DALL E & Chat GPT – how will they really impact creative
It’s fair to say that much recent mainstream AI news has focused on the pretty mind bending and exciting capabilities of these twins of content creation, DALL E, for imagery, and Chat GPT for textual content (including code). Both will take a textual prompt, with quite specific context, and output original creative content that in many cases is remarkably close to what we would expect to human output.
In the case of Dall E, multiple image options are offered, allowing the human in the loop to either choose one of those or further refine the brief.
One of the most interesting elements here for me, from a creative perspective, is for the prompt to indicate a stylistic approach. For example in the case of imagery specifying “in the style of Van Gogh”, or text, “as a Shakespearean sonnet” the outcomes can be remarkably true to this brief.
Does this mean creative professionals now have serious competition in the artistic realm, or even a threat to their very existence? Certainly, for some tasks i would say the former is true, but I still believe we are a bit away from the latter.
Not yet ready for real creativity but certainly for creative output
The potential of these tools to streamline process and stimulate creative has tremendous potential. When looking for imagery for this website there are a number of places where Dall-E generated images have created a shortcut to specific design output that would have taken hours for me to replicate, if at all.
To give an example, specifying particular items such as “clouds”, “computer binary”, “androids” or “data collection” allows the algorithm room to a number of stylistic options that, whilst may not always consitute the finished article, can consitute steps in that direction.
Additionally in the context of the written word it offers a tremendous shortcut to factually based content generation. Want a potted history of Coca cola, need to understand Mahatma Gandhi’s impact, or offer advice on fly selection for early season fishing, Chat GPT will offer all of these things in very short order. What is even more impressive is that if you validate that knowledge it is often remarkably accurate. However the rub is in that “often”. For quality content generation accuracy does matter, and once you veer into certain realms, medical or legal advice it can be literally life altering.
Where accuracy matters
Certainly where we have tested, and also explored others’s experiences online, those factual inaccuracies do occur. So proof reading is definitely required. Again leading us to the conclusion that the tool in it’s current state is a wonderful shortcut to content but if you value true accuracy it’s a shortcut not a finished product.
Of course that is not going to stop mass abuse by low quality content spinners and web spammers, unfortunately this is already coming to pass.
Coding is another area of Chat GPT strength with impressive output. However again this is a realm where accuracy can matter. Stack Overflow, where developers frequently turn with technical questions, has taken the decision to ban code answers generated by Chat GPT. The fact that these have even featured in the first place tells us that the quality of output has come a long way. We are well beyond basic chatbot style AI text applications, or replication of existing content and into the realms of genuine content creation.