I've written about AI before in an article titled, Is AI Bad For Landscape Photography?, and my conclusion was that it could be used for good practical application in terms of printmaking. Topaz Labs is a company that specializes in innovative photography software to overcome technical limitations. Many photographers use their services and I doubt that most people have any ethical concerns around its' use as the tools are not designed to deceive or be a replacement for creative input. This article is about a more controversial application of artificial intelligence however.
The current investment hype in Silicon Valley is around generative artificial intelligence. While the concept of AI has been around for decades, generative AI is about creating new content based on imagery, sound and text. This is a relatively untapped market that can have wide-ranging implications ranging from practical, philosophical, economics to ethics. Google and Microsoft are two tech giants that have have recently ramped up their investments around this type of technology. In Google's case, they appear to have been caught flat-footed here as they likely see ChatGPT, a recently-released chatbot, as a threat to their core search business. Microsoft has decided to get ahead of the curve by investing $10 billion into OpenAI which could possibly be integrated into their Bing search engine among other things.
What is DALL·E 2?
DALL·E 2 is a large-scale language model developed by OpenAI. It is a follow-up to the original DALL·E model, which was released in December 2020. DALL·E 2 is a transformer-based model, like the original DALL·E, but it has been further fine-tuned to generate more diverse and coherent outputs.
One of the main features of DALL·E 2 is its ability to generate images from text descriptions. This is known as "image synthesis", and it allows DALL·E 2 to create new images from text descriptions that have never been seen before. DALL·E 2 is also able to perform a wide range of language-based tasks, such as text-to-speech, language translation, and text summarization.
(The above section was written by ChatGPT based on several prompts by me.)
It's pretty wild that software can learn from crawling hundreds of millions of images on the internet to create entirely new images on demand. Out of curiosity I have decided to test this out for myself. I've included AI-generated photography & illustrations that I made in DALL·E 2 and Midjourney to accompany this article. Each took about ten seconds to generate. Is this art? You be the judge.