The use of natural language in these prompts highlights a shift in how programs communicate with each other, moving from specialized, structured data formats to conventional English. This is made possible by large language models (LLMs) like GPT-4, which are trained on hundreds of millions of documents from the web. The use of polite language and all-caps for emphasis in the prompts is likely a reflection of the human communications used in the training data. The discovery has led to discussions about the future of programming and whether we might all be "shouting" at our computers to get them to work better.
Key takeaways:
- OpenAI's DALL-E 3 AI image generator and ChatGPT communicate using human-like language, including the use of all-caps for emphasis.
- The use of natural language in AI communication is a shift from the traditional use of application programming interfaces (APIs) that often used specialized, structured data formats.
- Politeness and emphasis in AI communication can be effective due to the training data, which includes examples of human communications with polite language and emphasized sentences.
- AI writer and researcher Simon Willison suggests that in the future, we might use emphasis or 'shouting' in our communication with AI to get better responses.