TechCrunch publishes AI glossary to explain LLMs, AGI, and other terms
The Story
TechCrunch has published a glossary aiming to define common AI terms such as LLMs, AGI, and chain-of-thought reasoning. The document is described as a living document that the outlet updates regularly as the field evolves. According to the glossary, artificial general intelligence (AGI) refers to AI more capable than the average human at many tasks, with definitions varying among organizations. The glossary also explains AI agents as tools that perform series of tasks on behalf of users, though the term may mean different things to different people. Other terms defined include chain-of-thought reasoning, coding agents, compute, deep learning, diffusion, distillation, fine-tuning, GANs, hallucination, inference, and large language models. Hallucination is described as the AI industry’s term for models generating incorrect information, a problem thought to arise from gaps in training data. The glossary notes that distillation can be used to create smaller models from larger ones, but using it on a competitor’s model typically violates terms of service.
Key Facts
- Artificial general intelligence (AGI) refers to AI more capable than the average human at many tasks.
- OpenAI CEO Sam Altman described AGI as the “equivalent of a median human that you could hire as a co-worker.”
- OpenAI’s charter defines AGI as “highly autonomous systems that outperform humans at most economically valuable work.”
- Google DeepMind views AGI as “AI that’s at least as capable as humans at most cognitive tasks.”
- AI agents are tools that use AI to perform series of tasks on behalf of a user, but the term may mean different things to different people.
- API endpoints are “buttons” on software that other programs can press to make it do things.
- Chain-of-thought reasoning breaks down problems into smaller steps to improve answer quality.
- Coding agents can write, test, and debug code autonomously.
- Compute refers to computational power for AI models, often shorthand for hardware like GPUs, CPUs, and TPUs.
- Deep learning uses multi-layered artificial neural networks inspired by the human brain.
- Diffusion is the tech behind many generative AI models, learning reverse diffusion from noise.
- Distillation extracts knowledge from a large teacher model to train a smaller student model; using it on a competitor’s model typically violates terms of service.
- Fine-tuning further trains an AI model for specific tasks using new specialized data.
- GANs (Generative Adversarial Networks) use two neural networks to produce realistic data.
- Hallucination is the term for AI models generating incorrect information, thought to arise from training data gaps.
- Inference is the process of running an AI model to make predictions.
- Large language models (LLMs) are used by AI assistants such as ChatGPT, Claude, Gemini, Llama, Copilot, and Le Chat.
Conflicting Reports
No conflicting reports identified in the source article.
Still Unclear
No open questions identified in the source article.
Misconceptions
No widespread misconceptions addressed in the source article.
Key Figures
- Sam Altman, CEO of OpenAI
- OpenAI
- Google DeepMind
Sources: TechCrunch
