There has been a lot of online chatter recently about the Artificial Intelligence platform ChatGPT, partly sparked by an AI-generated post on Bordeaux expert Jane Anson’s online platform.
I was delighted that Anson had directed the spotlight to a subject I addressed late last year in Verona at the annual Wine2Wine conference, organised by Stevie Kim of Vinitaly and the Vinitaly International Academy.
Kim had thrown me the curve ball of scheduling my presentation on ‘Wine Communication Today’ at the end of the event - after a long list of other speakers had covered the subject pretty comprehensively. In particular, Erica Duecy, who was formerly the Chief Content Officer of first VinePair and then Pix, had outlined the situation in the US brilliantly during the opening hour of the event. Any gaps she’d left, had been filled by a number of other first class experts such as my friend and sometime colleague, Polly Hammond and former Meininger’s editor Felicity Carter. There was almost nothing left to say about ‘today’. Worse still, the presenter directly before me was Konstantin Baum, the dynamic young German Master of Wine who has become a wine YouTube sensation with millions of followers.
Playing games with the audience
So I decided to look at the future and, in the hope of holding the audience’s attention, to go off piste and to play some games with them.
First came a little quiz. I presented the following piece of text on the screen.
Why I Love Barbera
The first time I tasted Barbera was at the tasting room of Cantara Cellars in Camarillo... it quickly became one of my favorite wines. I was so impressed with its sexy personality and its ability to age beautifully... It's like a perennial favorite - it just gets better with age…”
I showed it alongside three unnamed faces labelled A, B, and C.
And said we'd come back to the text and faces later...
Instinctive decision making
Then, as briefly as possible, I introduced those who weren’t already familiar with it - the majority - to the economics Nobel laureate, Daniel Kahneman’s notion of Thinking Fast and Thinking Slow. Stated simply, Kahneman - like a growing number of others - believes that 95% of our decision making is more or less instinctive.
Then I invited the attendees to make their own instinctive decision and to identify the face they thought had written the words about Barbera. Around a third thought it was either A or B, with the remainder opting for C.
Why? I asked. Because, came one answer, “It looked like the kind of thing an Englishwoman might have said.” I asked anyone who agreed to raise their hands. Many did.
Clearly, and with absolutely no logical reason for doing so, they had presumed that because I carry a UK passport, the same might apply to the white female I’d labelled C.
Then I revealed that neither C, B nor A was British - or American or indeed of any other nationality. None was a real human being; all three had been created by a computer and sourced for $9 apiece from among the 2.5 million images on offer from an online service called Generated Photos.
And, of course, like the far longer piece published by Anson, the words about Barbera had been written by a presumably non wine-drinking AI bot in response to the simple prompt “why I love Barbera”.
Creepy texts
Next, I introduced the audience to Dave, a 52 year-old American graphic designer and Andi, his 25 year-old female assistant - via a piece of texted dialogue. Andi, who was evidently working from home, had apparently taken a 10 minute break to have a shower. When she returned, she’d written “perhaps I should have asked if you wanted to join me”. To which Dave had replied “Love the idea”.
Even in Italy, where attitudes to sex can be less strait-laced than in Anglo Saxon countries, this raised eyebrows.
But, I explained, Dave could no more have shared the hot water with his young employee than he could with A, B or C, because, like them, Andi has no physical presence. She’s a ‘virtual assistant’ - powered by AI that’s sufficiently lifelike for Dave to tell the New York Post that “I’ve actually wondered if I wasn’t talking to a real person.”
Miquela, the virtual influencer
Miquela, an online influencer who appeared in the 2021 video I then showed, makes no pretence of being a real person, but like Andi, she has an age: “I’m a 19 year-old robot who lives in LA and makes music” she says, before answering questions from real young people such as “What’s your worst memory?” and “What’s something about humans that you just don’t get?”
Miquela then talked about “travelling to Brazil” and “to Paris for Fashion Week” and having a very public break up with her flesh-and-blood boyfriend Nick (aka Angel Boy), on whose lap she is seen sitting in a piece about the split in The Cut.
If none of this makes any sense to you, please pause before dismissing it as nonsense. Miquela has over 3m followers and lucrative ‘partnership deals’ with Prada, Calvin Klein and Dior.
Maria Margarita
If we take her at her digital word, in 2021, Miquela was two years too young to drink legally in the US, so perhaps unsurprisingly, she - or her creators - hadn’t signed any similar deals with any wine or spirits companies. But Maria Margarita, the ‘virtual global citizen with Mexican roots’ Pernod Ricard has created for its Olmeca tequila brand is 25, so perfectly pitched at the crucial cohort of 18-30 year old consumers. Even so, her iammariamargarita Instagram page comes with the instruction “Please drink responsibly and don’t share with anyone underage.”
Curiously, given Pernod Ricard’s marketing muscle, and despite 137 costly video clips, Maria Margarita only seems to have got herself a thousand or so followers; fewer than countless real life 25 year olds.
AI-sceptics will be delighted by this apparent lack of viral success, just as they have relished the great Australian songwriter Nick Cave’s dismissal of an AI song ‘written in the style of Nick Cave’. AI writing is derivative, some have said, and ‘curiously flat in tone’ one critic noted, or spoiled by linguistic errors such as ‘human sommelier’ as others have said. A computer, the sceptics claim, will never produce a great work of art or the kind of wine writing we appreciate from Jancis Robinson, Tim Atkin or Andrew Jefford.
Early days of the internet
Seeing these comments reminds me of my experience in April 1997 when I gave a talk at the fourth Masters of Wine symposium in Perth Australia. My topic was how the internet might be relevant to the wine industry. The World Wide Web or WWW as it was known, was still in its infancy. Two years earlier, CERN, the European Organization for Nuclear Research had held a two-day seminar for 250 European journalists using 60 computers to show them what it was and how it worked. Thirty high school-aged students were on hand to show them how to ‘surf the web’.
When I went online in the Perth conference room, the connection was slow and shaky. Waiting for pages to load was painful and my audience was generally as unimpressed as modern motorists would be if they were transported back in time to the launch of the Model T Ford. Only a tiny number could imagine how this new technology would ever affect their business.
In historic terms, it did not take long to get from those primitive beginnings to online retail and social media, but things happen a lot faster today. Most of us are already reading text written by AI bots every time we see a brief sports or weather report, or one on the release of company quarterly results. Countless millions of us unthinkingly watch movies on Netflix or listen to Spotify playlists that would not exist without AI, but compared to what it will deliver, the technology is as embryonic as the internet was in 1997.
What does the future hold?
So here are a few predictions:
- No, AI won’t replace great journalism or literature, but it will be used to generate press releases and a growing volume of what we read in the media.
- Linguistic infelicities will disappear - AI is all about the technology learning from its mistakes - as will the ‘flat tone’. And it will also write tasting notes. I recently talked to a young woman who writes descriptions of en primeur offerings for a major German merchant. She is a skilled taster, but her client doesn’t send her the samples; all she does is amalgamate comments by leading critics. As we both agreed, a computer could do the job as well, or better, in a fraction of time. Of course, the bot will not replace the top tasters, but there aren’t many of them…
- And, bots will do the work of sommeliers. Again, they won't supplant the great somms I'm going to watch compete for the Meilleur Sommelier du Monde crown in Paris next month. But they'll do the job brilliantly in the millions of restaurants that currently have staff with little or no training in wine service.
Looking forward
- When it comes to analysis, given the breadth of data to which AI it has access, I can see it introducing far more lateral thinking than we are used to as humans. I am regularly surprised by how’s little wine industry members in one country know about what is happening elsewhere, both in their own business, in other beverages and in totally different sectors. A bot has no such constraints. A French wine business may learn from the experience of a whisky producer in Japan - or a luxury goods producer in Italy.
- Bots will help us with our daily business activities. Zoom already offers premium paying customers a service in which a bot will take minutes from online meetings and identify the most important moments in the discussion. We are not far from them also being able to offer us a daily selection of what we might find interesting too read and watch - and to give us a brief description that will save us time in making our selection.
At first, just as Spotify once said to me “You’ve been listening to Bruce Springsteen, maybe you’ll like George Michael”, the bots will get it wrong. But Spotify now generally gets out right for me and, just as online language translators have improved, so will the digital advisors.
- Virtual presenters - Miquela’s and Maria Margarita’s successors - will not only be reading the news and quite possibly conducting interviews, they will also be personalised to suit each of our tastes, either as virtual employees - like Andi - or ‘friends’. Over 10m people have already signed up for a Replika digital chatbot and as the offering becomes more sophisticated, that number will explode.
- AI will also create social media posts, including images and video clips, more quickly, efficiently and cheaply than human beings. And they'll monitor the success of their posts and tailor them accordingly.
- Finally, and perhaps crucially, right now we are being invited to play with AI technology like ChatGPT and Dall E images at little or no cost. The creators of this technology are happily learning from us as we do so, but sooner rather than later, someone is going to have to cover all of the costs that have gone into it. And, the people with the deepest pockets will benefit - and literally profit - from the best AI.
As Ronald Reagan liked to say, we ain’t seen nothing yet. The huge investment that is being pumped into this sector will reap astonishing - and almost certainly, in the case of deep fakes, terrifying rewards. But sticking one's head in the sand is not going to make it go away any more than one could wish away the internet in 1997.
I hope that a bot won’t be writing this column any time soon, but I have no doubt that it could happen and am making sure that I always have a plan B. Hopefully AI will help me to make it a good one.
This piece was briefly edited on January 21st to include a reference to social media that was inadvertently omitted