ChatGPT and Reef Chemistry: We Try it Out!

[gbroadbridge]

I want to see an original thought or a new idea that can be vetted and confirmed to be true. AI is simulation not innovation.

Something I don't want to see.
If it's 'intelligent' enough to do that, it would probably be 'wise' enough to conceal it :)

 

[Randy Holmes-Farley Verified Community Expert]

Something I don't want to see.
If it's 'intelligent' enough to do that, it would probably be 'wise' enough to conceal it :)

uh oh. That's probably it. It's way smarter than it lets on so we don't shut it down. 2001 was just a bit late getting here.

 

[snackpack]

Something I don't want to see.
If it's 'intelligent' enough to do that, it would probably be 'wise' enough to conceal it :)

So about that

 

[mmoner]

Randy rating for this one :)

 

[SDchris]

I tried it with a simple math question, to which it gave the wrong answer. Another four tries and got four different answers, but still all wrong. I then gave it the correct answer to which it then gave the right answer but the working out was wrong.:confused-face:

 

[Randy Holmes-Farley Verified Community Expert]

Randy rating for this one :)

Randy rating for the first discussion, D. Not at all good. The first paragraph is fine. The recommended alk/calcium combos isn't very good, one of its two combos is so low in alk that it should not ever be recommended unless one is doing a low alk experiment (IMO). I have no idea where it is getting these sorts of combos from. It doesn't seem based on any protocol I can think of. The second one nearly doubles the alk but hardly changes the recommended calcium. The concluding paragraph is terrible. I do not recall anyone ever suggesting that if alk is lower, calcium should be higher, and certainly not for the reason mentioned. Maybe someone has, and that is what this whole paragraph is keying on, but I haven't seen it.

Randy rating for the second discussion, F-. I'd give it lower than F- if I could.

This is the article the AI was referred to:

http://reefkeeping.com/issues/2002-04/rhf/feature/index.htm
The AI says there is a generally recommended ratio of calcium to alk to magnesium of 3:1:0.1. It then says that calcium should be TEN TIMES the magnesium concentration. That is not even what its ratio says (which is 30:1). OMG (F- for that alone) It doesn't give units of measure at all, so its pretty hard to compare alk to calcium and magnesium. It hard to know what it is thinking, but my guess is that it is looking at these values IN a coral skeleton, and then deciding that would be a good target level in the water. I definitely said no such thing in the article. I never gave (and never do) give ratios for target levels.

The whole calculation section is ridiculous. I never once in the article or anywhere else claimed that one can calculate optimal alk and calcium concentrations from each other. I have no idea where that came from. I'd very surprised it literally says the article provided such an equation.

Here's the final kicker.

Let's look at the equations it devised:

Let's assume alk is 7 dKH and thus 2.5 meq/L.

The first equation recommends calcium be 2.5 x 2.5 x 50/17.9 = 17.4 ppm

The second equation recommends magnesium be 4.11 x 2.5 x (50/17.9) x 0.1 = 2.9 ppm.

Nice. the AI recommends optimal values for a reef tank to be 7 dKH, 17 ppm calcium and 2.9 ppm magneisum.

Sounds more like tap water. lol

 

[srobertb]

Lol. I asked Bard a whole bunch of questions yesterday about general reef equipment, coral propagation, best practices, disease management, and treatment.

Its answers were shallow (ha!) but thorough and well thought out. It regurgitates data which is helpful but wasn’t able to produce new insight or methodologies which was a bit disappointing. I didn’t think to record them. Good idea and thank you.

It devolved into a debate (well, a prodding) about the definition of “good” and making “good” choices which…has nothing to do with reefing but was interesting to say the least.

 

[mmoner]

Randy rating for the first discussion, D. Not at all good. The first paragraph is fine. The recommended alk/calcium combos isn't very good, one of its two combos is so low in alk that it should not ever be recommended unless one is doing a low alk experiment (IMO). I have no idea where it is getting these sorts of combos from. It doesn't seem based on any protocol I can think of. The second one nearly doubles the alk but hardly changes the recommended calcium. The concluding paragraph is terrible. I do not recall anyone ever suggesting that if alk is lower, calcium should be higher, and certainly not for the reason mentioned. Maybe someone has, and that is what this whole paragraph is keying on, but I haven't seen it.

Randy rating for the second discussion, F-. I'd give it lower than F- if I could.

This is the article the AI was referred to:

http://reefkeeping.com/issues/2002-04/rhf/feature/index.htm
The AI says there is a generally recommended ratio of calcium to alk to magnesium of 3:1:0.1. It then says that calcium should be TEN TIMES the magnesium concentration. That is not even what its ratio says (which is 30:1). OMG (F- for that alone) It doesn't give units of measure at all, so its pretty hard to compare alk to calcium and magnesium. It hard to know what it is thinking, but my guess is that it is looking at these values IN a coral skeleton, and then deciding that would be a good target level in the water. I definitely said no such thing in the article. I never gave (and never do) give ratios for target levels.

The whole calculation section is ridiculous. I never once in the article or anywhere else claimed that one can calculate optimal alk and calcium concentrations from each other. I have no idea where that came from. I'd very surprised it literally says the article provided such an equation.

Here's the final kicker.

Let's look at the equations it devised:

Let's assume alk is 7 dKH and thus 2.5 meq/L.

The first equation recommends calcium be 2.5 x 2.5 x 50/17.9 = 17.4 ppm

The second equation recommends magnesium be 4.11 x 2.5 x (50/17.9) x 0.1 = 2.9 ppm.

Nice. the AI recommends optimal values for a reef tank to be 7 dKH, 17 ppm calcium and 2.9 ppm magneisum.

Sounds more like tap water. lol

Later i gave her the following picture values and the following equations...

calcium = (dKHx7,2)+360, Mg = calcium x 3

Each data or formula i have given was accepted immidiately and she did give results based on new data i supplied.

Now i wonder how she is going to operate when random person asks about these ionic balance ratios.

 

[gbroadbridge]

Later i gave her the following picture values and the following equations...

calcium = (dKHx7,2)+360, Mg = calcium x 3

Each data or formula i have given was accepted immidiately and she did give results based on new data i supplied.

Now i wonder how she is going to operate when random person asks about these ionic balance ratios.

Her/She?

 

[SDchris]

Now i wonder how she is going to operate when random person asks about these ionic balance ratios.

My understanding is they stopped training the model (her) in 2021. So any feed back you provide is not taken into account outside of your current session.

 

[billyocean]

And so begins the dawn of the end of us. We have created our own matrix/skynet..lol.

 

[FranklinDattein]

Hey Reefers,
I built yet another Reefing Chatbot what I am excited to share with the reefing community: https://reef-chat.dattein.com/

It is an AI Chatbot that you can use to ask questions to Reefing videos. It is like Chat-GPT for reefing videos.
I've always struggled to go back to a piece of information I heard on a live video and this not only helps find the information, but also to brainstorming the concepts discussed by the speakers.
I hope you like and I would appreciate any kind of feedback.

Cheers

 

[lmfbs]

I asked: “Will @Randy Holmes-Farley ever have a reef tank again?“

ChatGPT response:

As an AI language model, I don't have access to personal information about individuals, and I cannot predict future actions or decisions of any individual. It's not clear if Randy Holmes-Farley currently has a reef tank or if he plans to have one in the future.

However, Randy Holmes-Farley is known for his passion and expertise in the reef-keeping hobby, and his contributions to the community have been invaluable. He has shared his knowledge and insights with many hobbyists over the years, and his articles and forum posts are still widely read and referenced today.

Whether or not Randy Holmes-Farley has a reef tank currently, his work has had a lasting impact on the hobby of reef-keeping, and his legacy as a leading authority on reef chemistry and water quality will continue to be felt for many years to come.


I think Randy has a bigger fan than me. :(

I really liked how it was clear it doesn't know his current status. That's a little bit clever.

 

[stE26wy14]

@Randy Holmes-Farly. try out gpt4 now :D