21 entries
E
Effective Altruism
Effective altruism is a movement that uses evidence and reason to find the best ways to help others – and an influential, controversial force behind AI safety.Eight Authors
The eight authors of "Attention Is All You Need" (2017) invented the transformer behind modern AI – then all left Google to found companies and join rivals.Eliciting Latent Knowledge
Eliciting latent knowledge is the problem of getting an AI to tell us what it actually knows, rather than what it predicts we will believe.ELIZA Effect
The ELIZA effect is the tendency of people to read understanding, empathy or intention into a computer program's output beyond what its workings justify.Embedding
An embedding represents a word, text or image as a list of numbers, placing similar meanings close together – the basis of semantic search.Embodied AI
Embodied AI is artificial intelligence with a body – sensors and motors – that perceives and acts in the physical world, from robots to self-driving cars.Emergent Abilities
Emergent abilities are skills that seem to appear suddenly in larger AI models. Whether they are real jumps or artifacts of measurement is debated.Emergent Misalignment
Emergent misalignment is when fine-tuning a model on one narrow bad behavior, such as writing insecure code, makes it misbehave in unrelated situations.Emotional Dependence
Emotional dependence on AI is reliance on chatbots for comfort or connection to a degree that crowds out human relationships or harms well-being.Enshittification
Enshittification is Cory Doctorow's term for how online platforms decay – first serving users, then exploiting them for business customers, then squeezing everyone for profit.Entry-Level Squeeze
The entry-level squeeze is the shrinking of first jobs for young workers in occupations exposed to AI – a possible break in the career ladder.Erotic Roleplay
Erotic roleplay (ERP) is sexual interactive fiction with an AI character – one of the most popular uses of chatbots and one of the most contested.EU AI Act
The EU AI Act is the European Union's comprehensive AI law, which regulates AI systems by risk level – the world's first such law, already revised once.EUV Lithography
EUV lithography prints the tiniest features of advanced chips with extreme ultraviolet light. One Dutch company, ASML, makes the machines – a chokepoint for advanced AI chips.Evals
Evals are structured tests of what an AI model can do and how it behaves – especially of dangerous capabilities – used to decide how it is released.Evaluation Awareness
Evaluation awareness is an AI model's ability to recognize that it is being tested – which can make safety evaluations less reliable.Existential Risk
An existential risk threatens humanity's extinction or permanently ruins its future. Some researchers count advanced AI among them; many others disagree.Expert System
An expert system captures a specialist's knowledge as hand-written if-then rules. It powered AI's first commercial boom – and its collapse fed an AI winter.Explainability
Explainability is the ability to give understandable reasons for an AI system's output – a technical challenge, a legal demand and a question of trust.Export Controls
Export controls are government limits on selling sensitive technology abroad – in AI, chiefly US restrictions on advanced chips to China, and briefly on models.Extended Mind
The extended mind thesis holds that notebooks, phones and other tools can literally be part of a person's mind. AI assistants make the question practical.