Dave Asprey's Bulletproof podcast just featured Naveen Jain and Viome
. COI alert: Dave Asprey is an investor in Biome (which he does state along the way). I found it short on discussing just how the process works beyond references to AI, big data and the like. This page
at their website describes it a bit further.
Even if this technology may still be developing, is it really the way of the future? If humans
are designing the AI, can it really be that accurate in the insanely complicated world of the human/biome body
Is any AI really that accurate? How is AI being used, accepted and relied on today in other areas? They toss out 'AI' like that term alone should inspire confidence, but someone is writing and editing the AI algorithms.
Does anyone have any updates on their experience? My guess is we're 'not there yet', but some day, if we haven't already, we'll cross that threshold in the form of an evidence base in the scientific literature, and it'll be long after it actually became useful to prevent chronic disease.
The future of this tech is very promising. The current state of the art, less so, for my money.
After you sign up for the service, take the test, and run the math, what is it going to tell you to eat aside from "a bunch of organic plants and a little bit of high-quality meat / seafood" ? Perhaps it can make some top-down rule-based recommendations on those plants and steer some people further away from animal consumption, but those recommendations should have quite a bit to do with an individual's lifestyle and metabolic health, which I don't believe the current system can fully factor into the math. And again, it's difficult to imagine the system recommending more than "a bunch of nutrient-dense plants associated with health benefits." If it recommends that I don't eat sweet potato, but I know that I'm going to workout in the morning, I have a low A1C, and can handle it, who do I listen to? Or, if it recommends I eat more apples, but I know that I have a hard time digesting them, who do I listen to? If it recommends that I eat whole wheat, where does that recommendation originate?
Perhaps it can recommend some particular probiotics for some individuals, but I don't think the tech is able to really make a determination on why a particular probiotic would be health-promoting for one individual, but not for other individuals.
As someone who has completed a few Kaggle.com challenges (although admittedly, hasn't earned any medals in competition), the challenge as I see it, is acquiring the data. Lots, and lots, and lots of data. There's maybe some data that suggests that people who eat more fruits have less cardiovascular disease, and those who produce more TMAO (hence, meat + eggs) suffer from greater cardiovascular events. But, imagine if we had 1m data points for how people felt after eating an apple and 1,000 other foods, or how much deep sleep they were able to get after consuming some kind of tea at some point in the day, the recommendations would be so much more relevant and actionable. Imagine if we had 1m continuous heart rate heart-rate variability samples stratified into 10 patterns of eating along with subjective data points of stress / anxiety. From this starting point, we could start to make some really cool predictions. Then, we could train different AI agents to compete against each other with the recommendations for improving scores of subjective wellness + objective markers of health. But we currently don't have any data like this, or better yet, a system in place for incentivizing this sort of data production and mining collaboration.
To be honest, even with amazing levels of data acquisition and prediction modeling, it's fairly difficult to make large lifestyle changes. I wear an Oura ring that tracks every footstep during the day along with my respiratory rate through the night to try to encourage better sleep habits. Quite a lot of data is being crunched in the cloud to try to help me better recover through the night and push through the day. After sunset, it starts spouting advice on winding down. Yet, here I am, after 5 hours of sleep, drinking black coffee with a readiness score of 59.
A lot of the time, I think we understand what we should be doing, but we need some really convincing
recommendations more than recommendations