Several friends and colleagues have asked why Iโve been posting so often lately.
The honest answer: my account has not been hacked, and I didnโt suddenly start thinking. ๐๐ ๐ท๐๐๐ ๐บ๐ฎ๐ฑ๐ฒ ๐ถ๐ ๐บ๐๐ฐ๐ต ๐ฐ๐ต๐ฒ๐ฎ๐ฝ๐ฒ๐ฟ ๐ณ๐ผ๐ฟ ๐บ๐ฒ ๐๐ผ ๐๐ฎ๐ ๐๐ต๐ฎ๐ ๐ ๐๐ต๐ถ๐ป๐ธ.
As a non-native English speaker, writing a polished post on a professional platform used to take me hours. Most ideas never made it out of my head, because I decided to save my brain and hair for paper and proposal writing.
But with AI, my process now looks like this:
- I start with ๐ผ๐ป๐ฒ ๐ฐ๐ผ๐ฟ๐ฒ ๐ถ๐ฑ๐ฒ๐ฎ ๐ฎ๐ป๐ฑ ๐ฎ ๐ฟ๐ผ๐๐ด๐ต ๐ฑ๐ฟ๐ฎ๐ณ๐, sometimes even in Chinese.
- An LLM helps me ๐๐๐ฟ๐๐ฐ๐๐๐ฟ๐ฒ ๐ถ๐ ๐ฎ๐ป๐ฑ ๐๐บ๐ผ๐ผ๐๐ต ๐ผ๐๐ ๐๐ต๐ฒ ๐น๐ฎ๐ป๐ด๐๐ฎ๐ด๐ฒ.
- I go back and forth a few rounds, ๐ฑ๐ฒ๐น๐ฒ๐๐ถ๐ป๐ด ๐ฎ๐ป๐๐๐ต๐ถ๐ป๐ด ๐๐ต๐ฎ๐ ๐ฑ๐ผ๐ฒ๐๐ปโ๐ ๐๐ผ๐๐ป๐ฑ ๐น๐ถ๐ธ๐ฒ ๐บ๐ฒ, and sometimes words I donโt even recognize. (Iโm serious. I had to look up what โdelveโ means.)
- ๐ ๐ฐ๐ต๐ฒ๐ฐ๐ธ ๐ฒ๐๐ฒ๐ฟ๐ ๐ฐ๐น๐ฎ๐ถ๐บ. Sometimes the model makes my point sound more certain than I am, or adds arguments I never made. Those get cut.
Some people feel AI-assisted posts are less authentic. I see it differently. ๐๐ ๐น๐ผ๐๐ฒ๐ฟ๐ฒ๐ฑ ๐๐ต๐ฒ ๐ฐ๐ผ๐๐ ๐ผ๐ณ ๐ฒ๐ ๐ฝ๐ฟ๐ฒ๐๐๐ถ๐ผ๐ป, ๐ป๐ผ๐ ๐๐ต๐ฒ ๐ฐ๐ผ๐๐ ๐ผ๐ณ ๐๐ต๐ถ๐ป๐ธ๐ถ๐ป๐ด. The ideas, the judgment, and the responsibility for what gets published are still mine.
And step 4 is the same thing I keep writing about: ๐๐ ๐ผ๐๐๐ฝ๐๐ ๐ป๐ฒ๐ฒ๐ฑ๐ ๐๐ฒ๐ฟ๐ถ๐ณ๐ถ๐ฐ๐ฎ๐๐ถ๐ผ๐ป, whether itโs a single-cell analysis or a LinkedIn post.
For scientists who think clearly but write slowly in a second language, this is a real shift. ๐ ๐ฎ๐ป๐ ๐๐ผ๐ถ๐ฐ๐ฒ๐ ๐๐ต๐ฎ๐ ๐๐๐ฒ๐ฑ ๐๐ผ ๐๐๐ฎ๐ ๐พ๐๐ถ๐ฒ๐ ๐ฐ๐ฎ๐ป ๐ป๐ผ๐ ๐ฏ๐ฒ ๐ต๐ฒ๐ฎ๐ฟ๐ฑ.
Do you use AI in your writing? Where do you draw the line?
P.S. The downside of AI-assisted writing: if you meet me in person, you might wonder how the LinkedIn Shakespeare turned into the everyday regular normal guy next door. Same ideas, just fewer fancy words.