An amateur in photography, a beginner in AI arts

Photography isn’t exactly my hobbie – I engage in it to add an extra elemet of joy to my outdoors activities, and give my wife some motivations to join me. An ‘amateur photograher’ is is how I title myself when explaining to other hikers why I carry the expensive gears on the trail. However, one thing that I am always hesitant to disclose to others is that I earn a living through machine learning, because my expertise in ML is only valuable if I have access to an institution’s data. The emergence of generative AI may have altered this dynamic, but as a newcomer to this area, I am still exploring its potential.

Pictures I took using SONY A7III + G-master 24-75 mm F/2.8 lens V.S. pictures I created using prompt engineering on DALL.E

DALL.E is a multi-modal version of GPT-3 trained on image-text pairs on the internet. A particular contrastive language-image pre-training schema was employed to rank DALL.E’s output to select the more appropriate results. A diffusion model was used on top of DALL-E’s output to generate more vivid pictures.