After receiving an invitation to OpenAI's Dall-E, I tried things I set my imagination to. I asked it to "make a digital art of a cyberpunk person holding a digital brain as digital art. While similar images may exist from Googling, they do not hold all of the qualities at the same time, that is “being digital art”, “having a digital brain”, and “a cyberpunk” person simultaneously.
I was amazed and revelled in its beauty. But I thought any Tom-Dick-and-Harry equivalent of AI could do it, so I attempted to stretch it to its limits a bit. I tried to search for the same thing, but in the setting of an ancient Egyptian street. These were the results:
OK, perhaps I stretched it too far! But perhaps one day, and that one day will be very soon, these kinks will be worked out.
How does the AI learn how to draw?
Through various computer techniques and captioned image data, the AI for this type of program is able to predict the most likely image that could be outputted from a specific type of caption. AI like Dall-E have already processed millions of images in this manner!
What is useful about having this kind of AI?
Once it is advanced enough, it can speed up the process needed to design something needed for an artistic endeavour. It can bring into fruition that which you imagine but does not exist yet on a Google Search. It’s like a stock image generator. If you need to design a logo, it can help with that too!
What are the wider ranging issues and dangers with using this AI?
Some common dangers associated with faking anything really also exist with AI art. Images of various reportable offences on social media are also not allowed to be drawn on Dall E 2. You cannot have the AI try to draw a picture of anyone committing any sort of crime. While some AI art websites have taken measures to stop these ethical issues from happening, some have not. Oh technology, always a double edged sword…
There are also wider societal implications. Just like search engines, the AI is driven by the same biases people have when they use search engines. One could argue that certain biases may get pronounced by the use of this tech, but controlling for the biases, depending on who is doing the controlling could introduce a different type of bias itself.
Conclusion
There shouldn’t be a major fear with this type of technology. AI could be used to augment human output and even make it more efficient. The fear of AI currently is to make sure the dark side of humanity is not pronounced. Just like having children, we are the parents of AI and we are ultimately responsible for feeding it the information it needs to do good for humankind.