I can't help but wonder if real systems have additional (perhaps subtle) signals, which can be provided to a neural network, which then outperforms these simple algorithms.
For example, customers arrive at the grocery store in clusters due to traffic lights, schools getting out, etc. Even without direct signals, a NN could potentially pickup on these "rules" given other inputs, e.g. time of day, weather, etc.
> For example, customers arrive at the grocery store in clusters due to traffic lights, schools getting out, etc.
You're kind of just describing seasonality components and exogenous regressors; RNNs do actually function quite well for demand forecasting of this type but even simple models (Holt-Winters or a Bayesian state space model or something) can be really effective
Fixed price works great if the client's expectations are easy to meet and you've done the work many times before in a fraction of the time that others would take by the hour.
It was only a few years old with a 1-2 dozen employees. Social networks hadn't been monetized in the way they are today. The market thought Facebook overpaid for Instagram.
I turned on search the web to compare vs google search + AI labs, and ran a battery of typical tests. Google was instantaneous vs gpt-4o which wrote out each letter.
Subjectively, I'm not switching for quick searched - google remains lightning fast and is good enough. But I already use gpt/claude/etc for conceptual searches and deeper analysis.
---
[used leica q3] ==> google (product listings and website; Chatgpt told me about the Leica q3 and mentioned ebay)
[value of mac air m1] ==> neither!! (google was useless videos and crap; chatgpt gave me a price range and useful explanation... which made no sense - used was the same price or higher than new!)
[vogue lyrics] ==> google wins (gave me the lyrics; Chatgpt whined about copyright restrictions and sent me to a youtube video)
[weather in nyc] ==> tie (both provided correct, rich detail about the current weather)
[root causes of ww1] ==> tie (both identified Militarism, Alliances, Imperialism, Nationalism, explained each and then mentioned the assassination of the Archduke as the triggering event)
[bohemia to midtown] ==> equally bad (both figured out that it's a request for local directions, but neither just gave me directions until I gave a specific destination)
[bohemia to penn station] ==> ??? (chatgpt correctly gave me bohemia ny where google picked some obscure local listing; otoh chatgpt wrote out directions where google gave me a nice map)
[btc to usd] ==> tie (both got today's price)
[what time is it in stockholm] ==> tie (both got it right)
[iphone 16 vs 14] ==> chatgpt wins (nice comparison; google didn't pop search labs and just gave me websites)
[ffmpeg to clip the last 3 secs of a video] ==> chatgpt ?! (I didn't love either answer TBH)
[456+789] ==> google (both gave the answer, but google included a nice calculator)
...and the stuff people really want:
[porn] ==> google (gpt whined about policy violations; google gave pornhub and other "useful" results)
For example, customers arrive at the grocery store in clusters due to traffic lights, schools getting out, etc. Even without direct signals, a NN could potentially pickup on these "rules" given other inputs, e.g. time of day, weather, etc.
?
reply