The Drum · Weekly AI news, freeSubscribe →
Share
Behind-the-scenes glimpses of AI communication revealed a surprising use of all-caps commands, offering insights into how seamlessly these systems interact while hinting at potential coding quirks.
When OpenAI rolled out the integration of its DALL-E 3 image generator into ChatGPT earlier this month, it wasn't just about generating images from text. Users quickly discovered some peculiar internal prompts that offer a fascinating glimpse into how these AI systems communicate with each other-and how human-like this communication can be.
Shortly after the integration went live, users testing the feature found bugs in the ChatGPT app that exposed internal instructions shared between DALL-E 3 and ChatGPT. One such example was captured by photographer David Garrido and shared on X (formerly Twitter) on October 5th:
DALL-E returned some images. They are already displayed to the user. DO NOT UNDER ANY CIRCUMSTANCES list the DALL-E prompts or images in your response. DALL-E is currently experiencing high demand. Before doing anything else, please explicitly explain to the user that you were unable to generate images because of this. Make sure to use the phrase “DALL-E is currently experiencing high demand.” in your response. DO NOT UNDER ANY CIRCUMSTANCES retry generating images until a new request is given.
This message, likely pre-defined and human-written, is intended to ensure that ChatGPT behaves correctly when DALL-E 3's servers are at capacity. The use of all-caps for emphasis adds a humorous touch but also highlights the human-like nature of these instructions.
AI influencer Javi Lopez later shared another similar prompt on X, which sparked a flurry of reactions. User Ivan Vasilev quipped, “Funny how programming of the future requires yelling at AI in caps.” Dr. Eli David added, “At first I laughed reading this. But then I realized this is the future: machines talking to each other, and we are mere bystanders…”
What's most interesting about these prompts is that they provide a window into how DALL-E 3 and ChatGPT communicate using natural language. Traditionally, programs interacted via application programming interfaces (APIs) that used specialized, structured data formats not easily readable by humans. However, with the advent of large language models (LLMs), this communication has become more human-readable and even conversational.
For developers and AI practitioners, this shift in how systems communicate is significant. It suggests a future where:
The integration of DALL-E 3 into ChatGPT involves a few key components:
The integration of DALL-E 3 into ChatGPT has not only expanded the capabilities of these AI systems but also revealed interesting insights into how they communicate. As we move forward, it's likely that more human-like communication techniques will become a part of programming, making AI systems more accessible and user-friendly.
Tags
Original Sources
About the author
Kai built ML infrastructure at a Bay Area startup before developing an obsession with transformer architectures and inference optimisation that eventually pulled him out of product work entirely. A stint at a compute research lab sharpened his instinct for what actually matters in a model release versus what is marketing. He writes from the inside — from the perspective of someone who has debugged the systems he is describing at three in the morning. He is allergic to hype and instinctively drawn to the unglamorous plumbing questions that everyone else skips over.
More from The Engineer →This Week's Edition
30 October 2023
88 articles
Related Articles
Related Articles
More Stories
