+1 to Foundation Models. I don't share your concerns about LLMs, though, and often refer to the future involving LXMs where X could be images, audio, bioinformatics data, etc