In my day job (the one that actually pays the bills), I’m a tech guy. I stay on top of the emerging trends and try to see what’s coming around the corner, and something that seems to be scaring the heck out of print media folks (bloggers especially) is the idea that eventually an AI bot will be able to replace us all. It isn’t an outlandish idea, to be honest, so with the new release of ChatGPT I wanted to see how close it could come to actually replicating… me.
ChatGPT is the latest machine learning based algorithm from the same folks that brought us DALL-E 2 (which has been used to create some of the images used on this blog, like the one at the top of this article).
Early machine learning research focused on getting these algorithms to identify things: feed it a picture of a bird and it spits back “bird”. (Hence why you’ve probably been asked to identify traffic lights and cars to prove you’re a human — what you are actually doing is providing data that will then be used to train and improve those “models” to help them find those objects in the future.)
The next level of machine learning is going the opposite direction: ask it for a “bird” and it spits back a picture of a bird. (Or at least something resembling a bird.) Machine learning has been getting remarkably good at this recently, and ChatGPT is the latest text based model that has been fed everything from The Bible to random Tumblr posts and is trying to construct a text response to whatever prompt you give it.
Similar machine learning algorithms are already writing news articles in some publications, and it makes sense that eventually this technology might be good enough that long form articles like the ones you read here may be generated this way. To understand how close we are to that particular future, I fed the new ChatGPT algorithm two prompts to get it to write a full review. The prompts were:
Write an outline for a review of Jack Daniel's Tennessee Whiskey
Write a draft review of Jack Daniel's Tennessee Whiskey using the outline you just stated in the style of a snarky blogger
The results were… interesting. See below, and enjoy.
The AI Generated Review
Once upon a time, in a land far, far away (okay, just Tennessee), a man named Jasper Newton “Jack” Daniel decided to start distilling his own whiskey. And thus, Jack Daniel’s Tennessee Whiskey was born. Fast forward a few decades and Jack Daniel’s has become a household name, known for its smooth flavor and iconic square bottle. But is this reputation deserved? I recently decided to find out for myself.
First things first, let’s talk about the appearance of this whiskey. The iconic square bottle is a nice touch, but let’s be real – it’s not like I’m going to base my entire opinion on the bottle shape. What really matters is what’s inside, and unfortunately, Jack Daniel’s falls short in the appearance department. The whiskey has a pale golden color and lacks the rich, amber hues of other premium whiskeys. As for legs, well, let’s just say I didn’t see any.
But enough about looks, let’s move on to the aroma. I have to say, I was pretty disappointed here too. The aroma of Jack Daniel’s is quite mild, with hints of corn and oak, but nothing particularly noteworthy. I was expecting more complexity and depth, but instead I was left with a rather dull and uninspiring scent.
When it comes to the taste, Jack Daniel’s is definitely a sipper. The first sip is a bit of a punch to the palate, with a strong alcohol flavor that’s tempered by a sweetness that comes through in the form of caramel and vanilla. As you continue to sip, you start to pick up on some woody and smoky flavors, which give the whiskey some depth and complexity.
The finish on Jack Daniel’s is surprisingly smooth, with the alcohol burn fading quickly and leaving a nice, sweet aftertaste that lingers on the tongue.
Overall, I have to say that I was pleasantly surprised by Jack Daniel’s. It’s not the most complex or nuanced whiskey out there, but it’s definitely drinkable and has a nice balance of flavors. As far as value goes, it’s pretty affordable and widely available, so if you’re in the market for a decent, no-frills Tennessee whiskey, Jack Daniel’s is definitely worth checking out. Cheers!ChatGPT
Reviewing The Review
Right off the bat, I’m frankly surprised that ChatGPT was able to pull off the tone and the style so well. It reads just like something one of my colleagues would have written, and it seems to be using the terminology appropriately. Even the format of the review is appropriate — not exactly following the 31W style guide, but it gets fairly close.
Heck, it even included a brief history of the brand! I was floored when I saw it starting to write Jasper’s name at the top of the review. The ability to tie together not only information about the product but also the larger brand is impressive.
Where this starts to fall apart is the details. You can read our review of Jack Daniel’s Tennessee Whiskey for comparison, and while ChatGPT gets the general components that should be part of a whiskey taste and aroma, it misses the critical flavors that actually make a Tennessee whiskey unique. That fruity banana tone, like a bottle of Hoppe’s #9 cleaning solvent, is the hallmark of the Lincoln County process and what sets this bottle apart from others on the shelf. And the comment about a “smoky” flavor in the review is way off, more appropriate for a scotch whisky than an American bourbon.
What we’re getting seems to be the generic tasting notes for a whiskey, wrapped in some brand information about Jack Daniel’s, and with some snark layered in for good measure.
Will Whiskey Reviewers Be Replaced By AI?
Articles generated via machine learning might become more prevalent as time goes on, improving to the point where they might be indistinguishable from an actual human. But while you can do that for something like generic financial statements from corporations without much fuss, tasting (whether it’s whiskey, wine, food, etc) is one area where machines are unlikely to make a huge difference any time soon.
Spirits are something that requires actual human taste buds to sample and understand. There’s no way that a machine can synthesize thousands of reviews and accurately generate a new review of a brand new release bottle — that’s something that a reviewer needs to taste, consider, contextualize, and discuss. In short, an expertise that is unlikely to be replicated in code in my lifetime.