Technology & Medicine › Internet, genomes & AI › Deep learning (modern AI)
Large language models
2018 (GPT); ChatGPT November 30, 2022· Invention or discovery· Communication & computing
Well documentedSolid records, physical evidence or scientific consensus.
AI systems trained on vast amounts of text to predict the next word, which lets them converse, write and code.
See it on the Technology & Medicine map →How it works
A large language model is a transformer network trained on trillions of words of text to predict the next word. At enough scale this yields fluent writing, translation, coding and reasoning-like answers. OpenAI's GPT and Google's BERT (both 2018) started the trend; OpenAI's ChatGPT, released November 30, 2022 and tuned with human feedback, reached an estimated 100 million users in two months. Rivals include Google's Gemini, Anthropic's Claude, Meta's Llama and China's DeepSeek.
Why it mattered
They made computers usable through plain conversation, putting writing, coding and research help into everyday tools, and set off one of the largest investment booms in tech history.
Think of it like
Like an extremely well-read assistant who has absorbed millions of documents but can still misremember a citation with full confidence.
How we know
Published model papers and benchmark tests; usage estimates from analysts such as UBS; company disclosures; independent evaluations that also document confident errors, called hallucinations.
Still unknown or debated
Whether LLMs truly understand or reason, how far scaling will keep improving them, and their effects on jobs and misinformation are all hotly debated.
Inside this (1)
Connections (3)
▸Alan Turing proposed the test now applied to Large language modelsMostly agreed
The connection
In his 1950 paper 'Computing Machinery and Intelligence', Turing proposed the 'imitation game': if a judge chatting by text cannot reliably tell a machine from a person, the machine can fairly be said to think. For 70 years no program passed convincingly; large language models are the first to hold long, humanlike conversations, reviving the test.
When and how
Turing at the University of Manchester, 1950, in the journal Mind; modern tests by researchers such as Cameron Jones and Benjamin Bergen at UC San Diego, 2023–25.
How sure are we?
Turing's paper is published; in a 2025 UC San Diego study, judges picked GPT-4.5 as the human 73% of the time in five-minute chats.
Debates and open questions
Many experts argue that passing the test shows convincing imitation, not real understanding; whether it was ever a good measure of intelligence is debated.
▸Deep learning (modern AI) led to Large language modelsWell documented
The connection
Large language models are deep learning applied to text at enormous scale: the same layered neural networks, adjusted by backpropagation, run on clusters of graphics chips. Researchers found that bigger models trained on more data improved predictably ('scaling laws', 2020), which drove the jump to models with hundreds of billions of internal settings.
When and how
Word-meaning models such as Google's word2vec (2013); OpenAI's GPT (2018) and GPT-3 (2020); scaling-law studies at OpenAI (2020) and DeepMind (2022).
How we know
Model papers describe the deep learning methods used; scaling-law papers chart performance rising with model size and data.
Debates and open questions
Whether scaling alone can keep improving models, or new ideas are needed, is debated as high-quality text data runs short.
▸Transformer architecture underpins Large language modelsWell documented
The connection
Nearly every large language model is a stack of transformer layers, often dozens to over a hundred deep; GPT stands for 'generative pre-trained transformer'. Because transformers train in parallel, labs could feed them trillions of words, and the architecture's predictable gains with scale made multi-billion-dollar training runs worth the bet.
When and how
OpenAI's GPT in June 2018 and Google's BERT in October 2018, about a year after the transformer paper; ChatGPT in November 2022.
How we know
Model papers and technical reports, from GPT-1 to open models like Meta's Llama, describe transformer architectures explicitly.
Debates and open questions
Some researchers argue transformers' design limits long memory and reasoning, and hybrids are being tested; whether anything will dethrone transformers is uncertain.