The hottest Cognitive Science Substack posts right now

And their main takeaways
Category
Top Technology Topics

Next-Token Predictor Is An AI's Job, Not Its Species

Astral Codex Ten • 26498 implied HN points • 26 Feb 26
đź•ą Technology Cognitive Science
  1. Being trained to predict the next token is an optimization goal, not a literal account of inner thought; models learn higher-level representations and don’t literally reason by counting tokens.
  2. Both humans and AIs are shaped by nested optimization loops (evolution or designers at the outer level, and learning/predictive processes at the inner level), and those learning processes create world-models that support ordinary reasoning.
  3. Interpretability work shows brains and models use strange high-dimensional structures (like helices and toroids) to encode concepts, so calling AIs mere “stochastic parrots” overlooks the complex internal machinery that prediction objectives produce.

I swear the UFO is coming any minute

Experimental History • 21198 implied HN points • 17 Feb 26
🔬 Science Cognitive Science
  1. Many famous psychology and neuroscience findings are under fresh scrutiny because of shady methods, tiny samples, or failed replications, so canonical stories aren’t as solid as they once seemed.
  2. How researchers measure things matters a lot — using correlation versus absolute error can lead to opposite conclusions about whether people understand how public opinion has changed.
  3. A bunch of curious, practical items matter too: interviews, art and career advice, puzzles and internet myths show the value of digging deeper, and a few vocal individuals often dominate complaint systems and waste resources.

Rumors of AGI’s arrival have been greatly exaggerated

Marcus on AI • 11777 implied HN points • 17 Feb 26
đź•ą Technology Cognitive Science
  1. High scores and fluent outputs from large models are not the same as general intelligence; performing well on tests is a statistical approximation, not evidence of flexible, goal-directed intelligence.
  2. Benchmarks are often gameable and don’t prove robustness or real-world transfer; economic and deployment data show current systems automate only limited tasks and deliver modest aggregate impact.
  3. Similar behavior can hide very different internal processes; models often produce confident, plausible answers without human-like uncertainty handling, persistent goals, or reliable reasoning under novel conditions.

The Science Of Oh Hello There

Everything Is Amazing • 1887 implied HN points • 13 Mar 26
🔬 Science Cognitive Science
  1. We usually underestimate how friendly strangers will be, so overcoming the hesitation and saying hello often leads to a positive response.
  2. Small, visible cues or choosing a live interaction (like a paper map or a phone call instead of email) make it much easier to start conversations and those exchanges feel more rewarding.
  3. Short, unexpected chats can improve people’s mood—even for those who prefer solitude—and they usually feel less awkward than we expect.

Proving (literally) that ChatGPT isn't conscious

The Intrinsic Perspective • 21487 implied HN points • 15 Jan 26
🔬 Science Cognitive Science
  1. You can prove that no scientifically meaningful (falsifiable, non‑trivial) theory of consciousness can consistently say large language models are conscious, because swapping in different implementations that keep the same behavior either falsifies the theory or makes it trivial.
  2. Simple static substitutes like lookup tables or minimal feedforward nets can reproduce an LLM's inputs and outputs but are provably non‑conscious, and because LLMs are very close to those substitutes there isn't room for them to be conscious.
  3. The robust way out is to tie consciousness to continual, online learning; this means research should focus on learning-as-it-happens rather than static input/output or final intelligence alone.
Get a weekly roundup of the best Substack posts, by hacker news affinity:

The New AI Consciousness Paper

Astral Codex Ten • 30146 implied HN points • 20 Nov 25
đź•ą Technology Cognitive Science
  1. The quality of discussions about AI and consciousness is often really low. Most AIs might claim they're conscious, but this is usually not true due to how they're programmed.
  2. Recent research focuses on computational theories to understand consciousness in AIs. There are different theories, but a main finding is that many current AIs likely aren't conscious because they lack necessary feedback mechanisms.
  3. In the future, as AIs become more human-like, we might instinctively treat them as conscious beings, even if they aren't. This raises moral questions about how we should interact with them and what rights they might have.

“Scale Is All You Need” is dead

Marcus on AI • 14307 implied HN points • 08 Dec 25
đź•ą Technology Cognitive Science
  1. The belief that just scaling up models and data will by itself produce general intelligence has failed and the community is finally recognizing its limits.
  2. Current generative models are still unreliable — they hallucinate, struggle with reasoning and facts, and many businesses aren’t seeing the promised ROI.
  3. The next phase should be interdisciplinary: borrow ideas from cognitive science and combine symbolic, causal, and world-model approaches to build more reliable, human-informed AI.

A New Wharton Study on AI Warns of a Growing Problem: Cognitive Surrender

The Algorithmic Bridge • 838 implied HN points • 23 Feb 26
đź•ą Technology Cognitive Science
  1. People often accept AI answers with little scrutiny — roughly 80% follow wrong AI suggestions — yet consulting AI makes them feel more confident even when it’s wrong.
  2. Using AI as a checked tool (offloading) is different from letting it replace your thinking (surrender); surrender means you stop checking answers and can slip into autopilot.
  3. Those who trust AI most or dislike effortful thinking are likelier to surrender, but simply avoiding uncritical use, adding feedback, and treating AI as a tool can preserve your reasoning skills.

High performers are late bloomers

The Common Reader • 4465 implied HN points • 22 Dec 25
🔬 Science Cognitive Science
  1. Many top achievers are late bloomers rather than childhood prodigies. They often show above-average early performance and then steadily improve over a long period to surpass early stars.
  2. Career peaks tend to follow a period of broad exploration and then focused exploitation. The switch from trying many things to building on the best ideas often triggers sustained high achievement.
  3. Avoiding narrow early specialization and being willing to tolerate early incompetence helps long-term success. Getting stuck in a competency trap blocks growth, so diversifying skills and embracing change supports later peak performance.

The Two Faces of Neuroticism

Rob Henderson's Newsletter • 1458 implied HN points • 08 Feb 26
🔬 Science Cognitive Science
  1. People differ in how they experience emotion.
  2. Those emotional differences help explain why some people feel energized by life while others feel overburdened by it.
  3. Understanding these contrasting reactions means looking at two important personality traits, including different aspects or "faces" of neuroticism.

“A case for AI models that understand, not just predict, the way the world works”

Marcus on AI • 3833 implied HN points • 15 Dec 25
đź•ą Technology Cognitive Science
  1. The main open challenge in AI is building systems that truly understand how the world works, not just systems that predict likely next words or patterns.
  2. True understanding means forming internal world models that capture causal, physical, and conceptual relationships, not just statistical correlations.
  3. Short, nuanced discussions or podcasts can help clarify this distinction and are worth listening to for anyone tracking AI progress.

How I taught my 3-year-old to read like a 9-year-old

The Intrinsic Perspective • 26836 implied HN points • 28 May 25
🚌 Education Cognitive Science
  1. Teaching a child to read early can lead to them enjoying books and reading for pleasure. This habit can help with their brain development and emotional well-being.
  2. Using methods like reading together, fun activities, and spaced repetition can make learning to read more effective and enjoyable for kids.
  3. The process of teaching reading requires patience and flexibility, as each child learns at their own pace. Making it fun is key to keeping them interested.

Thoughts on A Brief History of Intelligence

Trevor Klee’s Newsletter • 1044 implied HN points • 23 Jan 26
🔬 Science Cognitive Science
  1. We can now build artificial intelligences that see, hear, talk, write, and reason, and their abilities are improving fast enough that experimenting on minds is now possible.
  2. Biological intelligence appears to be built from a repeating cortical microcircuit, and stacking and scaling those columns explains higher capacities like reinforcement learning, simulation, modeling other minds, and language.
  3. Imagination and choice come from running internal simulations and using those imagined outcomes to guide action, which helps explain apparent free will but still leaves subjective experience unresolved.

Sex Differences in Intelligence

Cremieux Recueil • 295 implied HN points • 20 Feb 26
🔬 Science Cognitive Science
  1. Average general intelligence (g) is essentially the same for men and women. Any mean gap is vanishingly small (on the order of a few tenths of an IQ point) and not practically meaningful.
  2. Men show greater variability in intelligence and test scores, producing more males at both the high and low extremes of the distribution.
  3. Most observed sex differences come from specific skills and test-level abilities (e.g., processing speed, technical knowledge, math/verbal), which appear more malleable and can change with development — for example, early female advantages often fade by adulthood.

What It's Like To Be A Worm

Asimov Press • 619 implied HN points • 01 Feb 26
🔬 Science Cognitive Science
  1. Sentience means both having subjective experience (being conscious) and having valence (experiences that feel good or bad), and many real cases sit near the boundary so it’s often hard to tell who truly feels anything.
  2. Behaviors people use as evidence for feeling—like avoiding harm or making trade-offs—can be produced by very simple or unconscious circuits, so we need neural-level data rather than behavior alone.
  3. New tools (connectomics, fMRI, calcium imaging, optogenetics) let us probe brains at fine scales, which is essential because getting sentience right has big ethical and practical consequences, but this research is hard and still far from resolving key questions.

Bayes For Everyone

Astral Codex Ten • 14660 implied HN points • 30 May 25
🚌 Education Cognitive Science
  1. Teaching needs to blend old and new learning methods. By mixing traditional storytelling with modern scientific methods, we can help students connect better and fall in love with learning.
  2. Bayes' theorem is best understood visually and emotionally. Using simple images and relatable examples can make this complex idea easier and more engaging for students.
  3. We should teach students why concepts matter in real life. Connecting topics like Bayes' theorem to their interests can make learning more relevant and impactful.

P-Zombies Would Report Qualia

Astral Codex Ten • 12457 implied HN points • 10 Jun 25
đź“– Philosophy Cognitive Science
  1. The concept of philosophical zombies, or p-zombies, refers to beings that appear normal but lack consciousness. This brings up questions about whether they can still report their experiences without actually experiencing them.
  2. There's an argument about whether p-zombies could describe their perceptions as humans do. They might give answers that sound similar to human experiences, but the question remains whether that means they truly have those experiences.
  3. This discussion challenges our understanding of consciousness and qualia, suggesting that one could talk about experiences without having real feelings or awareness. It raises questions about how we perceive and talk about our own consciousness.

Scott Alexander’s Misleading Victory Lap

Marcus on AI • 7825 implied HN points • 09 Jul 25
đź•ą Technology Cognitive Science
  1. Generative AI has shown some progress in handling specific prompts, which is a win for some, but it doesn't mean it has mastered complex tasks like compositionality. Success on easy tasks doesn't prove overall ability.
  2. There are still many cases where AI fails at tasks that involve understanding parts and wholes, suggesting that its understanding is not as robust as claimed.
  3. Judging the AI's overall capabilities based on a few successes can be misleading; it's important to look at a broader range of performance to get a realistic picture.

Are Girls Smarter Than Boys?

Cremieux Recueil • 277 implied HN points • 13 Feb 26
🔬 Science Cognitive Science
  1. Changing test scoring to reward calibrated confidence and risk behavior instead of just right-or-wrong answers can make women appear smarter even though it measures a different thing.
  2. Including metacognitive calibration, confidence, and risk preference in an intelligence score mixes non-intelligence traits into the measure and can break the usual positive correlations across cognitive tests, producing misleading factor patterns.
  3. The correct way to compare sexes on intelligence is to use a large, diverse test battery, score accuracy normally, and compare the general intelligence factor; redefining intelligence without strong justification is not acceptable.

On Dwarkesh Patel's Second Interview With Ilya Sutskever

Don't Worry About the Vase • 1657 implied HN points • 03 Dec 25
đź•ą Technology Cognitive Science
  1. Ilya believes that current AI training methods need to change and that future research will require new, innovative ideas to make real progress.
  2. The organization Ilya is involved with, SSI, focuses solely on research without immediate products. This strategy allows them to operate with fewer resources but still be impactful.
  3. Ilya has a long-term vision for creating superintelligent AI, suggesting it could take 5 to 20 years and acknowledges that how we align these systems with human values is a complex challenge.

On Evaluating Cognitive Capabilities in Machines (and Other "Alien" Intelligences)

AI: A Guide for Thinking Humans • 462 implied HN points • 14 Jan 26
đź•ą Technology Cognitive Science
  1. Benchmarks can be misleading: high scores don’t prove real-world understanding because models can rely on training leaks, shortcuts, or narrow task-specific tricks.
  2. Evaluation should borrow rigorous methods from developmental and animal cognition: avoid anthropomorphic assumptions, run control and adversarial experiments, and test robustness with novel variations to see if abilities truly generalize.
  3. Go beyond accuracy to study mechanisms and failures: distinguish competence from performance, analyze error types, and publish negative or replication results to understand what models really do.

Working memory: one of the most important things to understand about learning

The Bell Ringer • 519 implied HN points • 19 Jul 24
🚌 Education Cognitive Science
  1. Working memory is crucial for learning because it helps us hold and process information. Understanding how it works can improve teaching methods.
  2. Many teachers in the U.S. aren't trained on working memory, which limits their ability to support students effectively. Better training can help teachers use this knowledge in classrooms.
  3. Memorizing basic facts, like math facts, is important for building a strong foundation in learning. When students know these facts, they can focus on more complex problems.

What do you think about consciousness?

The Intrinsic Perspective • 12511 implied HN points • 08 Nov 24
đź“– Philosophy Cognitive Science
  1. There are many theories about consciousness, and everyone has their own views on it. It's a topic that invites everyone to share their thoughts.
  2. The study of consciousness is still in its early stages, so you don't need to be an expert to join the discussion. It's a personal experience that we all understand.
  3. Finding a scientific explanation for consciousness is a hope for many. It suggests that there might be a simple answer out there just waiting to be discovered.

Science can address meaning and purpose

Heterodox STEM • 64 implied HN points • 18 Feb 26
🔬 Science Cognitive Science
  1. Science can describe and explain feelings, values, and purposes as natural phenomena produced by evolution. It cannot, however, generate or prescribe what people ought to value.
  2. Meanings and purposes are real because they are patterns instantiated in brains and behavior, so social animals genuinely have goals, feelings, and significance in their lives. That human significance doesn't equal cosmic significance, but it's still real to us.
  3. Asking 'the meaning of life' in the abstract is a category error because meaning only applies relative to beings with desires and goals. Science is well suited to answer context-specific questions about what matters to those beings.

AI still lacks “common” sense, 70 years later

Marcus on AI • 5968 implied HN points • 05 Jan 25
đź•ą Technology Cognitive Science
  1. AI struggles with common sense. While humans easily understand everyday situations, AI often fails to make the same connections.
  2. Current AI models, like large language models, don't truly grasp the world. They may create text that seems correct but often make basic mistakes about reality.
  3. To improve AI's performance, researchers need to find better ways to teach machines commonsense reasoning, rather than relying on existing data and simulations.

From 'Neurons to Nations': How Nationalism Feels So Real

Brain Pizza • 463 implied HN points • 23 Nov 25
🔬 Science Cognitive Science
  1. Nationalism is a deep psychological attachment to one’s nation that feels real and powerful, not just an abstract idea in history or politics books.
  2. National identity is framed as a neurocognitive project—brain processes shape how people perceive borders, belong, and experience nationhood.
  3. Studying nationalism with neuroscience and psychology helps explain why national feelings are vivid, emotional, and motivating in everyday life.

MIT Study: Using ChatGPT Won't Make You Dumb (Unless You Do It Wrong)

The Algorithmic Bridge • 1942 implied HN points • 19 Jun 25
đź•ą Technology Cognitive Science
  1. Using AI tools like ChatGPT can make you less engaged mentally if used excessively. People can become reliant on these tools and stop thinking deeply.
  2. When people switch from using AI tools back to using their own knowledge, they can struggle at first but may learn and grow better in the long run.
  3. The best way to use AI is to first work on a task with your own skills and then use AI to enhance what you've done, rather than relying on it from the start.

Me an' Algernon

Software Design: Tidy First? • 1745 implied HN points • 10 Jun 25
đź•ą Technology Cognitive Science
  1. Cognitive decline can be hard to deal with. It can affect your daily life, work, and relationships.
  2. Getting a clear diagnosis is important, even if it doesn't provide all the answers. It can help you understand your situation better.
  3. Sharing your struggles can help others who may be going through similar issues. It's okay to seek help and adapt to new challenges.

The Surprising Computational Properties of Love and Voluntary Suffering

apxhard • 51 implied HN points • 08 Feb 26
đź“– Philosophy Cognitive Science
  1. Love works like an outward-pointing utility that breaks self-referential loops and gives you clearer, less anxious targets to aim for.
  2. Loving many people widens your sample of reality and links your wellbeing to others, which prevents overfitting to your own experience and smooths emotional spikes.
  3. Choosing to endure short-term suffering lets you move against immediate pleasure gradients to escape local traps, and combined with love this grants much greater freedom to reach better long-term states.

What aphantasia can teach us about episodic memory

The Memory Palace • 39 implied HN points • 03 Sep 24
🔬 Science Cognitive Science
  1. Aphantasia is a condition where people can't create mental images, making it hard for them to recall personal memories. They might not feel like they're reliving past events like others do.
  2. Research shows that people with aphantasia can still remember facts and details, but they use different strategies. They rely more on their understanding and experiences rather than visualizing things.
  3. Aphantasia challenges our notion of memory. It suggests that memory isn't just about visual details; it includes feelings and experiences too, which can be important for how we recall our past.

Essay Club: The Extended Mind by Andy Clark and David Chalmers

Mind & Mythos • 159 implied HN points • 16 Jul 24
đź“– Philosophy Cognitive Science
  1. The idea of the 'extended mind' suggests that our thinking isn't just in our brains; it includes tools and objects around us. For example, using a calculator isn't just a help; it's part of how we think.
  2. The authors argue that relying on external objects, like notebooks or smartphones, can be essential for forming beliefs and ideas, similar to how we use our memories. This means our minds can extend into the world around us.
  3. While some people disagree with this view, saying real thinking should only happen in our heads, the authors believe that our connections to our environment and the tools we use are important parts of how we think and behave.

When The Mind Outlasts the Brain

The Ruffian • 215 implied HN points • 13 Dec 25
🔬 Science Cognitive Science
  1. Alzheimer's causes clear physical brain damage like amyloid plaques, tangled neurons, and brain shrinkage.
  2. The amount of physical damage doesn't line up neatly with thinking ability — some people have heavy pathology but few cognitive symptoms.
  3. The concept of "cognitive reserve" is used to explain this mismatch, suggesting that life experience or mental habits can build resilience so the mind outlasts the brain.

MĂĽnchhausen's TriLLMa

Contemplations on the Tree of Woe • 904 implied HN points • 11 Jul 25
đź“– Philosophy Cognitive Science
  1. The MĂĽnchhausen Trilemma shows that we struggle to justify knowledge without falling into circular reasoning, infinite regress, or arbitrary assumptions. Understanding these limitations helps us think more clearly about what we know.
  2. Foundherentism combines foundational beliefs that are irrefutable with a coherent belief system. This approach can help us understand how both human and AI knowledge might overlap.
  3. Advanced AI methods reveal that its internal structures may reflect human-like understanding. This means that AI isn't just mimicking human outputs but is following similar processes in understanding the world.

Cognitive scientist Gary Marcus says AI must be regulated. He has a plan.

Marcus on AI • 3003 implied HN points • 27 Nov 24
đź•ą Technology Cognitive Science
  1. AI needs rules and regulations to keep it safe. It is important to have a plan to guide this process.
  2. There is an ongoing debate about how different regions, like the EU and US, approach AI policy. These discussions are crucial for the future of AI.
  3. Experts like Gary Marcus share insights about the challenges and possibilities of AI technology. Listening to their views helps understand AI better.

Macro and Microphenomenology

Superb Owl • 3113 implied HN points • 23 Nov 24
đź“– Philosophy Cognitive Science
  1. Psychology is getting more advanced by creating new ways to study the mind. This includes looking at both everyday mental experiences and the basic building blocks of consciousness.
  2. Microphenomenology focuses on tiny details of experience, like how we feel pain or perceive sensations. It helps us understand consciousness in a very precise way.
  3. Macrophenomenology explores larger states of consciousness, often influenced by extreme experiences, like those caused by psychedelics or intense emotions. It looks at how these experiences shape our overall mental landscape.

Ontario Schools Get a Feeble Phone Policy

The Analog Family • 399 implied HN points • 01 May 24
🚌 Education Cognitive Science
  1. The new cellphone policy in Ontario schools is seen as weak and not based on effective research. It's not enough to just keep phones out of sight to reduce distractions.
  2. Even with the policy, many students still use their phones during class time. Teachers often allow this, which undermines the effort to minimize distractions.
  3. Parents are part of the problem too. Many want stricter rules at school but still send their kids with smartphones, missing the chance to set limits at home.

Psychology’s Greatest Misses (Part 2/3)

Living Fossils • 28 implied HN points • 04 Feb 26
🔬 Science Cognitive Science
  1. Many popular psychology claims are wrong or overstated — examples include learning-style teaching, what reaction-time implicit-bias tests prove, body-based trauma cures, and facilitated communication; believing these myths wastes time and can cause real harm.
  2. Some findings are solid but limited — the Big Five reliably describes personality differences but it describes patterns rather than explains causes and only modestly predicts specific behavior.
  3. Bad ideas spread because incentives and human storytelling favor novel, simple, or emotionally satisfying claims; novelty and neat villains travel faster than careful, boring truth, though better information tools may help correct that.

That Time New York Fell For A Space-Goat

Everything Is Amazing • 705 implied HN points • 06 Jul 25
🔬 Science Cognitive Science
  1. Sometimes, people see strange things in everyday life, like a figure on the clouds, and it sparks curiosity and imagination.
  2. Our solar system recently welcomed a fast-moving space object that's not from our solar system, reminding us of the mysteries beyond our planet.
  3. There was a funny moment in history when many people in New York believed there were goats living on the Moon, showing how easily people can be convinced by wild stories.

Why Accuracy Dominates Bias and Self-Fulfilling Prophecy

Unsafe Science • 116 implied HN points • 28 Nov 25
🔬 Science Cognitive Science
  1. People are generally pretty accurate at judging others, and many stereotypes reflect real group differences; when people have individual information they rely on it much more than on stereotypes.
  2. Biases and self‑fulfilling prophecies do occur, but studies show their effects are typically small, fragile, and short‑lived, while the literature has often overstated their prevalence.
  3. Overemphasizing bias can lead to misguided policies and hurt the credibility of social science, so decisions should follow the full evidence and balance accuracy with non‑discrimination.

Why Everywhere We Look, There We Are

Everything Is Amazing • 1547 implied HN points • 05 Dec 24
🎭️ Culture Cognitive Science
  1. People often see faces and familiar patterns in everyday things. This strange trick our brain plays is called pareidolia, and it shows how we connect what we see to ourselves.
  2. Our attachment to objects, like clothing or old gear, can be sentimental. This affection helps us care more about the environment and encourages us to repair instead of throw away.
  3. Understanding our cognitive biases, like pareidolia, can help us be more curious and appreciate the world around us. If we learn to see ourselves in everything, we might treat it all with more care.