[numlock86]

Learn what ML and AI "really" is (start with neural nets, simple classifiers, get into GANs etc.) and just experiment with PyTorch or whatever else you like. Pick up Ray at some point and just go on from there. Rebuild models from papers and understand them. I wouldn't recommend "toying around" with ready to use stuff like premodeled and/or pretrained models (especially stable diffusion, the hype is really big right now) right at the beginning because that won't really teach an understanding of what is going on exactly. Compare with "script kiddies" as people called them back the days: They just use ready-to-use tools but have no idea what is happening and hence never surpass those tools or the level they are on, unless they bother to actually start at the beginning some day after all ... Like just using Linux won't make you an OS developer. What's all that generated prompts and content worth if you can't even tell at first glance if 0.05 or 0.0003 is a better learning rate for Adam optimizer?