Software update 19 is still in beta on google play store. Official release would come at end of next week, as we are weeding out bugs, with many thanks for the kind feedback from our beta testers. Sorry, it’s such a drag. We didn’t time it well with our previous schedule estimate.
Even our content team got a little grumpy, as they have finished porting the focus skills to audio and are impatiently waiting integration with the App. So we made a tiny compromise, a new ‘Focus Skill’ tab would be released along with software for both software update 19 and 20. And software update 20 would consequently move to mid September.
We get quite a number of backers writing to us, asking how does 'Train machine learning' of software update 20 work. How is self-report on emotion (which is notoriously unreliable) used to train machine learning? Wouldn’t it just skew machine learning? On the other hand, we get questions like, if I didn’t give any feedback to machine learning, how did it manage to give me good estimates in the first place.
The third question is easiest to answer. Machine learning uses similar mathematical techniques, as credit rating agencies use to rate institutions and individuals alike. For example, Moody downgraded the credit rating of University of Cambridge from ‘aaa' to ‘aa1’ last year. It didn’t require Cambridge University to give them a shoutout saying ‘hey, I’m less than impeccable, downgrade my credit rating’, nor would it be likely to heed any protest, if there is any, from Cambridge. It is just the way it is. What we do, what we do not do, is like our digital book. Machine learning is how machines like FOCI can learn to read and do something useful for us.
The second question is a difficult data engineering question. In short, self-report is unreliable, but can provide a huge boost to data when used judiciously. I.e. to use alongside counter metrics, to discount inconsistencies, and estimating the reliability factor of a user’s input to be used together with the input, a user’s self-report holds a huge amount of information that can be extracted.
The first question can be answered with the design of the 'Train machine learning’. For the emotion streaks of the past 12 hours, attribute how well they match your expectation, to 'Train' machine learning.
Mick & The FOCI Team
P.S. If you face any problems with shipment, message me at email@example.com and we’ll get it sorted.