Ask HN: Continue with SaaS type platform, or pursue AI?
My key takeaway after spending a couple months on it is that inaccuracy is really the biggest killer.
Here are some key takeaways after talking to some people as well as doing my own projects:
1. 1hr meeting's AI gen minutes (zoom/teams/etc). It makes a dozen or so points but some of them are not relevant to the meeting at all, and some folks have heavy accents and their input is mostly incorrectly captured. Folks ended up just using manual minutes again because these are important for their workflow.
2. Creating embeddings of files/documentation/messages/etc and then searching on them seems to work well. For the most part, it seems these can be 60-80% accurate because the alternative is even worse (try searching a huge slack for a file, or even software made for file management). However, when it doesn't work, it is really bad because I get wildly unrelated results.
3. Not very accurate for anything involving calculations or open ended questions about numeric data, granted this is a known limitation. Creating a graph or something like that? Sure. Asking it about those figures or trying to derive some additional information seems to be hit or miss.
So I'm a little stuck. I feel like if I stick with a SaaS platform, it won't be appealing in the current landscape where everyone is hammering AI into their products. Personally I find most of the gen AI not that useful - like when I open up a project management tool and the description box has an AI button.
NL input seems to instill a sense of false confidence - I will get what I ask. But when it doesn't work, especially after effort placed into the prompt, people just revert to the vanilla experience.
So what am I missing? Am I just thinking the wrong way? Maybe I just suck at setting vision for the future based on my own experience with these tools.
Any thoughts would be helpful. Thanks!
0 comments
[ 2.9 ms ] story [ 9.0 ms ] threadNo comments yet.