As a both AI expert and wine lover my humble opinion is that the main source of confusion is that AI believes labels too much in the training phase. On the other hand wine experts know winemakers' tricks which helps them to denoise judgments. Great video again @Konstantin Baum
Interesting comment, cheers! Yes, after tasting some wine, then reading the winemakers (fantasy) notes on the label, one is often wondering if we are tasting the same wine!
Hey Konstantin, it was a pleasure to see you live in the Masterclass (in Portugal's space) time was tight as usual, hence I didn't stop you for a picture - Fabrizio from Wines&Joy
can you make a video on the ageworthiness of wine? which wines are meant to be aged, which wines can be aged and which cant be aged. also the change in flavour for different grapes and wine styles. also a very interesting topic would be your view on alcohol itself. especially from you beeing in touch with this substance all day and every day. how you and the industry deals with the health issues and addiction of alcohol, thus tasting wine basicly almost every day of the week.
One aspect I see often missed when considering new applications of robotics and AI is that humans tend to prefer it when things are done by other humans. For example, a wine recommendation made by a human is more likely to be desired by another human than asking an AI version of WineSearcher to give them a generated verdict. AI, in my view has a brighter future assessing vineyard decisions based on physical factors, which is a more technical job were the client favors preciseness over human interaction.
Konstantin please would you make a video on ageing wine. How do you determine what wines should be aged and what makes some wines age worthy and others not. Thx
When you enter your tasting notes into ChatGBT, it tells me the wines are Sauvignon Blanc, Chardonnay, Syrah/Shiraz and Pinot Noir. Would be cool to see if you can beat AI in terms of concluding what the wine is! ChatGBT input: Which wine is this? The has a pale yellow colour with a slight touch of green. On the nose it is super aromatic, a lot of gooseberry, cassis, passionfruit aroma. ChatGBT output: Based on the description you provided, the wine you are describing is likely a Sauvignon Blanc. Sauvignon Blanc wines are known for their pale yellow color with greenish tones and are often described as having a very aromatic nose with notes of gooseberry, cassis, passionfruit, and sometimes even hints of herbaceousness or grassiness. Additionally, Sauvignon Blancs are known for their high acidity, making them refreshing and easy to drink.
I don't know if you get Canadian wines across the pond, but if you ever come across it, I wonder your opinion on Jackson Triggs or other wines from Ontario... a lot of them are made for mass market, but I know they specifically like to advertise their icewine
The development of the application of AI in the wine industry will be really interesting to follow. I don't believe we will see restaurants or wine bars in the near future use AI to help guests pick out a wine for their meal or just for enjoyment. There are so many terms different people use to describe the wine they'd like to have, and I believe AI is far away from being able to interpret all the imprecise descriptions people use. I can see AI being used in wine education though. Also, you have interviewed several winemakers on your channel! In your video on the Pfalz you had several interviews, this must be what the AI picked up on ;)
Love your channel! Would be great to hear your recommendations on high-scoring Bordeaux to drink now at say the £30-60 mark. So much of Bordeaux’s output seems to require years of laying down or sits way above that price (or is much cheaper but lacks complexity).
Great video, I like how your content oyen higlites how wine is/can interface with contemporary culture without being a gimmick. However, to take the AI's advice, it would be great to have you interview winemakers on the channel!
As a machine learning expert, I am not surprised; however too many people take the results from tools like GPT too seriously. It does show great potential though and can already be used if you are a subject matter expert.
Strange that Konstantin is confusing alc. content with body. To me body relates to the extract (the solid stuff). Heavy body meaning a lot of extract. I had cheap thin wines in Sicily some decades ago with probably 16% alc.
Great stuff! I am a wine student and thought about the same question recently. I think the machine / AI will beat humans in tasting / identifying wines one day, as long as it has enough data to deduce the wines. However, I don't really feel this is going to be a big 'threat' since wines are made for human consumption and the joy is always about sharing with family / friends, this alone is no machine can share!
I think wine is a too subjective topic for AI to really play a significant role as every human has a different taste. AI can help in analyzing human taste to categorize certain wines based on the wines profile..
One interesting experiment may be inputting your expressions on ChatGPT and ask what wine it would guess to see if the AI's vs your conclusions will differ (or align). I have GPT-4 (most recent paid version), and tried this on wine number 2, with a slight adaptation to your descriptions. Guess the wine (i.e., grape variety, vintage, region, producer) based on this description: Darker in color with a golden tinge, a little more profound in color than a sauvignon blanc; on the nose, ripe apple and lemon but not as expressive or aromatic as a sauvignon blanc; on the palate, rich and concentrated, there's a bit of alcohol at around 13.5 abv, the acidity is not too pronounced, pointing to a warmer region but not a hot region. On the palate there's concentration and richness. There's some banana flavors coming through. Quality is not too great. ChatGPT: "Based on the given description, I would guess the wine is a Chardonnay from a warmer region, perhaps a 2018 or 2019 vintage, due to the darker color with a golden tinge, ripe apple and lemon notes, and the higher alcohol content. The warmer region could potentially be California, USA or South Australia. The producer is hard to determine given the provided information, but it may be a lesser-known or budget-friendly producer, since the overall quality is not too great."
Here's wine #3: Guess the wine (i.e., grape variety, vintage, region, producer) based on this description: Purple core but a little lighter to the rim but quite concentrated and rich in terms of color; on the nose, it smells of blackberries, blueberries, and a little bit of pepperiness coming through, with a little rustic, herbaceous notes and flavors of black tea; on the palate, it's actually quite intense, the tannins are grippy, quite a lot of body, maybe 14% of alcohol, the acidity is rather low. Has a lot of bite and grip. All in all, it's a pretty good wine. GPT-4 "Based on the given description, I would guess the wine is a Syrah/Shiraz from a warmer climate, possibly a 2016 or 2017 vintage, due to the purple core, concentrated color, blackberries, blueberries, pepperiness, and rustic, herbaceous notes. The region could be the Barossa Valley in South Australia or the Rhône Valley in France, where Syrah/Shiraz grapes tend to have that profile. The producer is difficult to determine based on the description, but considering the overall quality of the wine, it could be a well-regarded winery from either of those regions, such as Penfolds or d'Arenberg from Australia, or E. Guigal or M. Chapoutier from France."
Konstantin wins! lol And here's ChatGPT's response when corrected: "Thank you for providing the correct information. Mullineux is a well-regarded producer in South Africa, known for their quality wines, especially Syrah. I apologize for not guessing the producer accurately earlier. I hope you enjoyed the 2018 Mullineux Syrah and found the tasting notes and description helpful."
Hi Konstantin! Thanks for another great video :) I literally just paused mid-way because something you said makes me want to ask you a question that's popped up for me several times already when watching yours and other videos on wine. It's about terminology: at 6:06 you say that Gruener Veltliner is 'not super aromatic, but it can be quite concentrated'. Could you elaborate on the difference between the two terms: what exactly would be the definition of 'aromatic' and what would be 'concentrated'?
I'm not an MW (currently studying for WSET3 with wset 1 and 2 already acquired) nor Konstantin but hope I can help. Aromatic wines are those that have a distinct and pronounced aroma or fragrance, often derived from the grapes themselves. Examples of aromatic wines include Riesling, Gewürztraminer, and Muscat. These wines typically have a lighter body and lower alcohol content, and are often served chilled. Aromatic wines are generally considered to be refreshing and easy to drink. Concentrated wines, on the other hand, are those that have a high concentration of flavor and intensity. This concentration can be achieved through various winemaking techniques such as longer hang time on the vine, lower yields, and selective harvesting. Concentrated wines are typically full-bodied and have a higher alcohol content. Examples of concentrated wines include Cabernet Sauvignon, Syrah, and Zinfandel. These wines are often aged in oak barrels to add complexity and depth. In summary, while both aromatic wines and concentrated wines can be flavorful and enjoyable, the former is known for its pronounced aroma while the latter for its intense flavor concentration. Chardonnay is the perfect example, it's a varietal that's not very aromatic, so winemakers usually apply various techniques to bring more concentration to it. Chardonnay with heavy oak and malolatic conversion are a prime example of a low aromatic wine but highly concentrated.
@@bgsouto2 Thanks, that answered my doubt really well! Thanks as well for the explanation @KonstantinBaumMasterofWine. I think in my mind I associated ''concentrated' more to the flavour (as in 'concentrated fruit') than body (fullness), so perhaps that made me confused.
If I get it right, AI is supposed to accumulate the previous knowledge - as such, it depends totally on wine experts, whose descriptions it should accumulate. However, if all the wine expert does can be reduced to analyzing wine technical qualities, so in the near future AI supposedly will be able to do it better than a human. I mean that AI just needs enough data. For example, beyond the chemical analysis of wine, there may be such an input as a terroir analysis, so that AI could possibly make a connection between a chemical and an agricultural data. In the end, my clients expect from me, as a wine seller, to explain them what to expect from a specific wine, why to expect that and how to predict the next wine should be nice even if they come to the store and suddenly I'm on my vacation. It seems like maybe your job will not be replaced nearly by AI, but mine will be in the most near future.
Would be fun to have an electronic pen handy that you could keep with you to quickly test what the wine was made of. Incase I had left a glass out for the next day already forgetting what it was. hahaha
It would be really interesting to hear what you have to say about companies like Cameron Hughes and if you actually think it’s a good value and quality wine
I think I've seen your video from somewhere! Anyway, I think there could be machine learning AI that could actually perform grading of wines. But still not sure how this will be developed. chatGPT would not be able to perform a proper tasting for this matter as first and foremost, it does not have any "tasting" capabilities. There are still loads of limitations to them. Question for you though, you think it's ok for wine makers to use image generator AIs like Bluewillow to create their wine labels?
Interesting video, result somehow to be expected, I would have wondered if a MoW or even myself woul be beat by an AI learning algorythm using just basic data and what the web offers. BUT this might change if someone is able to use plattforms like Cellartracker to train the AI, because I find that for wines where there are 10+ tasting notes Cellartracker has a tendency to be spot on with my own perception and to be at least as reliable as a MoW or other Professionals.
Interesting session. The AI notes at this point seem pretty generic, like you find in the text books on wine. Nothing like speaking to a real pro drinker with knowledge, than an AI, or someone just picking up and reading printed notes. Cheers, Marc D
I think there's some confusion in the discussion of the use of AI here and perhaps a lack of appreciation of the potential of Sillén's AI-based approach to affect wine making and wine culture. First, although chatGPT is an AI system, it is very different: the main thing here is that chatGPT is trained on all available internet data which is vast but ultimately still limited to text-based discussions and simple linear rankings of wine (e.g. from different wine magazines). ChatGPT and contemporaries would essentially just be summarizing multiple discussions (that might still have some practical utility already e.g. when considering purchasing an expensive bottle). Sillén's approach however is based solely (from what I understand) on the metabolic profile of the wine (I guess determined by mass spectrometry which just breaks down the wine into its constituent chemicals). Then from these molecular profiles, they attempt to predict different features of wine e.g. acidity, cellar length, etc. If you think in the extremes, this almost certainly has the capacity to predict how humans will broadly perceive the wine: extreme acid content alone is easy to detect and would almost always result in a low score, for example. Whether it can detect the subtle differences in acidity that help differentiate a 93 bottle from a 94 bottle is probably still in the future - this differentiation obviously requires a multivariate approach that considers many features of the wine simultaneously. Can any human consider 100 features of a bottle of wine simultaneously? It is clear however that the current version of the software (which is something probably like version 1.0) does seem to nail the obvious things (e.g. for the sauvignon and the gruenerveltliner). So it's probably operating at the "beginner sommelier" level! But version 5.0? I just want to also point out that the potential for such a system is huge: the ability to test wine at the molecular level in a cheap, low-resource manner would allow for tracking its evolution in an objective detailed manner, and the metabolite profiles would probably suggest interventions to "push" the evolution into desirable directions. I doubt humans will keep up and that's a good thing: AI will ultimately push wine into directions that we never thought possible. (Concrete example: watch a game of chess between two deep AI systems.)
I think there's a lot of future for IA in wine..PRODUCTION. We've seen great results in pest control and other viticulture enhancements through IA. But in terms of tasting itself not so sure really
I thought the wine was send to the lab and blind taste to find out the wine. It’s just the technical note from GPT. For me the AI is working based on the data collectively from searching hence why the aging has indicated 10 years 😅. Experts can share more detail knowledge and aging potential because they have tried the wine before but AI (general data) hasn’t. I’ll guess the AI can replace commercial sector for writing the tasting note which is a lot accurate than marketing manager 😂. Peace out
AI has a huge advantage over humans - it can be completely objective. WS, WA, JS, etc give a wine 98 points, I always wonder if either they directly benefit from sales financially, or they indirectly are biased as they like the winery, place, etc. Or they rate a wine badly because it is from an odd varietal or country. An AI wine critic is more trustworthy, especially if the model and code were provided as open source.
I think the challenge with AI is in the training. If you rely on what people write on the internet about wine to train your AI, you’re going to get a product built on a lot of incorrect information!
I'm not convinced this has much to do with AI at all. To me it actually smacks more of a party trick. It brims of commonplaces just like a horoscope does. Is there a future for AI in the wine tasting business? Yes, but a distant one. Unless a machine can "drink" (i.e. analyze) and gather and deduct further information as a human would do, I'm not impressed.
I haven't watched the video yet but I can tell you that wine is very subjective. There seems to be no "this is good, that is bad" as far as wines are considered. Now, can a computer get to know your palate? Yes, of course but then again so can a wine reviewer. As you know, I review wine and what I tell the people who follow my wine social media is that if I like for instance a Cabernet Sauvignon and you also happen to like it, we may have the same palate as far as that varietal is concerned. This "like" may not apply however to any other varietal. As foe ChatGPT, it's results are skewed torward....let's just say, a particular way of thinking. By the way, I agree with your thought process.
Yeah I don't think there's a reason to worry about AI in a while. Furthermore, AI doesn't produce objective knowledge, it's always influenced by its creator
I don't know if there's a future for AI, but I still prefer human interactions. I also think that what AI does, unfortunately, is to deskill people's abilities. Not a big fun of it 😁
When you go to the store in the future and there's 2 equal bottles of wines next to each other, same grape, same rating, same profile etc, but price difference is 50%, would you go for the more expensive bottle? This is why AI is the future
I asked ChatGPT if certain people were wines what wine would they be. Some of the answers were funny, some weird, some bonkers. Example I leave here is Donald Trump: Donald Trump would be a bold, full-bodied Cabernet Sauvignon. Cabernet Sauvignon is known for being assertive and powerful, with a bold and sometimes divisive character, much like Donald Trump's public persona. It's a wine that can be quite polarizing, with some people loving it and others finding it too intense. Additionally, Cabernet Sauvignon is often associated with luxury and wealth, which are also characteristics that have been associated with Donald Trump.
And of course: If Konstantin Baum were a wine, I might suggest a Riesling. Riesling is a grape variety that is often associated with Germany and is capable of producing wines that are complex, nuanced, and full of character, much like Konstantin Baum's expertise and knowledge of wine. Riesling can also be versatile and adaptable, which could reflect Konstantin Baum's ability to communicate with different audiences through his TH-cam channel. Additionally, Riesling can be a wine that rewards exploration and study, much like the journey of becoming a master of wine.
Machines don’t have a soul. It’s just wrong to link machines with anything beautiful eg art, wine, food. They can’t know what beauty is. They don’t have emotions etc.
I still can't believe master of wine just said: sulphites are too low for a "not natural wine". How you dare, as a passionate professional, call the wine "not natural"? The work and passion for past 8000 years, you just spit on.. not cool
It would be great to get you to travel to up & coming wine regions around the globe. Give your ideas and thoughts. In the South Eastern USA, Georgia has a relatively new AVA that's making awesome stuff. Super interesting video! AI is amazing, but it isn't quite advanced enough yet to break outside of "mass reading & averaging" labels yet. @konstantinBaum
![CAMPปลิ้น | EP.83[1/2] เปิดแคมป์หลอนต้อนรับ 2 ตัวพ่อแห่งวงการผี!](http://i.ytimg.com/vi/2uPuHYHgwiI/mqdefault.jpg)
As a both AI expert and wine lover my humble opinion is that the main source of confusion is that AI believes labels too much in the training phase. On the other hand wine experts know winemakers' tricks which helps them to denoise judgments. Great video again @Konstantin Baum
Interesting comment, cheers! Yes, after tasting some wine, then reading the winemakers (fantasy) notes on the label, one is often wondering if we are tasting the same wine!
The human palate. Does it taste good to the individual and meet their exoectations?
If the wine industry became more transparent in the future, the AIs will have an easier time
Hey Konstantin, it was a pleasure to see you live in the Masterclass (in Portugal's space) time was tight as usual, hence I didn't stop you for a picture - Fabrizio from Wines&Joy
can you make a video on the ageworthiness of wine? which wines are meant to be aged, which wines can be aged and which cant be aged. also the change in flavour for different grapes and wine styles.
also a very interesting topic would be your view on alcohol itself. especially from you beeing in touch with this substance all day and every day. how you and the industry deals with the health issues and addiction of alcohol, thus tasting wine basicly almost every day of the week.
The nod to Trojan condoms with the nipple hat is a nice touch. Love this show! Keep up the good work.
Interesting video. I had no idea that anyone was trying to use AI to analyze wines. Thanks for showing this.
This video was great fun. Thanks, Konstantin!
Love your channel....keep these interesting videos coming.
One aspect I see often missed when considering new applications of robotics and AI is that humans tend to prefer it when things are done by other humans. For example, a wine recommendation made by a human is more likely to be desired by another human than asking an AI version of WineSearcher to give them a generated verdict.
AI, in my view has a brighter future assessing vineyard decisions based on physical factors, which is a more technical job were the client favors preciseness over human interaction.
Woohoo! A South African wine in the mix. Thanks Konstantin. 🙏🙌
Well wine is a social thing, I think people will always prefer the opinion of another human being. I think your job is safe. Nice one Konstantin! 🍷🌟👍
Great content, as always. The results are.... as expected :)
magnificent idea, just another level 👏👏👏👏
Konstantin please would you make a video on ageing wine. How do you determine what wines should be aged and what makes some wines age worthy and others not. Thx
When you enter your tasting notes into ChatGBT, it tells me the wines are Sauvignon Blanc, Chardonnay, Syrah/Shiraz and Pinot Noir. Would be cool to see if you can beat AI in terms of concluding what the wine is!
ChatGBT input: Which wine is this? The has a pale yellow colour with a slight touch of green. On the nose it is super aromatic, a lot of gooseberry, cassis, passionfruit aroma.
ChatGBT output: Based on the description you provided, the wine you are describing is likely a Sauvignon Blanc. Sauvignon Blanc wines are known for their pale yellow color with greenish tones and are often described as having a very aromatic nose with notes of gooseberry, cassis, passionfruit, and sometimes even hints of herbaceousness or grassiness. Additionally, Sauvignon Blancs are known for their high acidity, making them refreshing and easy to drink.
Interesting video. Again.... Tstrike!
I don't know if you get Canadian wines across the pond, but if you ever come across it, I wonder your opinion on Jackson Triggs or other wines from Ontario... a lot of them are made for mass market, but I know they specifically like to advertise their icewine
The development of the application of AI in the wine industry will be really interesting to follow. I don't believe we will see restaurants or wine bars in the near future use AI to help guests pick out a wine for their meal or just for enjoyment. There are so many terms different people use to describe the wine they'd like to have, and I believe AI is far away from being able to interpret all the imprecise descriptions people use.
I can see AI being used in wine education though.
Also, you have interviewed several winemakers on your channel! In your video on the Pfalz you had several interviews, this must be what the AI picked up on ;)
Love your channel! Would be great to hear your recommendations on high-scoring Bordeaux to drink now at say the £30-60 mark. So much of Bordeaux’s output seems to require years of laying down or sits way above that price (or is much cheaper but lacks complexity).
Chat GPT does not go online and search tasting notes. It is a model that has been trained with lots of texts
Well, it has but it uses the information available online to create the responses. In this example it misquotes from Wine Spectator and RP
Great video, I like how your content oyen higlites how wine is/can interface with contemporary culture without being a gimmick. However, to take the AI's advice, it would be great to have you interview winemakers on the channel!
As a machine learning expert, I am not surprised; however too many people take the results from tools like GPT too seriously. It does show great potential though and can already be used if you are a subject matter expert.
Strange that Konstantin is confusing alc. content with body. To me body relates to the extract (the solid stuff). Heavy body meaning a lot of extract. I had cheap thin wines in Sicily some decades ago with probably 16% alc.
I think you can’t subtract alcohol from the body equation as it makes up the - by far - largest share of a (dry) wines body.
Great stuff! I am a wine student and thought about the same question recently. I think the machine / AI will beat humans in tasting / identifying wines one day, as long as it has enough data to deduce the wines. However, I don't really feel this is going to be a big 'threat' since wines are made for human consumption and the joy is always about sharing with family / friends, this alone is no machine can share!
Konstantin, which information was given to the artificial intelligence, and what was AI asked to do?
I think wine is a too subjective topic for AI to really play a significant role as every human has a different taste. AI can help in analyzing human taste to categorize certain wines based on the wines profile..
One interesting experiment may be inputting your expressions on ChatGPT and ask what wine it would guess to see if the AI's vs your conclusions will differ (or align).
I have GPT-4 (most recent paid version), and tried this on wine number 2, with a slight adaptation to your descriptions.
Guess the wine (i.e., grape variety, vintage, region, producer) based on this description: Darker in color with a golden tinge, a little more profound in color than a sauvignon blanc; on the nose, ripe apple and lemon but not as expressive or aromatic as a sauvignon blanc; on the palate, rich and concentrated, there's a bit of alcohol at around 13.5 abv, the acidity is not too pronounced, pointing to a warmer region but not a hot region. On the palate there's concentration and richness. There's some banana flavors coming through. Quality is not too great.
ChatGPT:
"Based on the given description, I would guess the wine is a Chardonnay from a warmer region, perhaps a 2018 or 2019 vintage, due to the darker color with a golden tinge, ripe apple and lemon notes, and the higher alcohol content. The warmer region could potentially be California, USA or South Australia. The producer is hard to determine given the provided information, but it may be a lesser-known or budget-friendly producer, since the overall quality is not too great."
Here's wine #3:
Guess the wine (i.e., grape variety, vintage, region, producer) based on this description: Purple core but a little lighter to the rim but quite concentrated and rich in terms of color; on the nose, it smells of blackberries, blueberries, and a little bit of pepperiness coming through, with a little rustic, herbaceous notes and flavors of black tea; on the palate, it's actually quite intense, the tannins are grippy, quite a lot of body, maybe 14% of alcohol, the acidity is rather low. Has a lot of bite and grip. All in all, it's a pretty good wine.
GPT-4
"Based on the given description, I would guess the wine is a Syrah/Shiraz from a warmer climate, possibly a 2016 or 2017 vintage, due to the purple core, concentrated color, blackberries, blueberries, pepperiness, and rustic, herbaceous notes. The region could be the Barossa Valley in South Australia or the Rhône Valley in France, where Syrah/Shiraz grapes tend to have that profile. The producer is difficult to determine based on the description, but considering the overall quality of the wine, it could be a well-regarded winery from either of those regions, such as Penfolds or d'Arenberg from Australia, or E. Guigal or M. Chapoutier from France."
Konstantin wins! lol
And here's ChatGPT's response when corrected:
"Thank you for providing the correct information. Mullineux is a well-regarded producer in South Africa, known for their quality wines, especially Syrah. I apologize for not guessing the producer accurately earlier. I hope you enjoyed the 2018 Mullineux Syrah and found the tasting notes and description helpful."
Interesting stuff!
Hi Konstantin! Thanks for another great video :) I literally just paused mid-way because something you said makes me want to ask you a question that's popped up for me several times already when watching yours and other videos on wine. It's about terminology: at 6:06 you say that Gruener Veltliner is 'not super aromatic, but it can be quite concentrated'. Could you elaborate on the difference between the two terms: what exactly would be the definition of 'aromatic' and what would be 'concentrated'?
I'm not an MW (currently studying for WSET3 with wset 1 and 2 already acquired) nor Konstantin but hope I can help.
Aromatic wines are those that have a distinct and pronounced aroma or fragrance, often derived from the grapes themselves. Examples of aromatic wines include Riesling, Gewürztraminer, and Muscat. These wines typically have a lighter body and lower alcohol content, and are often served chilled. Aromatic wines are generally considered to be refreshing and easy to drink.
Concentrated wines, on the other hand, are those that have a high concentration of flavor and intensity. This concentration can be achieved through various winemaking techniques such as longer hang time on the vine, lower yields, and selective harvesting. Concentrated wines are typically full-bodied and have a higher alcohol content. Examples of concentrated wines include Cabernet Sauvignon, Syrah, and Zinfandel. These wines are often aged in oak barrels to add complexity and depth.
In summary, while both aromatic wines and concentrated wines can be flavorful and enjoyable, the former is known for its pronounced aroma while the latter for its intense flavor concentration.
Chardonnay is the perfect example, it's a varietal that's not very aromatic, so winemakers usually apply various techniques to bring more concentration to it. Chardonnay with heavy oak and malolatic conversion are a prime example of a low aromatic wine but highly concentrated.
For me aromatic refers to the intensity of the wine’s aroma and concentration in this context referred to the wine’s alcohol and body
@@bgsouto2 Thanks, that answered my doubt really well!
Thanks as well for the explanation @KonstantinBaumMasterofWine. I think in my mind I associated ''concentrated' more to the flavour (as in 'concentrated fruit') than body (fullness), so perhaps that made me confused.
If I get it right, AI is supposed to accumulate the previous knowledge - as such, it depends totally on wine experts, whose descriptions it should accumulate. However, if all the wine expert does can be reduced to analyzing wine technical qualities, so in the near future AI supposedly will be able to do it better than a human. I mean that AI just needs enough data. For example, beyond the chemical analysis of wine, there may be such an input as a terroir analysis, so that AI could possibly make a connection between a chemical and an agricultural data. In the end, my clients expect from me, as a wine seller, to explain them what to expect from a specific wine, why to expect that and how to predict the next wine should be nice even if they come to the store and suddenly I'm on my vacation. It seems like maybe your job will not be replaced nearly by AI, but mine will be in the most near future.
Would be fun to have an electronic pen handy that you could keep with you to quickly test what the wine was made of. Incase I had left a glass out for the next day already forgetting what it was. hahaha
I think AI could be helpful keeping track of inventory. I can write better if very general tasting notes by looking at the bottle.
Can make you sure, that GV from Pfaffl Hommage will survive 10 Years minimum. GV is famous for the ability to age well
It would be really interesting to hear what you have to say about companies like Cameron Hughes and if you actually think it’s a good value and quality wine
Also what you think about the online wine clubs/subscriptions where you get like 3 bottles of wine from a region every month.
AI wouldn’t be able to make a channel so funny and informative as this one. The computer lacks the drinking experience and the wits of the MW…
😊
Probably the shelf-life of AI in wine tasting will also be 10 years!😂
Isn't barebera more opaque and less translucent? Or am I mixing up another grape variety from Piedmonte?
It can be darker but it is not known for producing very dark colored wines. The color of this wine could have taken me into the right direction.
10 years shelf life at a 7-11. Wonder what HAL's favorite wine was.
Chat AI doesn’t check the net as it thinks Queen Elizabeth II is still alive.
I think I've seen your video from somewhere! Anyway, I think there could be machine learning AI that could actually perform grading of wines. But still not sure how this will be developed. chatGPT would not be able to perform a proper tasting for this matter as first and foremost, it does not have any "tasting" capabilities. There are still loads of limitations to them. Question for you though, you think it's ok for wine makers to use image generator AIs like Bluewillow to create their wine labels?
The current standard for an advanced AI that can actually does ‘shit’ is ‘Kit’ from ‘Night Rider’ - David Hasselhoff is still the ‘funniest’ German🤓
I want tasting notes that inspire emotion, like an empty fruit basket, or object failure
Interesting video, result somehow to be expected, I would have wondered if a MoW or even myself woul be beat by an AI learning algorythm using just basic data and what the web offers. BUT this might change if someone is able to use plattforms like Cellartracker to train the AI, because I find that for wines where there are 10+ tasting notes Cellartracker has a tendency to be spot on with my own perception and to be at least as reliable as a MoW or other Professionals.
Hehe, felt like you saved us from AI invasion threat 💪fun and bubbly, as usual. Congrats on accurate South Africa guess 🥂
9:25 grüner bottle is flat
Interesting session.
The AI notes at this point seem pretty generic, like you find in the text books on wine.
Nothing like speaking to a real pro drinker with knowledge, than an AI, or someone just picking up and reading printed notes.
Cheers, Marc D
I think there's some confusion in the discussion of the use of AI here and perhaps a lack of appreciation of the potential of Sillén's AI-based approach to affect wine making and wine culture. First, although chatGPT is an AI system, it is very different: the main thing here is that chatGPT is trained on all available internet data which is vast but ultimately still limited to text-based discussions and simple linear rankings of wine (e.g. from different wine magazines). ChatGPT and contemporaries would essentially just be summarizing multiple discussions (that might still have some practical utility already e.g. when considering purchasing an expensive bottle). Sillén's approach however is based solely (from what I understand) on the metabolic profile of the wine (I guess determined by mass spectrometry which just breaks down the wine into its constituent chemicals). Then from these molecular profiles, they attempt to predict different features of wine e.g. acidity, cellar length, etc. If you think in the extremes, this almost certainly has the capacity to predict how humans will broadly perceive the wine: extreme acid content alone is easy to detect and would almost always result in a low score, for example. Whether it can detect the subtle differences in acidity that help differentiate a 93 bottle from a 94 bottle is probably still in the future - this differentiation obviously requires a multivariate approach that considers many features of the wine simultaneously. Can any human consider 100 features of a bottle of wine simultaneously? It is clear however that the current version of the software (which is something probably like version 1.0) does seem to nail the obvious things (e.g. for the sauvignon and the gruenerveltliner). So it's probably operating at the "beginner sommelier" level! But version 5.0? I just want to also point out that the potential for such a system is huge: the ability to test wine at the molecular level in a cheap, low-resource manner would allow for tracking its evolution in an objective detailed manner, and the metabolite profiles would probably suggest interventions to "push" the evolution into desirable directions. I doubt humans will keep up and that's a good thing: AI will ultimately push wine into directions that we never thought possible. (Concrete example: watch a game of chess between two deep AI systems.)
I think there's a lot of future for IA in wine..PRODUCTION. We've seen great results in pest control and other viticulture enhancements through IA. But in terms of tasting itself not so sure really
Would love to get your rating for the Mullineux Shiraz please
I thought the wine was send to the lab and blind taste to find out the wine. It’s just the technical note from GPT. For me the AI is working based on the data collectively from searching hence why the aging has indicated 10 years 😅. Experts can share more detail knowledge and aging potential because they have tried the wine before but AI (general data) hasn’t. I’ll guess the AI can replace commercial sector for writing the tasting note which is a lot accurate than marketing manager 😂. Peace out
AI has a huge advantage over humans - it can be completely objective. WS, WA, JS, etc give a wine 98 points, I always wonder if either they directly benefit from sales financially, or they indirectly are biased as they like the winery, place, etc. Or they rate a wine badly because it is from an odd varietal or country. An AI wine critic is more trustworthy, especially if the model and code were provided as open source.
I think the challenge with AI is in the training. If you rely on what people write on the internet about wine to train your AI, you’re going to get a product built on a lot of incorrect information!
I'm not convinced this has much to do with AI at all. To me it actually smacks more of a party trick. It brims of commonplaces just like a horoscope does.
Is there a future for AI in the wine tasting business? Yes, but a distant one. Unless a machine can "drink" (i.e. analyze) and gather and deduct further information as a human would do, I'm not impressed.
I haven't watched the video yet but I can tell you that wine is very subjective. There seems to be no "this is good, that is bad" as far as wines are considered. Now, can a computer get to know your palate? Yes, of course but then again so can a wine reviewer. As you know, I review wine and what I tell the people who follow my wine social media is that if I like for instance a Cabernet Sauvignon and you also happen to like it, we may have the same palate as far as that varietal is concerned. This "like" may not apply however to any other varietal. As foe ChatGPT, it's results are skewed torward....let's just say, a particular way of thinking. By the way, I agree with your thought process.
Yeah I don't think there's a reason to worry about AI in a while. Furthermore, AI doesn't produce objective knowledge, it's always influenced by its creator
I'll always trust you more than a machine, rest assured :D
🙏
I don't know if there's a future for AI, but I still prefer human interactions. I also think that what AI does, unfortunately, is to deskill people's abilities. Not a big fun of it 😁
Its for people with no imagination.
When you go to the store in the future and there's 2 equal bottles of wines next to each other, same grape, same rating, same profile etc, but price difference is 50%, would you go for the more expensive bottle? This is why AI is the future
I asked ChatGPT if certain people were wines what wine would they be. Some of the answers were funny, some weird, some bonkers.
Example I leave here is Donald Trump: Donald Trump would be a bold, full-bodied Cabernet Sauvignon. Cabernet Sauvignon is known for being assertive and powerful, with a bold and sometimes divisive character, much like Donald Trump's public persona. It's a wine that can be quite polarizing, with some people loving it and others finding it too intense. Additionally, Cabernet Sauvignon is often associated with luxury and wealth, which are also characteristics that have been associated with Donald Trump.
And of course: If Konstantin Baum were a wine, I might suggest a Riesling. Riesling is a grape variety that is often associated with Germany and is capable of producing wines that are complex, nuanced, and full of character, much like Konstantin Baum's expertise and knowledge of wine. Riesling can also be versatile and adaptable, which could reflect Konstantin Baum's ability to communicate with different audiences through his TH-cam channel. Additionally, Riesling can be a wine that rewards exploration and study, much like the journey of becoming a master of wine.
Beautiful!
Machines don’t have a soul. It’s just wrong to link machines with anything beautiful eg art, wine, food. They can’t know what beauty is. They don’t have emotions etc.
I still can't believe master of wine just said: sulphites are too low for a "not natural wine". How you dare, as a passionate professional, call the wine "not natural"? The work and passion for past 8000 years, you just spit on.. not cool
This is not a real AI😂 I'm interested to see a real olfactory sensor used in the analysis.
Human senses trump technology, every time.
So wrong 🥲
How utterly depressing. Please keep wine 'analogue'.
Who is ensuring that the wine tasting AI isn’t moonlighting as a driverless car AI?
It would be great to get you to travel to up & coming wine regions around the globe. Give your ideas and thoughts. In the South Eastern USA, Georgia has a relatively new AVA that's making awesome stuff.
Super interesting video! AI is amazing, but it isn't quite advanced enough yet to break outside of "mass reading & averaging" labels yet. @konstantinBaum
No thanks. No need to out-source thinking.