Echinosum
Posts: 604
Joined: 1/28/2021 From: Buckinghamshire, UK Status: offline
|
quote:
ORIGINAL: recotte Swiss scientists trained an AI on the chemical signatures of 73 wines from seven particular estates, then introduced seven new wines. Then they repeated the process 50 times. The algorithm identified the correct chateau every time, and got the vintage — the year of production, out of a possible 12 — right 50% of the time. It noted that wines from the right bank ... were chemically distinct from those on the left bank. It’s “coming close to proof” that local soils, climate, and traditions — known as the terroir — affect the flavor of wine, as wine buffs have long claimed, one researcher told New Scientist.” So it can tell 7 estates apart, if it knows it has to be one of those 7. Bordeaux has a lot of variety, more so than say red Burgundy which uses only one grape. If chosen to be as distinctive as possible in terroir and typical blend profile, I'm not surprised. Chemical analysis can easily distinguish wines from different grape varieties. So of course left and right bank wines are chemically distinct, just because of their quite different varietal blend profiles - very little Cab Sauv is used on the right bank. That would be enough to make them chemically very different. Even in the left bank, there are wines that are typically very high in CS, and those that are generally rather lower, and I expect those are reliably distinguishable for that reason, if those are the only options available to you to choose from. I'd be a lot more impressed if it reliably distinguish 70 estates. If it could tell the various chateau that use similar blend profiles apart from different areas. If it distinguish estates using similar blend profiles from the same area, like being able to distinguish the various different estates of Pauillac, or different merlot-dominated right bank wines. Of course, its hard to prove it is a difference in terroir rather than wine-making practice. If it could tell vineyards in Burgundy apart, despite them being from different makers, that would be a much stronger clue. AI isn't very good at telling us why it makes its choices, but maybe if could be so persuaded we might be able to tell the difference. Someone did an experiment once chemically testing grapes of the same variety grown in different soil types. It found consistent differences, that were due to the drainage structure of the soil. It did not find any differences from the chemical makeup of the soil, ie what kind of rock it was based on. Nevertheless that is kind of sufficient to prove that terroir exists, but perhaps not as richly as might be claimed. But the experiment had limitations, and there may be more to discover. Just because it was unable to find something, doesn't mean that it doesn't exist.
_____________________________
A discriminating palate can be a curse.
|