moat
In plain wordsA lasting business advantage that keeps competitors out, like a castle's moat; in AI, the question of whether any company can stay ahead when rivals and free models catch up quickly.
In business slang, whatever makes a company hard to dislodge: an advantage durable enough that rivals cannot easily copy it or lure away its customers. In AI the word frames a recurring debate over whether model developers can sustain an edge when capabilities diffuse quickly through publications, distillationknowledge distillation, n.: Training a smaller or cheaper 'student' AI to copy the behavior of a larger 'teacher' AI, so the student keeps much of the teacher's skill while being faster to run., staff movement and open-weight modelsopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private.. Candidate moats include proprietary data and user-feedback loops, distribution through existing products, access to computecompute, n.: The raw computing power used to train and run AI, usually measured by how many specialized chips are used and for how long. With data and algorithms, it is a main driver of AI progress. and capital, brand, and the engineering of products around models; skeptics argue that thin applications (see wrapperwrapper, n.: A product built mostly on top of someone else's AI model, adding a simple interface or a few instructions; often used as a put-down implying the product is easy to copy.), and even frontier modelsfrontier model, n.: One of the most capable AI models in existence at a given time, built by a handful of leading labs at great cost, and the main target of AI safety rules. themselves, have little protection from fast followers. The term became a fixture of AI commentary after a leaked 2023 Google memo argued that neither Google nor OpenAI had one.
Their moat isn't the model; it's the two hundred million people who already open the app every day.
Origin
Warren Buffett had popularized the moat metaphor for durable business advantages ('economic moat') by the 1990s. The AI usage took off in May 2023 with 'We Have No Moat, And Neither Does OpenAI', a leaked internal memo later attributed to the Google engineer Luke Sernau and published by the newsletter SemiAnalysis, which argued that open-source models were rapidly closing the gap.
Related
- wrapperwrapper, n.: A product built mostly on top of someone else's AI model, adding a simple interface or a few instructions; often used as a put-down implying the product is easy to copy.
- GPU poorGPU poor, adj.: Slang for researchers, startups or countries without access to the huge numbers of expensive chips needed to build the most advanced AI, as opposed to the GPU-rich big labs.
- open-weight modelopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private.
- frontier labfrontier lab, n.: One of the handful of companies building the most advanced AI models of the moment, spending billions of dollars on computing power to train them.
- knowledge distillationknowledge distillation, n.: Training a smaller or cheaper 'student' AI to copy the behavior of a larger 'teacher' AI, so the student keeps much of the teacher's skill while being faster to run.
- AI-nativeAI-native, adj.: Describes a product, company or person built around AI from the start, rather than one that added AI features to older ways of working.
Sources
- Sernau, L. (attrib.) (2023). We Have No Moat, And Neither Does OpenAI. Leaked Google document, published by SemiAnalysis, 4 May 2023.
Mentioned in
AI-nativeAI-native, adj.: Describes a product, company or person built around AI from the start, rather than one that added AI features to older ways of working., CUDACUDA, prop. n.: Nvidia's software system for programming its graphics chips to do general calculations. Most AI software is built on it, which is a big reason Nvidia chips dominate AI., wrapperwrapper, n.: A product built mostly on top of someone else's AI model, adding a simple interface or a few instructions; often used as a put-down implying the product is easy to copy., data flywheeldata flywheel, n.: A cycle in which a product's users generate data that improves its AI, the better AI attracts more users, and those users generate still more data., GPU poorGPU poor, adj.: Slang for researchers, startups or countries without access to the huge numbers of expensive chips needed to build the most advanced AI, as opposed to the GPU-rich big labs., reverse acquihirereverse acquihire, n.: A deal in which a big tech company hires a start-up's founders and top staff and pays to license its technology, without buying the company itself.