This is very cool. And I think the format is potentially very useful. But the questions are not that good, because most of the answers don't perfectly fit the exclusive labels presented as options
For example:
Going 600 km for a family event. How do you travel?
* Overnight train
* Flight
* Drive
* Sleeper bus
To me the answer depends. It depends on things like what kind of family event it is, the urgency, who is going, who is coming with me, availability and timing. So essentially, there's a lot of missed context, that is not being captured or represented well enough to produce good predictions
Note: For many (most?) of these, I'm not sure how to answer them, and am choosing arbitrarily; there is a trend of multiple (2-3 generally) all seeming reasonable and/or similar, and I am going with the flow of which I choose.
I think it would be useful. You could try matching people here on HN.
If you want to match similar people, aim to make them as geographically as close as possible. What you really want to find is "uncommon commonalities" between people and build groups around those. So yes, like attracts like, but it can be surrounded by stark contrasts as well. You could grade a match - percentage different, percentage same and then add some strength/idiosyncrasy/geographical/group-match weighting.
The progress bar at bottom of questions: it says "keep going" but eventually it reaches the end and there is a page refresh without any user confirmation to do so. That can be confusing and distracting.
agree on slop, all the questions were generated by AI
this was a fixed set, wondering if I should try with fully generated questions and see how that goes
This is very cool. And I think the format is potentially very useful. But the questions are not that good, because most of the answers don't perfectly fit the exclusive labels presented as options
For example:
Going 600 km for a family event. How do you travel? * Overnight train * Flight * Drive * Sleeper bus
To me the answer depends. It depends on things like what kind of family event it is, the urgency, who is going, who is coming with me, availability and timing. So essentially, there's a lot of missed context, that is not being captured or represented well enough to produce good predictions
agreed, the experiment was focused on what agents can learn when they have no context but I get your point
Note: For many (most?) of these, I'm not sure how to answer them, and am choosing arbitrarily; there is a trend of multiple (2-3 generally) all seeming reasonable and/or similar, and I am going with the flow of which I choose.
yeah, the questions and options were also AI generated - didn't cover extremes or unique decision options fairly
I got some questions that I didnt understand what it actually meant, so I had to choose randomly for those. But other than that, this is cool
maybe because they were written in an Indian context
This should be used for match-making! Like attracts like.
Needs to be a greater delay or key-press upon prediction status. Can it give a report and assessment?
Also useful for recruiting. Match candidates with their potential managers.
good idea, will add reports/export later Do you think match making will actually be meaningful if the AI matches "similar" people?
I think it would be useful. You could try matching people here on HN.
If you want to match similar people, aim to make them as geographically as close as possible. What you really want to find is "uncommon commonalities" between people and build groups around those. So yes, like attracts like, but it can be surrounded by stark contrasts as well. You could grade a match - percentage different, percentage same and then add some strength/idiosyncrasy/geographical/group-match weighting.
The progress bar at bottom of questions: it says "keep going" but eventually it reaches the end and there is a page refresh without any user confirmation to do so. That can be confusing and distracting.
Some questions might be relevant only for Indian audiences.
it started out good, but transitioned to slop question/answers
agree on slop, all the questions were generated by AI this was a fixed set, wondering if I should try with fully generated questions and see how that goes