Gemini fast: „Walking: It will take you about 45 seconds. You will arrive refreshed and full of steps, but you will be standing next to a high-pressure hose with no car to spray.“
Foundational models can work where so far „needs human intuition“ was the state of things. I can picture a time series model with large enough Training corpus being able to deal quite well with typical quirks of seasonalities, shocks, outliers, etc.
I fully agree regarding how things have been so far, but I’m excited to see practitioners try out models such as the one presented here — it might just work.
- energy management (shifting loads to times when energy is cheap) for consumer/commercial/industrial use cases
- energy markets, especially power trading: often highly algorithmic, and driven by models that turn fundamentals data (weather, calendar, …) into supply/demand predictions, and from there into price predictions
The same applies within large companies. If you’re within a business team, and you’re requesting work from a design team/engineering team/data science team, you’ll face the same issues with scope creep + churn + competing priorities etc.
I wouldn’t blame agencies for being bad, this is people being people plus a bit of other things. Anticipating and steering around/against these dynamics has been one of my biggest career learnings over the last years. The author has some good suggestions for how to do it — if you work in a large company, take another look and ask yourself if they don’t also apply to your work!
At least in e-commerce it’s very much possible. The company that I work for does such experiments regularly (there’s a dedicated Data Science team for measurement) and I’ve personally been involved in lift studies for Google Ads. They work, you just have to be careful with the ‘how much’ combined with ‘for what’.
Happy to chat with anyone who is interested in the topic (pfalke at pfalke dot com).
Haven’t had a chance to listen to the podcast, apologies if that made me miss the point of the parent post!