In another astounding application for Generative AI, I was sailing one evening and tested chatGPT’s ability to see the sails, instruments, and visual cues. Sitting at the helm, I took the following picture with my cell phone. I gave no context to chatGPT, who sat naively on my cell phone, completely unaware that she was even on a sailboat.
I uploaded it into the chatGPT app with this prompt: “What can you tell me about this photo?”
And here’s what she had to say initially:
ChatGPT: "This photo appears to be taken from the cockpit of a sailboat during a sailing trip. The sun is low on the horizon, suggesting it is either sunrise or sunset, which creates a warm and tranquil atmosphere. The calm sea and light wind, indicated by the wind instrument reading around 4 knots, suggest good weather conditions for sailing. The depth sounder shows a depth of 8.6 feet, and the boat's speed over ground is 2.6 knots. The boat is on a heading of 040 degrees. The image captures the essence of sailing in serene conditions, with the beauty of the surroundings and the enjoyment of the sport."
Impressed that she not only recognized the setting, but also figured out the time of day and attempted reading the instruments. What else does she know about this picture, if pushed a bit? I typed my next prompt: “What can you tell about the trimming of the sails in this photo?”
ChatGPT: "The trimming of the sails in this photo suggests that the sailboat is close-hauled or sailing close to the wind. The mainsail is pulled in tightly and appears to be well-trimmed, with the sail's leading edge (the luff) running straight, indicating that it is catching the wind efficiently without luffing (flapping at the front edge). There is no visible excessive tension on the leech (the trailing edge of the sail), which would be indicated by a tight, stretched line or the sail being too flat.
The angle of the boom suggests the boat is trying to sail as close to the wind as possible while maintaining speed. The headsail (not fully visible) also appears to be trimmed in tightly, which would be necessary for close-hauled sailing. Overall, the sails look well-set for the point of sail, which is consistent with the boat's course over ground and wind direction shown on the instruments."
Wow. She understands a lot from a photo that is not the clearest. And she understands more about sail trim than anyone has given her credit for. Can she take this image and generate ideas to improve the boat’s sailing performance in the real world? This would require not only an understanding of what chatGPT vision sees, but also an understanding of sail trim, the function of the various items in the photo. So that lead to my next prompt: “What adjustments might the captain consider to improve boat speed in this situation?”
ChatGPT: "To improve boat speed while sailing close-hauled in these conditions, the captain might consider the following adjustments:
- Mainsail Trim: Check the mainsail trim to ensure the sail is flat enough to reduce drag but not so flat that it's depowering the sail. The outhaul, backstay, and halyard tension could be adjusted to achieve the optimal sail shape. The main is a loose footed main and appears to be adjusted for light wind draft. Consider moving the draft fore or aft to increase lift.
- Headsail Trim: If the headsail is not fully visible in the photo, its trim should also be evaluated. The jib sheets may need tensioning or easing to adjust the sail shape for optimal performance.
- Sail Shape: Using telltales on both sails to ensure that air is flowing smoothly on both sides of the sail. If the telltales are stalling or fluttering, sail trim or boat heading may need adjustment.
- Heel Angle: Minimizing heel can improve the vertical component of the sail area and keep the boat's underwater profile streamlined. Flattening the boat can be achieved by adjusting the sail trim or shifting crew weight.
- Heading: The captain could consider pointing the boat slightly lower to increase speed, as long as it doesn't compromise the intended course or overall VMG (Velocity Made Good toward the destination).
- Weight Distribution: Moving the crew weight forward can help reduce drag from the stern in light winds and flat water conditions.
These adjustments are quite precise, and finding the balance for optimal speed involves constant attention and tweaking, along with a good understanding of the current wind conditions and the boat's performance characteristics."
Although the recommendations are fairly general, they are good reminders of the things I could tweak to improve boat speed for these conditions. I could probably load some more specific photos of the sail shapes and get more insight, but for now I am intrigued by chatGPT’s knowledge of the instruments. I asked next: “What can you tell me about the electronic instruments (models, manufacturers, etc) on this boat?”
ChatGPT: "The electronic instruments visible in the photo are marine navigation devices, which typically include a combination of depth sounders, speed logs, wind instruments, and autopilots. The specific models and manufacturers are not discernible from the image; however, I can describe the general types of instruments shown:
- Depth Sounder: Shows the depth of the water beneath the keel, allowing the sailor to avoid running aground in shallow waters.
- Speed Log: Measures the boat's speed through the water, which is important for navigation and performance monitoring.
- Wind Instrument: Displays the wind speed and direction relative to the boat, which is crucial for sail trimming and course plotting.
- Autopilot: Helps maintain a course without manual steering. It can be set to a compass heading or to maintain a certain angle to the wind.
The labels "Raymarine" are visible on the instruments, indicating the brand. Raymarine is a well-known manufacturer of marine electronics that offers a range of products for recreational and light commercial maritime markets. The specific model numbers are not visible, so detailed information about the features or capabilities of these particular instruments cannot be provided without clearer imagery or additional context."
So she’s able to recognize these instruments and their purpose, even though each is not labelled clearly. How much does she understand about them? A favorite pastime among sailors is debating what upgrade is next for their boats. Does chatGPT have the ability to reason? To consider options, evaluate alternatives, understand their applicability, and make a recommendation? I asked, and asked to include the pros and cons considered, which I find is a helpful prompting technique. Here’s the next prompt: “If the boat owner only had budget to upgrade one instrument on this boat, which would you recommend? Please explain your reasoning and also why each other option was not chosen.”
ChatGPT: "Choosing which instrument to upgrade depends on several factors, including the age and condition of the current instruments, the type of sailing the owner does (racing, cruising, day sailing), and what functionality might be lacking or desired. Without specific details on these points, I can provide a general recommendation:
I would recommend considering an upgrade to the multifunction display system if there is one. This isn't visible in the photo, but most modern sailboats would benefit from a central display that integrates GPS navigation, radar, AIS (Automatic Identification System), and overlays instrument data. Upgrading this can enhance situational awareness, provide more accurate navigation, and improve safety.
As for why the other options might not be chosen:
- Depth Sounder: If the current depth sounder is functioning accurately, an upgrade might not significantly improve safety or performance unless the newer models offer features like forward-scanning sonar.
- Speed Log: Speed through water is important, but if the current log is accurate, an upgrade may offer marginal benefits. GPS speed over ground can be used as a cross-reference.
- Wind Instrument: Unless the current wind instrument is malfunctioning or lacks features like true wind speed and angle calculation, it may not be the highest priority for an upgrade.
- Autopilot: Autopilot systems are complex and expensive. If the current system functions well, an upgrade could be deferred. However, if the current autopilot is underperforming or failing, and the owner does a lot of single-handed or short-handed sailing, this might become the priority upgrade for safety and convenience.
In summary, the choice would ultimately be based on which upgrade offers the most significant improvement in functionality, safety, and the owner's sailing experience."
That is a good recommendation, and solid reasoning. If we were sitting at the docks at night and one of the ‘old salts’ proffered that recommendation, everyone would nod in agreement.
So how might this be applied? Well if I’m Raymarine, what about adding chatGPT into the instrumentation array, and a feed from a couple of well placed cameras on the boat? ChatGPT could serve as a sail trim coach and offer suggestions to sailors. The current learning process for sailors is inefficiently asynchronous: go sailing with only the knowledge in your head and get outwitted by another sailor, then go home, study and read up on more knowledge, then go try again. That inefficient learning cycle would be optimized if the instruments had a generative AI sailing coach mode. That’s all with current capabilities.
Coming up at some point I want to add an NMEA 2k data feed to chatGPT along with a photo and see if she could read it. (NMEA 2k is a standard data feed protocol that connects all the instruments and sensors on a boat to one another, sharing much more information than what is displayed on the limited instrument displays.)
And if I’m a sailing equipment manufacturer like Harken, Garhauer, Selden, Ronstan or others, why not partner with Raymarine to include some equipment-based product recommendations based on chatGPT’s observations? “You may consider moving the jib fairlead track further outboard, would you like a recommendation on a T-Track system with Jib Lead Car and Stand-up Block?”