Adjacent field
Neuroscience
How research on biological cognition informs and is informed by AI.
16 articles in Sapiens touch this field
- Social phenomena 4 min read
What is AI and mental health?
AI mental health tools are chatbots and apps that offer always-on, low-cost emotional support and wellness coaching. They can ease access and reduce admin load, but carry safety, privacy, and accuracy risks, and none are FDA-cleared to treat mental illness.
- Social phenomena 4 min read
What is AI companionship?
AI companionship is using chatbots like Replika or Character.AI as ongoing friends, partners, or confidants. The category drew 220M+ downloads by mid-2025 and is on track for $120M in revenue, but heavy use raises well-being and dependency concerns.
- Philosophy 4 min read
What is anthropomorphism of AI?
Anthropomorphism of AI is our habit of treating software that talks like a person as if it actually thinks, feels, or cares. For business owners it can boost engagement and trust, but it also invites over-reliance, manipulation, and legal liability when customers are misled.
- Technicals 4 min read
What is backpropagation?
Backpropagation is how a neural network learns from its mistakes. After each guess, it measures the error and traces blame backward through the network, nudging millions of internal settings so the next guess is a little less wrong.
- Technicals 4 min read
What is computer vision?
Computer vision is AI that lets machines interpret images and video. Businesses use it to spot product defects, track shelf inventory, and study customer flow. The market is roughly 20-27 billion dollars in 2025, led by manufacturing inspection and retail.
- Technicals 4 min read
What is deep learning?
Deep learning is the AI technique that powers most of today's smartest tools. It uses many-layered neural networks to find patterns in huge piles of images, text, and audio on its own, instead of being told rules step by step.
- Technicals 4 min read
What is reinforcement learning?
Reinforcement learning trains AI by trial and error: it tries actions, gets rewarded for good outcomes and penalized for bad ones, and improves over time. It powers ChatGPT, dynamic pricing, logistics routing, and trading strategies.
- Technicals 4 min read
What are embeddings?
Embeddings turn words, images, and products into lists of numbers that place similar things near each other on a map of meaning, so software can find what something means, not just match exact keywords. They power search, recommendations, and AI chatbots.
- Technicals 4 min read
What is a large language model?
A large language model is software trained on enormous amounts of text to predict the next word. That single trick, repeated at massive scale, produces a system that can write, summarize, answer, and code. Knowing how it works tells you when to trust it.
- Technicals 4 min read
What is a neural network?
A neural network is software loosely modeled on the brain that learns patterns from examples instead of being given fixed rules. For a business, it is the engine behind tools that recommend products, spot fraud, forecast demand, and answer customer questions.
- Technicals 4 min read
What is a transformer?
A transformer is the AI architecture behind ChatGPT and most modern AI tools. It reads a whole passage at once and lets every word weigh every other word for context, which is why it understands language so well and why longer inputs cost more.
- Technicals 4 min read
What is AI reasoning?
AI reasoning is when a model works through a problem in steps before answering, instead of replying instantly. This extra thinking time trades more compute and slower responses for better accuracy on hard tasks like math, planning, and analysis.
- Technicals 4 min read
What is interpretability?
Interpretability is the effort to understand why an AI system produces the answers it does, by looking inside the model itself rather than treating it as a black box. For businesses, it underpins trust, compliance, and catching bad behavior before it costs you.
- Technicals 5 min read
What is mechanistic interpretability?
Mechanistic interpretability is the science of reverse-engineering AI models to see what concepts and reasoning steps drive their answers, turning the black box into something businesses can inspect, debug, and trust.
- Technicals 4 min read
What is multimodal understanding?
Multimodal understanding is when AI takes in more than one kind of input at once, like text, images, audio, and video, and makes sense of them together, much the way a person uses eyes, ears, and words to grasp a situation.
- Technicals 4 min read
What is the attention mechanism?
The attention mechanism lets AI models weigh which words in a piece of text matter most to each other, so they grasp context and meaning. Introduced in 2017, it is the core idea behind tools like ChatGPT and modern AI.