Beer expert GPT

Homebrew Talk - Beer, Wine, Mead, & Cider Brewing Discussion Forum

Help Support Homebrew Talk - Beer, Wine, Mead, & Cider Brewing Discussion Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
🙍‍♂️🙍‍♀️

The error(s) and/or error causing omission(s) in the sample are obvious to a human with actual brewing experience.







overview: What Is ChatGPT Doing … (link) tl;dr: subtitle: "It’s Just Adding One Word at a Time"

an experience report: A two-hour walking tour with ChatGPT: ... (link) "So all I manage to do is contribute free labour to ChatGPT by refining its data on Sydney’s public transport options. Given my long tenure as an unpaid content creator for Meta, that’s probably the least of my big tech complicity."







There is obviously a lot of active research around "hallucinations" in LLMs.

Mean while, in the real world, quality curated content includes time proven techniques like peer reviews and technical editors.

Paying an intermediary with a known ability to provide incorrect information doesn't have a lot of appeal to me at the moment.
 
GitHub Copilot - Wikipedia (edited for length, for focus related to this topic, and emphasis added).

"Reception

"Since Copilot's release, there have been concerns with its security and educational impact, as well as licensing controversy surrounding the code it produces.[17][11][18]

[...]

"Licensing controversy

[...] a class-action lawsuit filed in November 2022 called this "pure speculation", asserting that "no Court has considered the question of whether 'training ML systems on public data is fair use.'"[20] The lawsuit from Joseph Saveri Law Firm, LLP challenges the legality of Copilot on several claims, ranging from breach of contract with GitHub's users, to breach of privacy under the CCPA for sharing PII.[21][20]



There is also a "follow people (and their motivations), not events" idea, but that (I think) is way out of scope for this home brewing forum topic.

Can we limit this topic to "LLMs in the context of home brewing"?
 
After reading the example conversation, I'll say that I have seen worse content in forum posts, but with the latter, there are usually challenges made. For example, "Heat your water to around 152°F (67°C) and steep the grains for 60 minutes" in a forum post would result in an avalanche of corrections regarding mash temp vs strike water temp (and consideration of mash pH, etc.). Relying directly on the quoted statement from a beer "expert" could result in a largely unfermentable wort.

The need for peer review mentioned by @BrewnWKopperKat is real, whether that review is by forum members or (in other media) by technical editors or other credentialed experts.

(Similarly (but a little off topic), don't ever rely directly on the advice of the Dude-At-The-LHBS, unless you know for certain that he/she is an expert, because the mere holding of that position isn't evidence of expertise.)

This is not a criticism of @pepindavid's effort, but rather of ChatGPT and similar models. Fascinating technology, but not ready for prime time for critical application.
 
Last edited:
How can something with zero sensory perception offer advice on creating anything that is 100% sensory? You cannot learn anything from AI. You must use your own critical thinking skills.
I think we’d all agree that research is useful for learning in general, and for making beer in specific. Do Google searches, read books, look at published recipes, go through blogs and forum posts…

The potential of AI is that it’s essentially doing that: gathering a tremendous amount of reference information and processing it (in difficult to comprehend ways) to synthesize a result. It doesn’t matter that the computer has no senses, because it’s relying on the records produced by humans that do have them.

In practice, I’m impressed by how far it’s come, and how quickly. The results produced by GPT and its ilk are often good starting points, and in some cases can save a lot of time. But we’re still at the stage where discrimination between data is iffy, and so you need to take that starting point and look over it pretty carefully if you don’t want to go badly wrong.

This is amateur speculation, but my guess is that it won’t be too long before the answers you get from AI will be at least as good as what you’d come up with yourself after hours or days or more of intensive research.

On a different topic, I’d also guess that advances in data analysis, paired with multi-D chromatography/mass spectrometry, are going to crack the code relating chemical composition to “this tastes good.” And that large breweries will ultimately completely replace sensory panels and consumer research with analytical labs.

I, for one, welcome the arrival of our automated brewing overlords.
 
do LLMs understand what is generated?
1699908761338.png
 
How can something with zero sensory perception offer advice on creating anything that is 100% sensory? You cannot learn anything from AI. You must use your own critical thinking skills.
I think AI can do a good job after a couple of beers (on my part)!
 
Back
Top