Ever Wondered What DeepSeek, Toasters, and Adam Smith Have In Common

DeepSeek, Toasters and Adam Smith… DeepSeek is a Large Language Model. Large Language Models predict the next word in a sequence, given the previous words in that sequence. A number of companies built Large Language Models that were very good at predicting the next words in a sequence. That enabled them to score highly on a selection of written tests. The first batch of companies to build large language models were primarily US-headquartered and they spent a lot of money building their models. DeepSeek is Chinese-headquartered and did not spend a lot of money building their model. Despite the low cost of development, DeepSeek’s model is nearly as good at predicting the next words in a sequence as the best US-made models when judged by how well they score on a selection of written tests.

Large Language Models themselves are remarkable. But DeepSeek shows that they can be built inexpensively. The exciting bit in AI is working out what to put in the models and what to do with the stuff you get out of the models. In other words, how you apply them in the real world. That is where all the value is going to be created as the models themselves become increasingly commoditized.

Large Language Models are a general-purpose technology like electricity. It is farcical to imagine that one country could monopolise the creation and distribution of electricity. The production of electricity was never exclusive to a single nation. Power stations were built worldwide.

Cheap, readily available electricity led to an explosion of technological innovation and industrial growth. Washing machines, refrigerators, vacuum cleaners, radio, television – electricity is everywhere in the modern world. There is no moat in the deep tech of AI – the barriers to entry are low and getting lower.

But AI will be in everything over the next decade, as electricity is today. An electric toothbrush is not an electric drill. They both use electricity but only one is good for cleaning your teeth.

Specialisation drives prosperity. Adam Smith worked this out in the 18th century and new technologies don’t change that fundamental truth.

As the AI-era unfolds we will see a wide range of specialised applications tailored to specific tasks and industries. They will use LLMs as a toaster uses electricity. Each application requires a unique approach based on what you are trying to achieve. And just as you probably wouldn’t use an iron to toast bread, you probably shouldn’t use a chat-bot to review legal documents.