Asking AI....

[Mont Y. Märzen]

Any freedom I sacrifice by using Siri, or AI is almost surely already known info. I'm ex Army, so they already know quite a bit about me. I have some title 2 stuff, so there's that. I've lobbied and testified at fed & state level years back.

I should probably offer a correction to my 'not worried about more surveillance'. We're veering way off here, but I'm vehemently opposed to DHS, it's a boondoggle. TSA, which is nothing but security theater, FISA and patriot act powers. So I probably should have chose better language but I'd be genuinely surprised if Siri is ratting me out to "The MAN". Gas buddy did, that's easily searchable. Ditto with GM's on board data sold off to data brokers, bought by insurance companies jacking up lots of driver's rates. And so on.

I'll continue using AI - work kind of requires me to know it. I may even do some research with it about beer, so please don't excommunicate me.

I don't think it is a security threat or any other kind of threat. People can be though. And folks are already blaming stuff on it, claiming 'it' messed up. No, no, the person using it for other than its designed and intended purpose messed up.

This tech isn't going to wipe us out. People are going to do that because they stopped thinking and taking ownership of their actions and turned it over to a Magic 8-Ball because they think it is more than that.

I also don't think this could have come about sooner. Folks were already dumbed down after 20 years of addiction to their tethers. That couldn't have happened back in the early 2000s or prior.

The future, however long it lasts, is a Venn Diagram of Nineteen-Eighty-Four, Brave New World, Animal Farm, and Fahrenheit 451. Some would argue we are already there.

Of course, the world won't implode based on how you formulate a beer recipe.

 

[NorthCoastBrewer]

... going back to an earlier reply ...

We've implemented AI tools at work and it's been an absolute gamechanger. Some of my team are performing tasks in half the time using a AI chatbot skill/tool that I designed. I use it daily.

Like any other tool it has strengths and limitations and its up to the user to understand those weaknesses to use it most effectively.

One of the things I noted, over a couple of years of using 'Generative AI' for hobby and professional subjects is this: the models / services tend to change noticeably every six months or so.

And it looks like we're in the "things are changing" period.

Six months ago, for code generation, "prompt harder" (with layers of 'prompts' to establish context) seemed to work better. Recently, I started seeing suggestions that those prompt layers may be model version specific.

Recently, tech companies with large 'Generative AI' usage bills has been in the news. Will be there a return to measuring 'return on investment'?

Paying for the service also seems to be changing - I'm starting to see comments about changes in 'terms of service' from some providers that hint at advertising models, perhaps even in lower end subscriptions.

 

[Brew Cat]

I don't think it is a security threat or any other kind of threat. People can be though. And folks are already blaming stuff on it, claiming 'it' messed up. No, no, the person using it for other than its designed and intended purpose messed up.

This tech isn't going to wipe us out. People are going to do that because they stopped thinking and taking ownership of their actions and turned it over to a Magic 8-Ball because they think it is more than that.

I also don't think this could have come about sooner. Folks were already dumbed down after 20 years of addiction to their tethers. That couldn't have happened back in the early 2000s or prior.

The future, however long it lasts, is a Venn Diagram of Nineteen-Eighty-Four, Brave New World, Animal Farm, and Fahrenheit 451. Some would argue we are already there.

Of course, the world won't implode based on how you formulate a beer recipe.

You forgot Soilient Green

 

[west1m]

You forgot Soilient Green

And the Hal-9000

 

[Dave Y]

When I see this thread, two things happen:

First, I think of a guy named Al (Alfred), with whom I used to work. What a character. But I’d never ask him anything, the answer would typically be what we now call a hallucination.

Then I think of John Valby’s song that has the chorus ‘Ai yi yi yi…”. Somewhat like the Frito Bandito song, with lyrics inappropriate for this forum.

And IDK if Siri is AI. Maybe AS. She’s pretty dopey.

Great. Now, thanks to this response, it makes me think of ALF
So... it started out as AI.
Then Al and now ALF, soylent green and HAL.

Open the pod bay doors, HAL !

 

[Brew Cat]

Great. Now, thanks to this response, it makes me think of ALF
So... it started out as AI.
Then Al and now ALF, soylent green and HAL.

Open the pod bay doors, HAL !

" I'm sorry Dave I'm afraid I can't do that"

 

[Donoroto]

We digress.

As usual…

 

[Ward Chillington]

Agreed @Mont Y. Märzen "novelty parlor tricks"!

I think most of us see what we put into a glass as a form of art and what we are part of is a craft and a set of skills that stands on the shoulders of over 10,000 years of other humans tinkering around with water, something that ferments and something that was thought to be a miracle until about 200 years ago!

Anyone the believed that anything AI produced and considered "art" is missing the fact that it's plagiary at best. It is only copying what the large language model has scoured the internet for often including copywritten works.

What we brew has human hands on it and in it. What we brew has human thought and judgement in it.
What we brew is created within systems and processes we as humans have built and perfected as best we can to create our art.

You'd be hard pressed to find a craft brewer that considers their finished product as something that could be equally created by a box that only knows the difference between a one and a zero.

 

[Megary]

I have not used AI (mostly because I really don’t know how best to do so)…but I’m not opposed to taking ideas from wherever I can get them. And I’m not sure I see any honorable distinction between searching recipes on Brewer’s Friend or the AHA or reading them in Designing Great Beers or How to Brew or asking AI.

It’s all a pool, weight accordingly.

 

[Ozarks Mountain Brewery]

when asking a how to question, its a better start than youtube, so an example, how to get weeds out of your yard or fix a ford that wont start, this part of ai works good

 

[Mont Y. Märzen]

I have not used AI (mostly because I really don’t know how best to do so)…but I’m not opposed to taking ideas from wherever I can get them. And I’m not sure I see any honorable distinction between searching recipes on Brewer’s Friend or the AHA or reading them in Designing Great Beers or How to Brew or asking AI.

It’s all a pool, weight accordingly.

I'm not sure 'honorable' for distinction is necessary, as opposed to accuracy, but BF and other recipes are those of others being brewed; Book sources are the same, and more likely to isolate award winning examples, probably even multiple-award winning, and probably high-profile/score award winning. (that is, those that consistently win and get high scores in big competitions from very experienced judges) Those sources (especially the books) are the closest you can get to an 'amateur example of a style' similar to a 'commercial example'.

A chatbot is just going to regurgitate some probabilistic mash-up of whatever it has available as 'training/source data'. And it will do that with zero understanding or comprehension of what you asked or what it spit out.

Those two are in alternate realities from each other.

 

[Mont Y. Märzen]

You'd be hard pressed to find a craft brewer that considers their finished product as something that could be equally created by a box that only knows the difference between a one and a zero.

Even worse, the box doesn't "know" the one or the zero, what that means, or anything at all. The "I" in "AI" is entirely a ruse.

 

[Megary]

I'm not sure 'honorable' for distinction is necessary, as opposed to accuracy, but BF and other recipes are those of others being brewed; Book sources are the same, and more likely to isolate award winning examples, probably even multiple-award winning, and probably high-profile/score award winning. (that is, those that consistently win and get high scores in big competitions from very experienced judges) Those sources (especially the books) are the closest you can get to an 'amateur example of a style' similar to a 'commercial example'.

A chatbot is just going to regurgitate some probabilistic mash-up of whatever it has available as 'training/source data'. And it will do that with zero understanding or comprehension of what you asked or what it spit out.

Those two are in alternate realities from each other.

Like I said, it’s all a pool, weight accordingly. If you don’t want to give any value to an AI recipe, I’m good with that. Others may find them useful to varying degrees.

 

[Zefram]

I guess I have two disagreements with water profiles in general and using AI for coming up with a profile for you.

Water profiles have often been exaggerated among homebrewers and pro's alike. Yeah, it's important, but you really have to have everything else in your process nailed down hard. Even when everything else is perfect, it has a small impact on the beer. If any thing is off on the mash, fermentation, ingredients, etc., it will skew the results of the mineral profile to the point of negation.

Second, everyone perceives flavors slightly different. High levels of sulfate don't taste dry to me, it adds a mineral finish. Others say something else. Understanding you palate will help you discern flavors more accurately, so when some says this finish is dry or not dry or even sweet, you will be able to have a more intelligent conversation about. Unfortunately, you may have to "guess and check" as a process of learning. It's really to important to understand not only the "what" in brewing, but the "whys" may be even more important. AI can get you started, but in the end it's a cooking art done by humans for humans.

I suppose I should clarify. Having no clue where to start with the water additives, I use AI to give me a starting point based on a profile. From there I adjust. So there's still some "guess and check" going on, but from a reasonable starting point.

I think of it no differently than wanting to cook something I've never cooked before. Use the interwebs to get an idea of a recipe, then tinker with it until it suits my preferences (or at least the family's, in this scenario).

 

[Dave Y]

Agreed @Mont Y. Märzen "novelty parlor tricks"!

I think most of us see what we put into a glass as a form of art and what we are part of is a craft and a set of skills that stands on the shoulders of over 10,000 years of other humans tinkering around with water, something that ferments and something that was thought to be a miracle until about 200 years ago!

Anyone the believed that anything AI produced and considered "art" is missing the fact that it's plagiary at best. It is only copying what the large language model has scoured the internet for often including copywritten works.

What we brew has human hands on it and in it. What we brew has human thought and judgement in it.
What we brew is created within systems and processes we as humans have built and perfected as best we can to create our art.

You'd be hard pressed to find a craft brewer that considers their finished product as something that could be equally created by a box that only knows the difference between a one and a zero.

A recipe that someone else _created_ that you're / we're using is exactly what you're describing from an AI source.
You're recreating something that someone else made available.

How many of us here create their own recipes; for better or worse? ( raises hand sheepishly )
Is it disqualified as 'art' if we post our efforts in the 'recipes for feedback' subforum ? Because we're basically asking for help.

...

A chatbot is just going to regurgitate some probabilistic mash-up of whatever it has available as 'training/source data'. And it will do that with zero understanding or comprehension of what you asked or what it spit out.

Those two are in alternate realities from each other.

It very much does not do that, it is far, far more capable.

Which is not to say it is right, proper, correct, ethical, hono(u)rable or cheating, it is only to say the current models are better than this.

 

[Mont Y. Märzen]

A recipe that someone else _created_ that you're / we're using is exactly what you're describing from an AI source.
You're recreating something that someone else made available.

Something some*one* else made available, that they have brewed. That someone has cognition and understanding. They know at least vaguely their malt from a car tire or a Daffodil. The ChatBot knows ZERO.

Using someone else's recipe as a starting (or even ending) point is not the same as requesting an assemblage of *possibly* connected or related words that happen to resemble a recipe. Those two are entirely different things. One is the real deal. The other one is entirely artificial and not even designed to be real, but only have the appearance of being real.

How many of us here create their own recipes; for better or worse? ( raises hand sheepishly )
Is it disqualified as 'art' if we post our efforts in the 'recipes for feedback' subforum ? Because we're basically asking for help.

When you start a chat session with an LLM, you are NOT asking for help. You are prompting an algorithm to spit back words that appear to be a reply based on the relation of the words you typed to words in its dataset.

This isn't even a 'search engine' because it doesn't return citations to actual human knowledge resources written by real people who have some clue about the subject. (and maybe lots of clues, and lots of experiences)

It's far far worse. It is mashing up bits and pieces from all of those sources and spitting them out to you, as if it were a compiled summary. But it has no idea what it is doing, or have any idea about the topic at all, or any parts of it. If a human tried to do the same thing, I'd trust that result more because at least the human does have cognition of some things and can apply and extrapolate and at least to their ability, think logically. An LLM does not even think.

It very much does not do that, it is far, far more capable.

Which is not to say it is right, proper, correct, ethical, hono(u)rable or cheating, it is only to say the current models are better than this.

I have yet to see that LLMs have moved on from the 'Attention' model. Yes, some are refining this, but they are still this at the core. Are you familiar with the concept? If not, it is a very interesting read. Wikipedia can certainly provide the 30k foot overview to get anyone started.

Until and unless a new model is developed, the current systems are all just word generators. There is not even a path for cognition and understanding and thus no path for learning, and thankfully no path for sentience.

 

[Dave Y]

...

I have yet to see that LLMs have moved on from the 'Attention' model. Yes, some are refining this, but they are still this at the core. Are you familiar with the concept? If not, it is a very interesting read. Wikipedia can certainly provide the 30k foot overview to get anyone started.

Until and unless a new model is developed, the current systems are all just word generators. There is not even a path for cognition and understanding and thus no path for learning, and thankfully no path for sentience.

I've mentioned previously, my employer is heavily invested in AI development, use and implementation. My team and I have been using it for some time, and I do have a grasp on it.

And that's a reason why I'm disagreeing with your assessment. Another reason is I have used it and mentioned previously it came perilously close to my 6c hop recipe I submitted for feedback here and brewed. All the updates I made to it based on feedback were from subjective 'that hop might just suck how you're going to use it' and similar. All "subjective" but all still valid input. Also, based on how well this beer turned out, I'll concede that every change recommended and implemented was right. The AI agent came within a couple of recipe stylistic differences and also prompted for 'crystal malts or no?' which was also mentioned in my recipe thread.

So while the agent cannot know what the malt smells or tastes like, it does aggregate and recommend based on sources available to it or referenced. You are able to provide it your own sources to compare and it absolutely does report citations. Ditto for hops or yeast, though I think I got better hop feedback here.

Yes or no; Have you used or tried to use a current model AI agent to help design a beer recipe or check against a known good recipe?