We know: usefully coding robots is 100x harder than building them. Now learning: getting intel out of data is 100x harder than gathering it
Conversation
Replying to
not sure I agree. Data engineering, cleaning, and munging is about 80% of the work.
2
1
Replying to
Most insights are trivial, flawed or both and as yet, can't even compete well even with intuitive, low-data human narrative.
1
Replying to
Not sure what realm you're talking about. A/B testing and recommendations, two enourmous data domains, are not trivial nor intuitive
3
Replying to
In 20-30 years, when the tech matures, it'll put low-data narrative out of business altogether. Like drones, driverless cars
1
Replying to
I think by "data" you mean "BI", which I've never seen to require more than a few dashboards. Hard pressed to name uses for ML in BI.
3
1
Replying to
there's a Moravec's paradox in data too. The seemingly hard-to-mine intel is easy, seemingly easy intel is hard.
1
1
Replying to
Many "obvious" questions are so hard are we pretend they're the wrong/unimportant/ill-posed instead of admitting we can't answer
1
Replying to
that's an interesting point. I definitely wouldn't mind a machine telling me what I should be when I grow up...
1
Replying to
exactly. Yet, it's actually in the addressable set. I'm fiddling with a lifelogger for just such q's. github.com/vgururao/lifel

