Possibly the Final AI Post on Beervana
Midjourney. Prompt “AI Beer, v 6”
Two years ago, I wrote a post on a topic that hadn’t received a ton of media attention. I’d heard about a text generator on an obscure site and discovered that many who had used the program—it was still in a beta test form—were really impressed. It could generate human-sounding text, and all you had to do was give it a boost by including a few prompt words. The program, also obscure, was called GPT-3.
It would be difficult to imagine a human living in the US who has not heard the word ChatGPT, the current version of GPT-3, today. It feels like it has been a part of our vocabulary for thirty years. Googling the word returns a billion results; “beer,” a thing that has existed for 10,000 years, returns just two billion. Every major tech platform is racing to place their chatbots on the market. The chatbots’ capacity to return grammatical, sensible text has already transformed education, and estimates of their impact in the workplace range from substantial to existential. All of this has happened in the space of two years.
Last year, things were so hot in AI that I followed up with a three-post series (“AI is Here. Now What Do We Do With It?” “Using AI Ethically” and “The AI Nightmare Scenarios”). I was shocked that in one short year I could go from being impressed that a chatbot could pass the Turing test to wondering whether AI would be a more society-changing tech than the internet. A few months later, during the apocalyptic heat wave rolling across the planet, we had moved on to calculating our p(doom), or the probability that ChatGPT was a beta version of Skynet. Well, here we are two years out, and I thought it would be worth returning to the topic one last time. Of course, no promises—if Skynet arrives, I reserve the right to discuss it again. (Favorably, of course, to appease our robot overlords.)
It’s shocking how this technology, which once seemed so advanced it might have been magic, now is used in applications that fill us with the wonder of a socket wrench. I can, for example, use my Squarespace text composer to help me write my posts: This implementation of AI technology on Squarespace marks a fundamental shift in how users interact with the platform, promising more efficient and intuitive website development processes. I don’t love that previous sentence, which sounds like the usual corporate spin, but it is well-constructed and free of errors. For most of the sentences we write or speak, that’s fantastic. And it comes with the push of a digital button. (People focus a lot on AI’s inaccuracies and “hallucinations,” which is appropriate to a point. But most of our communication doesn’t require facts—and even when it does, they’re easy to check.)
Large language models are already so effective that they are almost certain to replace routine, professional communication the way calculators replace long division. For certain tasks, writers will continue to compose sentences the old-fashioned way, but for the millions of regular writers, chatbots will save enormous time and improve their communication. The average human makes tons of mistakes in written communication—probably more than fifty percent of the time, even in short, simple blocks. AI never makes mistakes. I am just old enough to remember the dying debate over calculators in the classroom, which were identical to the current debate on chatbots. Calculators won out because we all agree that speedy, flawless math is better than slow, mistake-ridden long division.
The Brewery of the Future
AI has been so magnetizing—even beer bloggers are writing about it—because it may transform so many different different human activities. I used the example of text-generation because it’s the first major application of the tech, but it will penetrate every workplace. Take breweries. They exist in meatspace and produce a tangible product, but they’re still ripe targets for the kind of efficiencies AI offers. Breweries spend a huge amount of time managing inventory and ordering supplies. AI could automate this by monitoring ingredients and supplies, ordering based on productions schedules, and even troubleshooting when supply chains are tangled. They could be optimized to search for deals during ordering, finding the lowest prices, or quickest shipping, or both. Brewhouse automation is already pretty sophisticated, but AI might help improve efficiencies, particularly in the cellar. Basically every step along the way is part of a coordinated effort AI could streamline and minimize errors.
In brewery offices, AI will improve sales and marketing communications, analytics, strategic planning, and branding. Much as brewhouse ordering could be automated, routine beer sales and payments could be as well. Wholesalers, who have to manage the complex dance of SKUs, could use AI to streamline their own processes. Retailers could in turn improve their efficiency in ordering, interacting like gears with newly-automated wholesale systems to get beer from brewery to shelf faster and in volumes that more accurately mirror demand. Beer could arrive fresher at the retailer and less of it would fall out of code. Of course, all of this will reduce the number of workers needed to put a case of beer on the shelf—there’s that workplace effect—which will further save money.
In centuries past, breweries were quick to adopt new tech because they have always worked with narrow margins. AI is going to be a valuable tool for cost-saving, and large breweries will lead the way using it to automate systems. That doesn’t mean it will displace the human hand entirely. AI is great at doing routine things, but it has limitations. It’s not going to replace human creativity and decision-making at higher levels. Brewers are still going to be tinkering with new styles, ingredients, and processes, and business managers are still going to be crafting brands and setting long-term strategies. As with text-generation and calculators (and push-button brewhouses), AI will be a tool people deploy and manage.
It’s Already Here
The reason I suspect this will be my last post on AI because is because it’s already becoming old hat. I ran a quickie poll on Twitter/X to see how many people use AI. The large majority said they don’t use it at all (77%). Yet put another way, that means 23% use it for personal or work use, which is actually a big number for a tech that is brand new. And moreover, my poll understated results found in other, scientific polls.
According to one poll, a majority of workers have used AI, at least infrequently. Only a quarter use it daily or frequently, but another 30% use it occasionally.
The younger you are, the more likely you are to use AI: “Gen Zers (37%) and millennials (35%) are the most likely to have used AI in their jobs. Just 25% of Gen Xers and 17% of baby boomers report using AI tools like ChatGPT at work.”
Students use AI a lot. 40% have used AI, and 20% use it daily.
Workers of color use AI more than White workers: “41% of Asian employees, 38% of Black and 36% of Hispanic workers hav[e] used AI software in their roles. That compares with 23% of white workers who said the same.”
Part of the disparity in numbers is definitional. What constitutes “AI?” Google uses it in search and email apps we all use daily. That’s quite a bit different than generating an AI video. Still, the upshot is this: a lot of people are using AI already, and adoption rates will accelerate.
For what it’s worth, I agree with the folks who think AI is going to radically transform our lives. To take one example that I bumped up against in Budapest: translation. This isn’t new tech—I was absolutely delighted to be able to use Google to translate historic texts when I was writing The Beer Bible a decade ago. But it has gotten more accurate and efficient. You can translate signs with Google Translate, and we’re getting very close to being able to translate each other’s words in real time, Star Trek universal translator-style. (I actually pulled off a rudimentary version of this in Poland five years ago and managed to get into and see a doctor by speaking into a phone, having it translate to Polish, and then speak the words to the people at the clinic.)
Where my thinking has changed is that I don’t think the transition will be as shocking as I did a year ago. This isn’t going to be the car replacing the horse. It will be a slower, more orderly process as various tools emerge that we plug into our daily/work lives. The internet was a massive, transformational technology, too. But it wasn’t transformational instantly. We had to take different technologies and knit them together. The modern smart phone is a wondrous tool not because it harnesses so many different applications and knits them together. I suspect AI is going to look the same. It will bring together a bunch of small tools that will, eventually, create a world that we couldn’t have imagined or predicted back when I wrote my first post here.