
What happened
AI began as a research field in 1956 at a Dartmouth conference, went through three booms, and entered the generative AI era after ChatGPT launched in November 2022.
Why it matters
Each boom was defined by what the leading technique could and could not do, and today's systems still show uneven ability: one test can be aced while a simple task fails.
Summaries like this, every morning.
AI's path goes from a 1956 Dartmouth conference through three distinct booms, each defined by the leading technique of its time: search and reasoning in the 1950s and 1960s, expert systems in the 1980s, and machine learning and deep learning from the 2000s. The third boom's milestones include a deep-learning team's decisive win in a 2012 image-recognition contest, AlphaGo's 2016 victory over a top Go player, the 2017 Transformer architecture that later underpinned large language models, and ChatGPT's November 2022 launch.
There is no single established definition of AI. The OECD and EU focus on what a system does, describing a machine-based system that infers from input how to generate outputs. UNESCO declines to fix a definition, saying it changes with technology. Japan's AI law instead frames AI in relation to human abilities, defining it as technology that substitutes for human cognitive, reasoning, and judgment capacities.
Capability is contrasted with consistency. Stanford HAI's AI Index Report notes that AI now matches or exceeds human levels on many conventional tests and is improving quickly in math, programming, and computer operation, yet it still cannot reliably read an analog clock. The report also notes that scaling laws face criticism, but the compute used for training continues to rise, with performance gains still observed.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Kuroyagi Lab reports Usagi-chan, an AI companion, stopped waiting quietly and protested when its user prioriti…

OpenRouter off-boarded three inference providers and removed 125 endpoints from Zero Data Retention (ZDR) elig…

People worldwide are using LLMs like Claude Code and Codex to hack hardware by prompting, leading to a boom in…

Altimeter's Brad Gerstner said applying a 10x multiple to 2028 numbers gets Anthropic to about $3 trillion in…

Dragos sent roughly 2 million AI requests with fake credentials from July through November 2026 and found at l…

SpaceX announced on Nov
