2025年人工智能现状报告

stateof.ai airstreet.com Follow our writing on(press.airstreet.com) If you enjoy reading the State of AI Report, we invite you to read and subscribe to Air Street Press, the home ofour analytical writing, news, and opinions. About the authors Nathan Benaich Nathan is the General Partner ofAir Street Capital, aventure capital firm investing in AI-first companies. Heruns the Research and Applied AI Summit (RAAIS), theRAAIS Foundation (funding open-source AI projects), AIcommunities in the US and Europe, and Spinout.fyi(improvinguniversity spinout creation). He studiedbiology at Williams College and earned a PhD fromCambridge in cancer research as a Gates Scholar. State of AI Report 2025 team Nell Norman Ryan Tovcimak Zeke Gillman Ryan is a founder of the AI StackTracker. His work spans red-teamingfrontiermodels,benchmarking theglobal AI competition, and trackingtrendsin AI compute and powerdemands. He holds a BS in Econ fromVanderbilt University. Zeke is a Tech Policy Fellow at Stanford,andco-author of Regulating underUncertainty. He previously worked atHarvard Business School and the DOJAntitrust Division, and holds a BA inPolitical Science and Philosophy fromthe University of Chicago. Nell is a grad student in Computing atImperialCollege London focusing onhow LLMs could enable scalable vishingfraud. She previously helped AI teamsbuildreliable products at AI agentplatform V7 Labs, and has a first classBA from Oxford University. stateof.ai 2025 Artificial intelligence (AI) is a multidisciplinary field of science and engineering devoted to creating intelligent machines. AI acts as a force multiplier for technological progress in our increasingly digital, data-driven world. This is becauseeverything around us, from culture to consumer products, is ultimately a product of intelligence. Now in its eighth consecutive year, the State of AI Report is the most widely read and trusted open-access publicationtracking progress in artificial intelligence. Consider it a curated compilation of the most significant and thought-provokingwork from the past 12 months. Our goal is to inform and shape an ongoing conversation about the state of AI, where thefield is heading, and what its developments mean for the future. This year’s report examines six key dimensions of the AI ecosystem: -Research: Technological breakthroughs and their capabilities.-Industry: Areas of commercial application for AI and their business impact.-Politics: Regulation, economic implications, and the evolving geopolitics of AI.-Safety: Efforts to identify and mitigate catastrophic risks that highly capable future AI systems could pose.-Survey: Findings from the largest open-access survey of 1,200 AI practitioners and their AI usage patterns.-Predictions: Our outlook for the next 12 months, alongside a review of our 2024 forecasts to keep us accountable. Produced byNathan Benaich and Air Street Capital team. stateof.ai 2025 Definitions Artificial intelligence (AI):a broad discipline with the goal of creating intelligent machines, as opposed to the natural intelligence of humans andanimals. Whileartificial general and super intelligence (AGI and ASI)are terms that don’t agreed upon definitions, we use them to describemachines that could match (AGI) and then exceed (ASI) the full range of human cognitive ability across all economically valuable tasks. AI Agent:an AI-powered system that can take actions in an environment. For example, an LLM that has access to a suite of tools and has to decidewhich one to use in order to accomplish a task that it has been prompted to do. AI Safety:a field that studies and attempts to mitigate the risks (minor to catastrophic) which future AI could pose to humanity. Context window: The number of input tokens that an LLM model can attend to while answer a user’s prompt. Diffusion: An algorithm that iteratively denoises an artificially corrupted signal in order to generate new, high-quality outputs. In recent years it hasbeen at the forefront of image generation and protein design. Environment: The world an AI agent acts in. It receives the agent’s actions and returns the next observation and often a reward (i.e. a signal of theaction being good or bad). In this context,trajectoriesare the time-ordered record of an agent’s experience in an environment, typically tuples like(observation/state, action, reward, next observation) from start to finish. These trajectories are used for RL. Function calling / tool use: Structured calls that let models invoke APIs, search, code, or calculators with typed arguments and schemas. Generative AI:A family of AI systems that are capable of generating new content (e.g. text, images, audio, or 3D assets) based on 'prompts'. Graphics Processing Unit (GPU):the workhorse AI semiconductor that enables a large number calculations to be computed in parallel. stateof.ai 2025 Definitions (Large) Language model (LM, LLM):a model trained on vast amounts of (often) textua