[OpenFracture]

Why? When there are so many tried and true recipes available at your fingertips... why? When you can get recipes directly from craft brewers... why?

Because I want to learn new things.  I want to experiment with small changes.  With Chat GPT I can put in a recipe from another source and then ask for it to have a little more citrus, or maybe a little more malt character and it will adjust it.  My go to for recipes has always been BCS.  I am also interested in brewing some of the Brewdog recipes that they published.   I only get to brew once in a while, but I can play with recipes and read about this hobby everyday.  Looking at ChatGPT was just something else to make me think about it.

You are not going to learn anything from AI. You need to get your hands dirty and learn the flavors of each ingredient yourself. If you have time to read something read Randy Mosher's Mastering Homebrew. It will walk you through the recipe design process and how to use ingredients that work together well.

Please don’t let the number of posts next to my name confuse you.  I have brewed almost a hundred batches and have designed my own recipes using all of the great resources that have been mentioned here.  I have listened to hundreds of hours of brewing podcasts and enjoyed hundreds of homebrews of my own and from members of my local club.  Saying that someone will learn nothing from AI is ignorant of the power of AI.    When I say I want to learn from it, I mean that it can point me in the right direction because no matter how much I have read, watched, or listened to, it has more. 

[BrewnWKopperKat]

Please don’t let the number of posts next to my name confuse you.  I have brewed almost a hundred batches and have designed my own recipes using all of the great resources that have been mentioned here.  I have listened to hundreds of hours of brewing podcasts and enjoyed hundreds of homebrews of my own and from members of my local club.

My guess back in #1 was that you were aware that a number of craft breweries had used ChatGPT to generate recipes.  I would also guess that you were aware that there are a number of topics over in HomeBrewTalk or /r/homebrewing where home brewers with deep knowledge in  AI and LLM explained the strengths and weaknesses of LLMs.

When I say I want to learn from it, I mean that it can point me in the right direction because no matter how much I have read, watched, or listened to, it has more.

And, currently, it gets that for free, from the public internet.

This is had a "chilling" effect on public discussion in software development Q&A forums (like Stack Overflow). 

[Kevin]

Why? When there are so many tried and true recipes available at your fingertips... why? When you can get recipes directly from craft brewers... why?

Because I want to learn new things.  I want to experiment with small changes.  With Chat GPT I can put in a recipe from another source and then ask for it to have a little more citrus, or maybe a little more malt character and it will adjust it.  My go to for recipes has always been BCS.  I am also interested in brewing some of the Brewdog recipes that they published.   I only get to brew once in a while, but I can play with recipes and read about this hobby everyday.  Looking at ChatGPT was just something else to make me think about it.

You are not going to learn anything from AI. You need to get your hands dirty and learn the flavors of each ingredient yourself. If you have time to read something read Randy Mosher's Mastering Homebrew. It will walk you through the recipe design process and how to use ingredients that work together well.

Please don’t let the number of posts next to my name confuse you.  I have brewed almost a hundred batches and have designed my own recipes using all of the great resources that have been mentioned here.  I have listened to hundreds of hours of brewing podcasts and enjoyed hundreds of homebrews of my own and from members of my local club.  Saying that someone will learn nothing from AI is ignorant of the power of AI.    When I say I want to learn from it, I mean that it can point me in the right direction because no matter how much I have read, watched, or listened to, it has more.

I stand by my statement no matter what your experience or the number of posts. You simply can not develop critical thinking if you rely on AI to give you the answers. You have to roll up your sleeves, experience problems and use your brain to overcome them. However, you do you. As for others reading this with less experience; don't rely on AI. Roll up your sleeves, get your hands dirty, use your brain and think for yourself. The I in your head will serve you much better than AI.

[pete b]

I can see someone who is an experienced brewer benefitting from AI. It's not hard to imagine an AI recipe giving someone an idea that they hadn't considered then using their own experience and perhaps research to work with that idea. It's not something I am interested in myself but I think it's black and white thinking to say either you are doing critical thinking or blindly following what AI says to do. 

[OpenFracture]

I can see someone who is an experienced brewer benefitting from AI. It's not hard to imagine an AI recipe giving someone an idea that they hadn't considered then using their own experience and perhaps research to work with that idea. It's not something I am interested in myself but I think it's black and white thinking to say either you are doing critical thinking or blindly following what AI says to do.

This right here.

[dbeechum]

I can see using ChatGPT in the same way as writing prompts - a proverbial kickstarter for an idea. Give me something to kick the tires around, chew on, etc

I think my allergic reaction is the way some folks, at least from public statements, are using ChatGPT recipes as written without processing through our own experiences. (See the number of stunt "AI" brews that have been released commercially)

[BrewnWKopperKat]

/r/homebrewing has a current topic on creating/finding recipes: https://old.reddit.com/r/Homebrewing/comments/17qj0l6/creating_recipes_vs_finding_recipes/

[OpenFracture]

/r/homebrewing has a current topic on creating/finding recipes: https://old.reddit.com/r/Homebrewing/comments/17qj0l6/creating_recipes_vs_finding_recipes/

For the record, I did not post anything about ChatGPT there.  Haha.  I am usually a hybrid approach to recipe creation.  I will brew a recipe from another source, usually from Jamil's BCS, then make changes based on how I want the beer to change.  As I said in an earlier post, I only recently found the BrewDog database, and I would like to brew one or two of those to see how they come out. 

[BrewnWKopperKat]

/r/homebrewing has a current topic on creating/finding recipes: https://old.reddit.com/r/Homebrewing/comments/17qj0l6/creating_recipes_vs_finding_recipes/

For the record, I did not post anything about ChatGPT there.  Haha.  I am usually a hybrid approach to recipe creation.  I will brew a recipe from another source, usually from Jamil's BCS, then make changes based on how I want the beer to change.  As I said in an earlier post, I only recently found the BrewDog database, and I would like to brew one or two of those to see how they come out.

I thought the /r/homebrewing topic would be interesting as it covered a lot of ideas for how people approach recipe design. 

My first experience with the BrewDog recipes (a number of years ago) left me unimpressed.  At that time, the recipes I looked at left me with too many "blanks to fill in".  The recipes at the current web site are much better.  But it's a crowded market for recipes - and the recipe books I have don't require me to "fill in the blanks".     

What's missing from the "I used ChatGPT to create a recipe" topics I have seen is a discussion on the prompts that were used.  From what I've seen and experienced, 'quality' prompts can do a good job of focusing the statistically generated response (but the response is still subject to 'making stuff up').   

[HighVoltageMan!]

Maybe I'm the only one who finds recipe formulation to be one of the most enjoyable parts of brewing. Some of my recipes have won at the NHC, which I was super stoked about. I never leave well enough alone and I start tinkering with them again. It's certainly a learning opportunity for me, especially when I make a mistake.

I think letting AI write your recipes is a little like buying a sports car only to let the auto-pilot drive it for you. It takes all the fun out of it. 

[denny]

Maybe I'm the only one who finds recipe formulation to be one of the most enjoyable parts of brewing. Some of my recipes have won at the NHC, which I was super stoked about. I never leave well enough alone and I start tinkering with them again. It's certainly a learning opportunity for me, especially when I make a mistake.

I think letting AI write your recipes is a little like buying a sports car only to let the auto-pilot drive it for you. It takes all the fun out of it.

I agree with you. I brew for the process, not the beer. I have no desire to eliminate any part of that process.

[Megary]

I can see ChatGPT / AI as a jumping off point for recipe development.  No harm in that.  I would run it through my own personal filter, just like any other recipe I see on-line. 

[neuse]

I just found a feature in AI Chat that surprised me - I was using the Chat on Bing. When I use it, I normally get a response that has specific references for the statements. I felt that the AI Chat basically did detailed searches much faster than a human could. Today I did a chat about UV rays and eye damage. I got a helpful answer, but it didn't have any references. So I asked "What is the reference for this information?". The AI Chat answered:
"Certainly! The information I provided is based on widely accepted knowledge about UV radiation and eye health. Here are some reputable sources you can refer to for further reading:
1.   The New York Times: An article titled “How Tall Is Mount Everest? For Nepal, It’s a Touchy Question” discusses the height of Mount Everest and its significance.
2.   Wikipedia: The Mount Everest Wikipedia page contains comprehensive information about the mountain, including its elevation and historical measurements.
3.   USDA and Other Sources: For banana length, you can refer to the USDA’s average measurement of 7 to 8 inches for bananas. Additionally, articles like “How Long is a Banana?” and “What Is the Length of a Banana?” provide insights into banana dimensions.
Remember to always verify information from multiple reliable sources, especially when it comes to health-related matters."

First, I noticed that it does more than just a detailed search. Somehow, it uses "widely accepted knowledge" without any specific references. How does it know all this without specific references?

Second, the further reading it suggested was completely off-base.

[BrewBama]

Widely accepted knowledge ≈ the programmer’s narrative.


History doesn't lie. People simply lie about history.

[Clint Yeastwood]

It's important to remember that ChatGPT is a full-blown moron.

I tried it when it came out. I beat it in arguments because it lied, backtracked, and contradicted itself. It made up responses like a toddler who thinks dad will believe anything.

It can't write a simple Pascal program that works. Give it a try and see. I gave it a whole bunch of chances, and it was never able to do it.

I can't see trusting recipe construction to a machine that can't taste or smell. To me, it's useful mainly for simple stuff, like asking what's best for cleaning mineral stains off toilets.

It sounds like Denny and I are in agreement about one thing: the whole process is fun, and for me, the creativity of writing recipes is essential. If I had to use recipes other people, or my toaster, wrote, I'd probably quit.