Jonathan Beri @jonathanberi.com · Dec 3 📚 Fascinating paper! tl;dr Google Research found that using *images* of Timeseries data is more effective and cheaper than working with raw data.
8 likes 3 replies
Replies
📈 Why is that important? Most of IoT is Timeseries and so far most models (LLM and others) are not generalizable or effective. This could significantly change that.
Anecdotally I’ve seen this to be true haha. I once worked on a project where an intern started taking screen shots of traces and feeling those into a neural net and it worked better than any of the hand-coded analysis for a certain kind of problem!
That’s super interesting! And a little worrying. 😜 By the time I’ve figured out how to plot something clearly I usually know what to look for. What I really want is help finding information in data when it’s not obvious in a plot…