The recipe that artificial intelligence has recommended

[Willaim james]

Every Italian winery has a logbook that the winemaker regularly updates with information from the previous day and a record of the measures conducted. An experienced eye can deduce from this data a kind of recipe that was used to create a singular wine, which is obviously distinguished by a special and distinctive taste. In wine production, this is a difficulty of balancing taste, nature, and human expertise. As the variables vary, the outcome varies.

In light of this, Viberti has integrated all the information and recipes utilised over the previous ten years into the new 4.0 fermenters' management centre.
The AI may identify similarities across years in this way: "It does not offer answers but reminds us of which year we can be motivated by. As a result, Claudio Viberti noted, "on the one hand, we know which activities are correct to reproduce and which may not have produced the desired effects.

Winemakers will continue to judge fermentation kinetics by looking at the colour of the must and hearing it bubble, as they have always done, but now they may analyse the specifics with a more accurate historical memory: The goal of this tool's creation was to enhance the wines' quality and favour their expression with minimum intervention.

 

[winemaker81]

The AI may identify similarities across years in this way: "It does not offer answers but reminds us of which year we can be motivated by. As a result, Claudio Viberti noted, "on the one hand, we know which activities are correct to reproduce and which may not have produced the desired effects.

This gives me hope that the system will be used intelligently.

Note: There is no such thing as "artificial intelligence", as there is no intelligence in any of these program. The correct term is "machine learning", as these programs look for patterns in the data which they are fed, and make decisions that depend totally upon the correctness of the inputs, and the likelihood of identifying trends in the data.

If anyone has heard the phrase "Garbage In, Garbage Out"? That applies to machine learning.

But as I said, if this system is used to look for trends and make recommendations regarding where to look for similarities, it's a good thing.

 

[bearpaw8491]

Artificial Intelligence will never overcome Natural Stupidity!!!!!

 

[Jim Welch]

!Danger Will Robinson, Danger!
Unplug the PS now!

 

[Willaim james]

thanks for suggesting me

 

[ratflinger]

When machines actually think, it will be synthetic intelligence. AI is nothing but extremely good look-up tables.

 

[winemaker81]

Artificial Intelligence will never overcome Natural Stupidity!!!!!

I read an article about some of the answers coming out of ChatGPT ... I expect there is true competition in this arena!


@Willaim james, thanks for posting this. Detailed analysis of winemaking processes and results can help us to see patterns we can use. As @ratflinger indicated, ChatGPT produces really good lookup tables, and as long as we keep in mind that there is no intelligence involved, just pattern matching, it's a really useful tool.

 

[BigDaveK]

One could argue that primitive man had "tables" - this berry very good, that berry very bad. Somewhere along the timeline bits of information from different tables were combined in new ways. Is that where we are? I've seen examples of songs and poems written by AI and they're actually quite good. Where will it lead? Who knows.

Instead of an interesting tool it may become a crutch, enabling mass stupidity at worse or just plain lethargy.

https://www.zerohedge.com/personal-...ificial-intelligence-not-near-it-already-here
Like Bryan said, GIGO. But who's putting in the garbage?

https://www.theburningplatform.com/...troy-humanity-and-establish-global-dominance/

 

[winemaker81]

Like Bryan said, GIGO. But who's putting in the garbage?

The reverse is the problem -- how to keep the garbage out?

 

[winemanden]

Artificial Intelligence will never overcome Natural Stupidity!!!!!

Some deluded hacker will undoubtedly insert that into AI

 

[winemanden]

Natural Stupidity is putting the garbage in.

 

[BigDaveK]

Natural Stupidity has no limits. Read today that one of the climate "experts" claims that growing rice accounts for about 10% of methane production and that we should consider not growing it. Seems every day I read something and think, "You can't make this stuff up!"

 

[BernardSmith]

I think you are imagining that AI at both the input side and the output side has far more agency than in fact exists. The corporation behind Chat GPT apparently spent BILLIONS of $$$ purchasing billions of pages published online and that is the database that this program uses. When you ask it a question or interact with it, it uses a sophisticated statistical program to look for words you used in documents it can access and then its program is designed to suggest a string of words that make statistical good sense in light of this vast database. Until very recently (and I have not accessed the program for about 3 weeks), it could not understand a simple logical problem that a 9 year old could solve; if it could not find good material in its database it offered pure BS - a phenomenon known as "hallucinating". That aside, the larger problems are that any current AI program can only use data it has access to which is why a little while ago Amazon had to toss away an AI program it was using in its HR dept, because the data base that undergird its data was systemically biased against female employees (if women rather than men take time off because of pregnancy, and if laws in some states only allow mothers and not fathers to take time off after a birth then women as a population are more likely to take more time off than men when you compare the amount of time off taken by workers aged between, say 18 and 40... so, a program that notes "female" in a job application will reject the applicant (and it did). Similar problems of bias have been found with biases in AI programs used in Corrections. Garbage in.... garbage out. but since the algorithms are not transparent, and the data sets employed are rarely subject to academic scrutiny, most problems with AI are only discovered after they have made terrible decisions that humans have followed over the cliff (we know, too that sometimes the GPS in your car will tell drivers to drive across rivers and off mountain roads... those who use AI need to have some expectation for what counts as a reasonable result and not treat any and every result as a good one - as when I asked Chat GPT to give me a recipe for a gallon of mead and I was told to take a gallon of honey... ( as a maker of mead, I KNOW that two to three pounds would be reasonable , a gallon is 12 lbs...

 

[balatonwine]

Viberti has integrated all the information and recipes utilised over the previous ten years

Ten years.

Golly jeez wiz.

Wine making just out of diapers.

I have been growing grapes and making wine for over two decades, and I am just starting to figure it out.

In my humble opinion, (IMHO) Viberti is maybe an order of magnitude off in their data time line.

 

[winemaker81]

I have been growing grapes and making wine for over two decades, and I am just starting to figure it out.

In my humble opinion, (IMHO) Viberti is maybe an order of magnitude off in their data time line.

True. But they gotta start some place.

I have records going back to my beginnings, but up until the last 10 years or so, the records were usually sparse, 1-line descriptions of activities with date and SG. Especially for organizations that have been making wine for decades, the idea of detailed record keeping may not have occurred to them as useful, not until relatively recently.

 

[balatonwine]

AI is nothing but extremely good look-up tables.

Well... I think we can all assume it is probably a lot more complicated. There is probably a huge dose of somthing akin to Bayesian stats going on there too.

 

[balatonwine]

True. But they gotta start some place.

Agree.

But staring too soon has costs.

The past ten years of data may be a biased sample, for example, and trying to use it forward could be a huge mistake.... Potentially incomplete data is garbage in garbage out.

Thus, seems more a PR stunt. Or a serious endeavor by people who have not thought it through from all angles. AI can not replace basic knowledge requirements for decision making, or statistics, which requires a decent and unbiased sample size. And a sample size of "10" may not satisfy either requirement.