Would add to this list a pitfall of forgetting to step of the hype-train at times and see what you can realistically utilize NOW.
What would work perfectly today with the promise of being able to add more to it in the future as the technology takes leaps forward tomorrow.
Like bigger context (as an example) is cool and all, but one does get into same position as it is with some people and wealth: how much context is enough or are you using context as a crutch where alternatives are available which you could utilize already today (or perhaps you can reframe the problem you're solving). Is there even a problem to begin with?
We get excited on every new release / improvement where there are actually use cases for the last generation of models already that do not get the attention they deserve.
In the rush to adopt the latest generative AI technologies, it's easy to get swept up in the excitement and make mistakes.
However, understanding and addressing the typical hurdles in AI projects can increase your chances of project success.
#AI #datascience
XR development Artist_Sound Design_Créatrice de contenu_Communication_Web3
10mo🫀