Using AI (Artificial intelligence) for dive planning

[Wibble]

Having a play with AI by asking ChatGPT via Siri on an iPhone for some answers.

Q: "What Gradient Factors to use for a 60m technical trimix dive"

It responded with:

"For a 60-meter technical trimix dive, selecting appropriate gradient factors is crucial for safe decompression. Gradient factors determine the maximum allowed supersaturation in tissue compartments during ascent, affecting decompression time and risk. A common starting point is a 30/85 gradient factor, meaning 30% of the maximum supersaturation at depth and 85% at the surface. However, individual factors like dive duration, gas mix, and personal physiology should be considered. Consulting with a diving professional and using dive planning software is recommended to tailor the gradient factors to your specific dive profile."​

Interesting answer; well written in terms of grammar and spelling, but rather old-school information as the current thinking for several years (e.g. Dr Simon Mitchel et al) would be around 50/75.

The way it's written appears very authoritative which could be challenging if people get used to asking it and getting wrong answers without any ummms and errmms.

Wondering what other questions would get, lets say, interesting answers from this toy?


Edit: edited the thread title.

 

[Wibble]

Q: "What is the best rebreather to buy"

Answer:

"The best rebreather to buy depends on your specific needs and experience level, but popular options among divers include the AP Inspiration, the JJ-CCR, and the Hollis Prism 2. These models offer reliability, good performance, and safety features. It's essential to consider factors such as your diving experience, budget, and the kind of diving you plan to do before making a choice. Additionally, ensure you receive proper training on the specific rebreather you choose."​

Definitely a diplomatic answer!

 

[BoltSnap]

Q: "What is the best rebreather to buy"

Answer:

"The best rebreather to buy depends on your specific needs and experience level, but popular options among divers include the AP Inspiration, the JJ-CCR, and the Hollis Prism 2. These models offer reliability, good performance, and safety features. It's essential to consider factors such as your diving experience, budget, and the kind of diving you plan to do before making a choice. Additionally, ensure you receive proper training on the specific rebreather you choose."​

Definitely a diplomatic answer!

Disregarding the "diplomacy," what is your assessment of the validity and suitability of this answer?

 

[Wibble]

You're right, it's quite a good answer, probably not far off of the standard answer to that question here on SB -- which it's probably "learned" from.

The JJ is definitely the easy one to recommend. The Inspiration was the most popular rebreather made as the early adopter. Hollis Prism is quite rare in the UK.

 

[atestani]

For what it's worth...
Here is what DeepSeek (the Chinese AI tool that got a lot the press recently) said.:
(I'm not a tech diver and am not into this at all but thought the level of detail in the response was notable. Whether it is right or wrong is a whole other question!)

"What Gradient Factors to use for a 60m technical trimix dive?"

When planning a 60-meter technical trimix dive, selecting appropriate gradient factors (GF) is crucial for managing decompression stress and minimizing the risk of decompression sickness (DCS). Gradient factors are used in conjunction with Bühlmann decompression algorithms to adjust the conservatism of the dive profile.

Key Considerations:​

1. Depth and Gas Mix: At 60 meters, you're likely using a trimix (e.g., 18/35 or similar) to manage narcosis and oxygen toxicity. The choice of gradient factors will depend on your risk tolerance, dive conditions, and personal physiology.
2. Gradient Factors (GF): Gradient factors are typically expressed as two numbers (e.g., GF 30/70 or GF 40/80):
   
   • GF Low (e.g., 30 or 40): Controls the initial ascent and how close you get to the first decompression stop.
   • GF High (e.g., 70 or 80): Controls the final ascent and how close you get to the surface.

Common Gradient Factor Settings:​

• Conservative Profile (e.g., GF 30/70): Suitable for cold water, strenuous dives, or if you have a history of DCS.
• Moderate Profile (e.g., GF 40/80): A balanced approach for most technical dives.
• Aggressive Profile (e.g., GF 50/85): For experienced divers in ideal conditions, but increases DCS risk.

Recommended Gradient Factors for a 60m Trimix Dive:​

• GF 30/70: A conservative choice, especially for longer dives, colder water, or if you're less experienced.
• GF 40/80: A good middle ground for most divers, balancing safety and decompression time.

Additional Tips:​

• Plan with Software: Use dive planning software (e.g., Subsurface, MultiDeco, or Shearwater Desktop) to model your dive with your chosen gradient factors.
• Monitor Real-Time GF: If using a dive computer with real-time GF tracking (e.g., Shearwater), adjust your ascent to stay within your planned GF limits.
• Gas Switching: Ensure proper gas switching to higher oxygen content during decompression to accelerate off-gassing.
• Safety Stops: Always include deep stops (if planned) and safety stops to further reduce DCS risk.

Final Note:​

Gradient factors are not one-size-fits-all. Your choice should reflect your experience, dive conditions, and personal risk tolerance. Always dive within your training and certification limits.

 

[Wibble]

Interesting

Again, the current thinking is to use a GF low around 50, so not using the old deeper stops, and GF high around 70.

The tips and notes look good though.

What's most interesting is how persuasive the wording is in these AI responses.

 

[therookie]

I decided to have some fun with this recently too. My "prompt" (the technical term for the question you ask the AI) was this:

I am a scuba diver with a PADI AOW certification. I am planning minimum gas requirements for my upcoming dive trip. Assuming my SAC rate is 14 liters per minute and I am diving with an AL80 cylinder, tell me what my minimum gas requirements to bring me and a buddy up to the surface safely with a maximum ascent rate of nine meters per minute and a three-minute safety stop at the following, from the following depths, 10 meters, 20 meters, and 30 meters. What should the reading on my SPG in imperial units be rounded to values that I can easily read on the gauge?

(The question at the end is because I like metric for calculations but the rented SPGs I use are all Imperial)

I won't give the full response, but it began with this (showing how well it was able to parse my not-very-well-worded prompt):

To calculate the minimum gas requirements, we will factor in:

1. Surface Air Consumption (SAC) rate: 14 liters per minute.
2. AL80 cylinder: A standard aluminum 80 cubic foot cylinder holds 11.1 liters of air at 200 bar (or ~3000 psi).
3. Two divers: You and a buddy will be sharing gas.
4. Ascent rate: 9 meters per minute.
5. Three-minute safety stop at 5 meters.
6. Depths: 10m, 20m, and 30m.
7. Convert values into imperial (psi) units and round to readable values.

And then it proceeded to show the step by step calculations and came up with numbers that were, in my opinion based on my own calculations, too low. So I asked it to double the SAC rate (really what I should have done in the first prompt), and got numbers I was more comfortable with. Then I asked it to account for inaccuracy of the SPG, so it padded with an extra 100psi (AI decided on the number to add). And then I asked it to remove the safety stop from the calculation (because in reality there is a good chance I would just skip it, depending on our mental state at the time and how long we were at depth...). At the end of the chat I had minimum gas requirements that were more conservative than my own, with and without the safety stop.

It was a fun exercise in AI prompting. I'm still going to use my own numbers, though. AI is a super powerful tool when used correctly, and super dangerous if you just take the results at face value without examining them carefully - not just in the context of scuba diving, but in everything I have used it for.

 

[Trilobite]

Edit: edited the thread title.

Needs more editing for spelling.

 

[Wibble]

...AI is a super powerful tool when used correctly, and super dangerous if you just take the results at face value without examining them carefully - not just in the context of scuba diving, but in everything I have used it for.

Being a bit on the old side, I originally studied electronics where we used log tables and a slide rule. Scientific calculators were very expensive and quirky (e.g. HP reverse Polish notation).

Anyway, the big thing about using slide rules & tables was doing the approximation test to get the magnitude of the result and to roughly validate that the answer to be within the expected range.

This is probably the same challenge with AI. The answers are well written, but you cannot guarantee their validity unless you do the approximation and have an understanding of the subject.

Can see some horrid disasters ahead if people unconditionally rely upon AI.

TBH this is like all dive planning. When you've done enough dives you know what the expected profiles are going to be. We also trust our (Shearwater) dive computerS to know that they get you out unbent, physiological quirks aside. We also know that MultiDeco et al give good dive plans which are close enough to the real dives.

 

[BoltSnap]

Scientific calculators were very expensive and quirky (e.g. HP reverse Polish notation).

They're the best! I still have a couple of them that I use regularly.