A year ago, OpenAI released its chatbot to the public: “We are excited to introduce ChatGPT to get users’ feedback and learn about its strengths and weaknesses” the OpenAI blog announced on November 30, 2022. The subsequent run on the prompt certainly taught the company a lot – about its tool and even more about its users.
I also learnt something. Namely, what a prompt is. And that generative AI is similar to autonomous driving: Great promise, but implementation is still a problem. ChatGPT feels a bit like the first generation of a lane keeping assistant. It was good until you really needed it. Then, however, it said goodbye.
As long as I’m on the highway with ChatGPT, the tool delivers: Outlines, ideas for presentations, blog posts and social media post suggestions. Everything is familiar and predictable – the crash barriers and median strip are always in view. However, as soon as the terrain becomes a little more demanding, inaccuracies, errors and misfires become more frequent. Sometimes the system switches off without warning.
Of course, autonomous driving cannot be achieved with the lane departure warning system alone. Nor can a tool like ChatGPT do the writing, photography or painting for us. However, there is now a wealth of AI systems that assist us with creative processes. For example, searching for relevant content, organising knowledge, generationg ideas, producing and publishing content.
A year ago, OpenAI gave us a practical AI assistant in the form of ChatGPT. It has found a permanent place in my work. But I think that it will be a while before we can let go of the steering wheel and just type in the destination.