Are you getting advice from AI on Reefing questions?

[Hatuey]

Yes, I use AI for reefing advice, but more as a tool to validate information and refine my approach rather than a primary source. My technique involves asking targeted, practical questions related to my specific tank setup, equipment, and livestock. I focus on real-world problem-solving, such as managing nutrient levels, optimizing equipment use, and troubleshooting reef tank challenges.



To get more fine-tuned responses, I screenshot my parameters from Aquarimate along with my recent activities (like water changes, replacing GFO, etc.). This gives AI more context, allowing for tailored recommendations. Additionally, if I disagree with a suggestion, I explain why and provide my preference. For instance, when it recommended dosing 5ml of NoPox, I pushed back, saying, “No way!” and started with 1ml instead, also telling it to be more conservative with future dosage recommendations.



I also created a dedicated Reef Tank folder (I believe this is only available on the paid version), where I store all my reefing-related questions. This keeps everything in one place and allows AI to give more informed responses based on past interactions.



I love AI and try to use it as much as possible in everything I do. In fact, I even created a Reef Tank GPT, though I never published it—now considering it.



Appreciate your post!
(Response Powered by AI)

 

[klc]

OMG why all of a sudden is everything about AI? And who coined the term "AI"? Everyone talks about it and uses it in conversation like it's something new. It isn't and it's really annoying.

 

[gbroadbridge]

OMG why all of a sudden is everything about AI? And who coined the term "AI"? Everyone talks about it and uses it in conversation like it's something new. It isn't and it's really annoying.

You can't do that, Dave.

 

[Gumbies R Us]

I use AI as more of as a tool and not so much as a “end all be all method”. If I need help on something with my tank, I go straight to reef2reef for information not AI. Generally if I ask questions on google now it’ll give me an “AI” response but it’s usually a summary from 3 websites with the website posted directly below the response. It’s nice in those scenarios when I just need a quick answer to a question and I don’t want to spend time digging through a website to find some random information. Also, I find AI useful when I need to identify plants just to figure out what kind they are, haha.

 

[TacticalBacon]

Based on your picture you must like AI, but AI will never have years of experience and actual eyeballs to look at the animals reactions and just reefing instinct. Yes it will give general answers, but it's a no good imo.

I have a co-worker who loves AI and the major problem I’ve found is, AI doesn’t take into account the entire situation.

People say you have to be as specific as possible but you can’t always take into account minor details that experts at their job are able to account for without writing 20 paragraphs of background.

I use AI to locate some agency regulations that aren’t always easy to find and still it will miss the mark because of the language barrier, specifically, an Agency regulation may refer to set of regulations with different jargon than industry does.

It is also very interesting to be in a conversation with someone on the phone or web-meeting when they’ve relied on AI for an argument and when you start parsing it out and are argue to validate the point, the AI user is quickly at a disadvantage as they do not have the base line foundation to know they AI response is actual valid or invalid.

Or as Randy has proven. AI is not built to challenge experts and can be wrong when their is unknown variables at play.

In a Perfect Simulation with all known facts and the ability to scour millions of data sources, AI in theory should be accurate and powerful but Perfect doesn’t exist in a Saltwater Tank or situation that require a certain level of intuition, experience, and common sense.

 

[Randy Holmes-Farley]

You can't do that, Dave.

lol

 

[Randy Holmes-Farley]

At least with google chrome search AI, I have found thet the exact way the question is asked will give entirely different answers. That has come up in threads. Here's one that we had to sort AI fact from fiction, where the exact way the question is asked changes the answer to the 100% opposite answer.

Search on:

"zinc does not leach into water from PVC pipes"

and the google search AI says:

This statement is true; zinc does not typically leach into water from PVC pipes because PVC, by its chemical composition, does not contain zinc; the primary source of zinc in drinking water is usually from corroding galvanized pipes which are coated with zinc to prevent rust.
Key points to remember:
PVC composition:
Polyvinyl chloride (PVC) is a plastic material that does not naturally include zinc in its chemical structure.
Zinc source in water:
When zinc is found in drinking water, it usually comes from galvanized pipes, which are made of steel coated with zinc.
Leaching definition:
Leaching refers to the process where a substance dissolves and moves from one material into another, like chemicals from a pipe into water.

Then ask it
"does PVC pipe contain zinc?"

Yes, zinc is sometimes used as a stabilizer in polyvinyl chloride (PVC). Zinc stearates are a key component in PVC production, and they help to improve the quality and durability of PVC products.

Then ask it:

"Can zinc stearate leach from PVC pipe?"

Yes, zinc stearate, which is commonly used as a stabilizer in PVC pipes, can leach out of the plastic and into the surrounding environment, particularly when exposed to certain conditions like high temperatures or prolonged contact with liquids, meaning it can potentially leach from a PVC pipe.

 

[Northern Flicker]

heck no

 

[BeanAnimal]

You can't do that, Dave.

lol - one of the classics

 

[brandon429]

I asked chat gpt how to transfer a reef tank and guarantee success

After it stole + repeated 90% of the things we do in posts, the 10% omitted brings the certainty part which guarantees success.

That gap is where youd want to consult a human vs a bot, unless the reef in question is expendable.

It's final half page answer gave the reader the impression there were multiple options they could pick from, attaching no real matter either way. Wishy washy is chat gpt, wishy washy will kill some reefs who consult it for xfer strategy.

On the flipside it's fun to use it as a reflection of society to make a point on who's getting ripped off in reefing:

 

[brandon429]

Article bot #56777?

Coral bleaching in marine aquariums – causes and prevention | Aquaforest

Hi, today we have prepared for you an article on coral bleaching. If you are interested in how to avoid it read it. The full article is available on our website here. Coral bleaching in marine aquariums – causes and prevention Coral bleaching is a serious issue that can turn a thriving marine...

www.reef2reef.com

 

[drewzaun]

Parse my post and find out how AI can’t always be right. It m

Exactly. At least in its current iteration, AI will never give you the correct answer, only the most popular one.

One aspect of AI I am playing with is in a newer app called Reefability. I use Aquarimate to track all the things I care about, but I’m experimenting with Reefability also. They claim it uses AI and idk, so far I’m not getting anything out of it but I assume it needs more people and therefore data.

But the ai bot answer in Google is always suspect. And to me that’s terrible. If I ask a question here I can search up whoever answered me to see more about who they are.

In Wikipedia a decent article lists its sources. Peer review publications always do this.

I haven’t seen an AI answer provide anything to validate its answer. Even a simple Google search points you to pages that you yourself curate for value.

To me right now AI is the equivalent of forming an opinion based on reading the headline and not the article.

 

[brandon429]

And it is allowed to write articles now too.

 

[drewzaun]

with google chrome search AI, I have found thet the exact way the question is asked will give entirely different answers

That right there is scary given the amount of people who will quickly begin to rely on the bot answers as facts.

I learned the hard way that you need to consider how to word your search engine query’s when my boss asked me why HR was asking him why I was searching pornography on my work computer. I wasn’t but the terms were close enough the Googles figured I meant something else. That was funny…but this may not be.

You are smart enough to either know what someone is asking even though they may not ask the question the way someone with a more specialized education would. At the least you will ask questions to narrow down what they want.

AI will answer with confidence and I fear stupid people on the internet much more than someone from Churchill CA fears running into a bear while taking out the trash…

 

[rtparty]

OMG why all of a sudden is everything about AI? And who coined the term "AI"? Everyone talks about it and uses it in conversation like it's something new. It isn't and it's really annoying.

 

[BeanAnimal]

Exactly. At least in its current iteration, AI will never give you the correct answer, only the most popular one.

That is not correct either. You don't know what answer it will give you, and 4o connected to the internet will give you a different answer than a version not connected.

I asked (not reefing) yesterday:
"Does Glock manufacture an OEM slide for G17 Gen 3 with MOS cut". Not only was the answer "no" but it was emphatic with 7 paragraphs about why Glock would not do that and how 3rd party parts could cause reliability issues and tips and tricks and other options for optics on a G17 Gen... none of what I asked for.

Now, I know for a fact, that these were announced a few months ago, as OEM Glock. Iogged into a second GPT account and simply said....
"Glock manufactures an OEM slide for G17 Gen 3 with MOS cut." -- The responses went into great detail about why Glock has introduced the OEM replacement slide with the MOS cut and the benefits of upgrading, as well as Glocks ethos regarding reliability and that even if the slide was not availabel, how well regarded many 3rd part slides with optics cuts are.

TWO POLAR OPPOSITE RESPONSES -- BOTH STATED NOT ONLY HAS HARD FACT -- BUT WITH paragraphs full of supporting arguments, also presented as fact.

It does not regurgitate what is popular, it just makes it up as it goes, filling in blanks to what sounds good.

To me right now AI is the equivalent of forming an opinion based on reading the headline and not the article.

Worse that that, it just make stuff up to make you feel like you are having a conversation with a human.

 

[Randy Holmes-Farley]

That is not correct either. You don't know what answer it will give you, and 4o connected to the internet will give you a different answer than a version not connected.

I asked (not reefing) yesterday:
"Does Glock manufacture an OEM slide for G17 Gen 3 with MOS cut". Not only was the answer "no" but it was emphatic with 7 paragraphs about why Glock would not do that and how 3rd party parts could cause reliability issues and tips and tricks and other options for optics on a G17 Gen... none of what I asked for.

Now, I know for a fact, that these were announced a few months ago, as OEM Glock. Iogged into a second GPT account and simply said....
"Glock manufactures an OEM slide for G17 Gen 3 with MOS cut." -- The responses went into great detail about why Glock has introduced the OEM replacement slide with the MOS cut and the benefits of upgrading, as well as Glocks ethos regarding reliability and that even if the slide was not availabel, how well regarded many 3rd part slides with optics cuts are.

TWO POLAR OPPOSITE RESPONSES -- BOTH STATED NOT ONLY HAS HARD FACT -- BUT WITH paragraphs full of supporting arguments, also presented as fact.

It does not regurgitate what is popular, it just makes it up as it goes, filling in blanks to what sounds good.




Worse that that, it just make stuff up to make you feel like you are having a conversation with a human.

Have you tried asking it whether bolus dosing is better than kalkwasser? lol

 

[Randy Holmes-Farley]

Here are some fun tidbits from the google ai search:

search:

is bolus dosing better than kalkwasser

Answer:


In most cases, bolus dosing is generally considered a more controlled and potentially better option than kalkwasser for maintaining calcium and alkalinity in a reef tank, as it allows for precise dosing of the required amount at once, while kalkwasser can be more difficult to manage due to its potential for pH swings and imbalances between calcium and alkalinity levels if not carefully monitored.

• Stability:
  Bolus dosing provides a more stable environment by delivering the entire daily dose at once, minimizing fluctuations in pH and other parameters.
• With bolus dosing, you can accurately calculate the necessary amount of calcium and alkalinity to add, whereas kalkwasser can be more challenging to dose precisely due to its dependence on water changes and potential precipitation issues.
• • pH management:
    Kalkwasser can significantly raise pH if not carefully managed, which can stress corals, while bolus dosing can be designed to minimize pH swings.
    
    
  • Complexity:
    While bolus dosing may require a slightly more complex setup with dedicated dosing pumps, it often provides more fine-tuned control over water chemistry.
  
  However, kalkwasser can be beneficial in certain situations:
  • Lower maintenance:
    For smaller tanks or those with minimal coral growth, kalkwasser can be a simpler option to maintain calcium and alkalinity.
  • ​
  • Phosphate reduction:
    Kalkwasser can help precipitate phosphates out of the water column, which can be beneficial in tanks with algae issues.
    
    
    ​

Randy comment: some of those assertions (in red) are utterly ridiculous.

 

[charleydavis]

Artificial intelligence

Computer algorithm to gather information and it is supposed to learn I guess but it is not doing to great at some things.

Asking a bot questions about your reef


To answer the OP's questions, I don't use it purposely but when you google something it gives a AI summary. I usually skim that then actually look for something more trustworthy.

AI cannot distinguish what is trustworthy and what is not. The human has to do that part. AI can summarize a lot of information very quickly and can be helpful, but the human has to make the decisions.

 

[drewzaun]

That is not correct either. You don't know what answer it will give you, and 4o connected to the internet will give you a different answer than a version not connected.

To be fair what I have said so far is very simplistic and entirely based on my own experiences, colored by my continued disappointment in how tech is employed by the masses. It sounds like we agree but you are far more invested lol.

FWIW I think there is great potential here, but the developers need to do at least 2 things to ease my concerns.

First, send this slop back to the kitchen, it isn’t done cooking yet.

Second, point the crawlers only to sources that are reliable. If we take nutrition as a subject, the internet is filled with millions of examples that range from useless to dangerous, however there are places where information can be trusted. Most people haven’t the time, education or interest to parse good data out of peer reviewed articles or studies. An AI though, this could grab all the facts proven and present an easy to read common language answer. Hopefully we are moving in that direction, but the cynical side of me says it is easier and more lucrative not to, at least in the short term, which is all anyone seems to care about anymore.