The Problems of AI (in the Whisky World)

First, what *is* AI really?

What we now call 'AI' is not, in fact, 'Artificial Intelligence', despite the name. Chat GPT, Dall E 3, and Open AI are all ' Large Language Models ', and the difference isn't just semantic. What we now call AI is not 'intelligent' by any definition. It scrapes the internet for information, then recombines and replicates what it finds. 

These 'AI' programs therefore cannot create an idea or a thought. The real innovation - their ability to take inputs and give outputs in plain speech rather than programming languages ​​- gives a false impression that these programs actually comprehend everything they see. It's really important to always remember this when dealing with AI - it doesn't actually know anything.

Worst of all, because AI works by trawling the internet for input, it catches other AI-produced information in the process. That means that AI is feeding on itself and creating recursive loops of nonsense . AI programs now make up medical statements that no-one ever said. They can't tell you how many 'r's there are in the word strawberry. They pour increasingly Lovecraftian images onto Facebook (which I will not be linking to, for both your sanity and mine).

Where does Whiskey come into it?

AI comes from the same ecosystem of Web 3.0 slop and cons as cryptocurrency and NFTs. The whiskey industry also tried to jump on the NFT bandwagon during its peak (although one writer also called NFTs a 'new frontier in whisky collecting’ in June 2023, long after the concept was widely discredited).

Ardbeg released Fon Fhòid “as” an NFT… where you could exchange the given token for a bottle. An NFT is a glorious number, a unique data string, that's it. Yet this whisky, aged in casks buried in a peat bog, were sold for 1 ETH each (about 2800 Euros at the time). 

See above - Glenlivet's own dabbling in the world of NFTs and AI

If you don't know what the problem with NFTs is, check out Dan Olson's ' Line Goes Up - The Problem with NFTs ' which thoroughly explains and then discredits the whole concept (and this came out in January 2022). Thank me later - it's a great watch and really well put together.

You may know of Mackmyra's Intelligens , supposedly an 'AI-produced' whisky, but that was actually produced in 2019 using programs which we now wouldn't call AI at all. Nowadays, you have Promethean Whiskey (based in North Carolina) claim that their AI will 'meticulously select the finest grains and the purest waters' to make whisky. Think about that for a second - what would that even mean in practice? Would you be importing water for distilling? 

So far, this seems to be the furthest that AI has infiltrated into whiskey production. However, the same cannot be said for the aesthetics of the whiskey world…

These Images are not Art

AI-generated images have this recognisable style which, once you notice it, becomes immediately uncanny. Search for 'whisky ai art' on Instagram to see what I mean. If you ever need a reminder that AI doesn't understand what it's producing, try to read any text which isn't the main focus of the image. You can't . It's all nonsense words or shapes which aren't letters at all.

Two examples among many taken from Instagram.

Not here to call out the people who made them, but to use these as examples

The whiskey industry surely has some of the world's best photographers and graphic designers , yet AI completely sidesteps them and their work. We have to ask why people will use Chat GPT instead of knocking up a quick poster with Canva (for free), or just use a photo they have lying around. Almost anything is simpler, more ethical, and more human than using AI-generated images for your whiskey brand. 

If ethical questions surrounding art and labor seem a bit abstract, you might find the other problems with AI more concerning…

When You Use AI, You Pour Carbon and Water into the Atmosphere

Did you know that typing a prompt or search into an AI program uses enormous amounts of energy? It doesn't take much effort to do, so you'd be forgiven for not knowing. 

Asking AI a question requires a biblical amount of computation , compared to a Google search. According to OpenAI researchers, the amount of computing power required to train AI has doubled roughly every 105 days since 2012! Poznan-based Whiskey Project AI boasts how they test theoretical blend combinations before production using AI, with one dram involving 97,656,250,000,000,000 combinations. 

That dwarfs the number of stars in the observable universe , or the number of cells in the bodies of 1 million people . That's one AI prompt, for one whiskey blend. More computing work means more electrical power is needed. Kate Crawford of Nature claims that searches driven by generative AI use four to five times the energy of a conventional web search: the International Energy Agency claims that the true ratio is ten times .

Global AI energy demand is projected to exceed the annual electricity consumption of Belgium by 2026. By that same year, 6% of the USA's entire energy production will be dedicated to AI data centers if current trends continue. Ireland's leading role within the European tech sector means that it might need to devote 35% of all its energy production to this infrastructure by 2026!

Local power grids can't suddenly ramp up to meet such spikes in demand, so data centers like Elon Musk's 'Memphis supercluster ' have taken to wheeling in portable natural gas generators. Microsoft just got the nuclear reactors at Three Mile Island turned back on, all to power new data centers, but such clean energy is the exception rather than the rule here.

Most of this insatiable demand for AI data centers is being filled by countries where wages are lower and materials are cheaper. Places like the Philippines are seeing massive diesel generators appear to power new centers, so they get all the pollution with none of the AI ​​boom's financial rewards. Google recently revealed that its carbon footprint jumped 48% from 2019 due to data center energy demands. 

Here's an example - the new JHB1 data center in Johor Bahru, which started running in July this year. Run by Australian firm AirTrunk, but located in Malaysia. CC 4.0 Licensed .

These centers also need enormous amounts of cooling water . They use evaporative cooling, which means that every time you use Chat GPT, you're throwing gallons of fresh, treated water into the atmosphere as vapor (where it also acts as a greenhouse gas, by the way). 

As with the power problem, these centers are often sucking up clean water for cooling in places where there just isn't enough to go around. Just one building at the new Goodyear center (near Phoenix) uses 56 million gallons of drinking-quality water per year - enough for 670 local families who live in a desert . 

All this means that each time you use an AI program, your request is being met with real, tangible amounts of water and fossil fuels. Local communities already hard-pressed to deal with pollution and drought are now having to deal with even more, not to mention that all of us are affected by the subsequent acceleration of climate change.

What Can You Do?

I know we've strayed a little ways from talking about whiskey directly, but this applies to any time you use AI, whether it's for making art of an imaginary whiskey bottle or promoting a tasting.

It's easy to avoid using AI, but do you want to? That's what it really comes down to. Despite all the hard facts about pollution and waste, this is almost a philosophical question . Now that you know, do you care about the problems created by pressing that button, hidden from you by distance? Imagine running the bath and turning on the heating every time you ask Chat GPT for an image, or a description of something. Would it really be worth it?

I understand the desire to push work off ourselves and onto machines, especially for busy work at a job you don't personally care about. But whiskey should be different - it's a luxury good, something fun we all enjoy. It should be an easy space to cut AI out of! Every time we take AI shortcuts, we have environmental decline. It's not about individual guilt. It's just that the results aren't worth it. 

From uncanny 'artwork' to just plain wrong information, AI means ceding control to something unaccountable, inscrutable, and wasteful. You can do better - you can make something better, and feel all the better for it. Slàinte Mhath!