus issues ↳ Gaming interfaces using brain signals ↳ Cognitive fitness tracking The learning acceleration: AI identifies patterns humans miss ↳ Optimizes treatment automatically ↳ Predicts response before starting ↳ Personalizes therapy to individual circuits ↳ Reduces trial and error dramatically Challenges remaining: Signal quality degrades over time ↳ Brain tissue responds to foreign objects ↳ Individual variation in brain organization ↳ Long-term safety still being studied ↳ Cost and accessibility issues The accessibility question: Currently limited to severe conditions ↳ Insurance coverage expanding slowly ↳ Costs dropping with technological advances ↳ Simpler versions for consumer market ↳ Could become common as pacemakers Ethical considerations: Who controls the technology? ↳ Privacy of neural information ↳ Enhancement vs treatment boundaries ↳ Equality of access important ↳ Need frameworks before widespread adoption 💬 Comment if you'd consider brain technology for medical needs ♻️ Repost if brain interfaces will transform medicine 👉 Follow me (Reza Hosseini Ghomi, MD, MSE) for neurotechnology advances Citations: Willett FR, et al. High-performance brain-to-text communication via handwriting. Nature. 2021. Musk E, Neuralink. An integrated brain-machine interface platform with thousands of channels. Journal of Medical Internet Research. 2019.

Innovation happens when AI meets neuroscience and lives are changed. After surviving a stroke, Ann was unable to speak or express herself for 18 years. Today, that has changed. Researchers at University of California, San Francisco developed a groundbreaking brain computer interface that reads neural signals directly from her brain and decodes them in real time. Powered by AI, those signals are transformed into text, synthetic speech, and even facial expressions through a digital avatar that reflects how Ann would naturally communicate. There’s no typing. No word selection. Ann simply thinks about what she wants to say and the system converts those thoughts into full sentences, complete with emotional expression like smiling or raised eyebrows. This is one of the first demonstrations of a brain implant restoring both voice and emotional expression together. It marks a major step forward in using AI models to map brain activity to language and movement. The future of communication for people with severe paralysis is no longer theoretical. It’s here and it’s profoundly human. #AI #Neuroscience #Innovation #BrainComputerInterface #HealthcareAI #HumanCenteredAI #medtech #device #ml #human #life #world

One of the most bizarre + complex talks at NeurIPS [1] was given by my fellow Yorkshireman, the inimitable Prof Karl Friston [2], explaining active inference to a room full of #AI people who are not really neuroscientists. This was interesting to me because: 🎇 Karl Friston is a one-of-a-kind pioneer in brain mapping + neuroimaging analysis methods (co-developing SPM + DCM and to some extent Gaussian Random Field theory) but then he totally switched his research to study "active inference" - a theory of how brains work - which is somewhat related to Reinforcement Learning (RL) in #AI 🎇 He gave a talk at OHBM that was the most complex I have ever seen given at any conference, explaining the Free Energy Principle (FEP). which describes how systems (biological or artificial) minimize uncertainty about their environment by inferring the causes of sensory inputs + acting to fulfil their prior beliefs or reduce uncertainty. As far as I can tell, the main point is that agents (such as humans) try to minimize "variational free energy", actively measuring the difference between their brain's model of the world + its sensory experiences. 🎇 The AI people who do reinforcement learning [3] (e.g., programming self-driving cars) are doing a related thing, where an agent learns a "policy" to maximize a cumulative reward signal, given a state + actions; the agent balances trying new actions (exploration) with seeking known rewards, so the AI people in the room are using somewhat similar frameworks that focus on adaptive behavior to achieve the "best" outcomes. 🎇 Even so, these frameworks (active inference and reinforcement learning) aren't completely similar: RL maximises an external reward signal, but active Inference minimises a combination of "surprise "(discrepancy between observations + expectations) + uncertainty (lack of confidence in beliefs) but also (2) creates a generative model of the environment that predicts sensory inputs, given internal states. 🎇 Presumably the AI people in the room are not incorporating intrinsic motivation into their algorithms but they may do now. [1] I didn't go, but if you want to watch it, you can watch the whole thing for 200 USD and it's well worth it. The NeuroAI workshop was great, as was Arnaud Doucet's flow matching talk, which will give you a new perspective on Generative AI [2] https://lnkd.in/ghYNb-ke [3] actually not all AI people are doing RL, and most AI books with predictive and generative AI have RL as the last chapter or the last 1/3, which can you can ignore if you are not designing autonomous agents (it is still interesting if you have time to read it).