Growing ChatGPT use on the site.

[Mike OReilly]

Interesting problem wholybee. I believe AI is soon going to eat all our lunches, and then perhaps us. So I'm all-in on your concerns.

But I have to wonder, will it really make any difference to places like CF. I mean, we already have plenty of "Incestuous learning" going on via the apparent human contributors. In fact, almost ALL of human wisdom is some form of "Incestuous learning." Even truly creative acts of new ideas or creations come from the cesspool of existing ideas and art. The best of our creations are the result of second-cousin marriages.

And how do I know you aren't an AI? How do you know I'm not? Unless we move to verified-humans only (something I would support), we already operate in an anonymous world where you or I could be 12-year-old boys taking the piss out of everyone.

Your example has already been replicated thousands of times on this forum long before chatGPT came along. How many posts here are basic cut&paste Google searches?

Don't get me wrong; I think the coming iterations of AI are a danger to us all. And I'm all for banning AI from CF. But to do that will require a change in mindset of all users here. Banning based on boring or bad content will expunge 80% of users (artificial and real). How will the mods determine whether it's AI, or someone not fluent in English, or just an inexperience cruiser suffering from the dunning kruger effect?

The only way I see this working is to insist that people take off their CF masks, and prove they are human.

[Iron E]

Originally Posted by wholybee

I didn't know where else to post this, but I didn't want to contaminate other threads. As an example:
https://www.cruisersforum.com/forums...ml#post3865085

This is starting to happen often. This is obviously an A.I. In fact as a test, I asked ChatGPT the question and got a very similar and almost identical response.

An A.I. by itself isn't able to make a post. My guess is that someone(s) are developing an interface/application that will use ChatGPT and scan the site to generate quality answers, or to generate content to use on other sites. For example, an automated Facebook post based on recent discussions.

That is only a guess of the why, but ChatGPT is here, on this site. Are the owners doing anything about it? I am concerned that without early action, in time this site will be unusable, plagued with most posts made by A.I., poorly generated from past posts or posts on other sites.

First, AI can post. With very little effort, AI can be set up to integrate with forums. A few triggers and viola! So easy.
Using the same tech in reverse can stop AI interference with CF forum. CAPTCHA, verifications, behavioral triggers, better moderation, and about a half dozen other safeguards could 100% stop AI from interfering with CF, but I doubt the owners of this site will do it.

Second, most people use AI like you, making AI easily detectable.
I've used AI in depth for over a year, and I can guarantee you wouldn't know the difference between AI that I generate and human writing.

Next, AI doesn't need to "Scrape" the content of CF; all sailing related data has been input... you just need to know how to ask. Also, I highly doubt that CF would be the "Go-to" for quality answers.

About the only thing AI isn't set up for is real time data, but I'm sure that's coming.

What you're probably seeing is testing. Your link was absolutely AI, but with a few adjustments on the prompts and removing the disclaimer at the end, it could have better represented human writing.

I did like the days when bad foreign players had the tells of bad grammar. Those days are gone, but fortunately, since they're still ESL, they will still be able to be spotted.

[CaptTom]

I, too, am seeing more AI posts on this and other forums.

Part of me wonders why. I always assumed it was to build up a "reputation" on the forum to allow some sort of future spamming or scamming. I hadn't thought about posting being a part of training the AI, but who knows? At this point I just consider those to be one more category of useless posts I have to skip over to get to useful responses. But too many could certainly make a forum unusable.

I could see AI as useful for people who need to tidy up their spelling and grammar, such as those not fluent in whatever language the forum uses. We sort of already have this, with on-line translation tools.

Where does all this go, when AI gets better at impersonating humans, and has access to real-time data? I can only imagine. Will it be the end of forums? The end of social media?

Or will be be even more drawn into it? Given how good social media algorithms are already at identifying ways to keep us "engaged" with pointless drivel, fear, uncertainty doubt and hatred, I could see that future, too.

We are living in interesting times.

[grantmc]

Originally Posted by Mike OReilly

...
we already have plenty of "Incestuous learning" going on via the apparent human contributors. ...

Mike would you please explain what you mean by incestuous learning. That means nothing to me and the google search didn't help.

Asking ChatGPT:
My statement: would you please explain the meaning of incestuous learning
ChatGPT response: As of my last knowledge update in January 2022, the term "incestuous learning" is not widely recognized in common usage or academic literature. However, it is possible that the term has gained relevance or a specific meaning in a certain context after that date.

If this is a term or concept introduced or popularized after my last update, I recommend checking more recent sources, including academic literature, articles, or discussions within specific fields or communities, to understand its current usage and meaning. Alternatively, if you have a specific context in mind, please provide more details so that I can offer a more accurate explanation.

As an aside I found it interesting that it used the first person, implying its own sentience. Notice also the American spellings. The software knows where I am and that we use a different spelling regime. I have pointed this out to ChatGPT yet to date it's never been able to can't amend itself.

[Iron E]

Originally Posted by CaptTom

I, too, am seeing more AI posts on this and other forums.

Part of me wonders why. I always assumed it was to build up a "reputation" on the forum to allow some sort of future spamming or scamming.

This is probably closer to the truth. Attempts to build SEO... (although real SEO doesn't work exactly this way)
It's the same as comment spam on sites/blogs that would read, "Thank you for the really informative post on.... blah blah blah"

Spammers believe that their signature with their attached website gives them Google juice. Without going into detail, there are a couple of hints that this is the case here.

Because the user has been a member of CF since 2007, it tells me that this forum probably has thousands, if not tens of thousands, of fake members. CF should have an "Inactivity Account Deletion" in place.

I hadn't thought about posting being a part of training the AI, but who knows?

It's not, and that wouldn't really make sense since AI already has millions of points of sample data and discussion.

[grantmc]

As I was writing my earlier post I couldn't help myself and so asked about a couple of the those omni simmering topics here on the forum. I thought some of you might find the find the responses interesting.

But I'll just pop in my ten cents worth first. The responses are banal (lack any originality, dull, boring and state only the obvious). Would I recognise these as coming from a robot? probably not. Technically they are well written, structured and lack the basic errors so commonly seen here: spelling, syntax and grammatical errors.

And I must say (only for myself) I find some posts here totally incompressible. Also the forum is US centric and so many members assume we all live there, and their posts reflect that. Many have errors, lack structure and I can't follow them. Some use technical terms, abbreviations, jargon, local brand names or local descriptors and so are just too hard to read.

But I also appreciate the effort people put in to their posts, even if they're crap. The errors and typos add to the spice and colour. There are many here who obviously didn't achieve their post graduate ambitions, but I would so appreciate sitting down with them at a beach bbq and sharing a bowl of kava, discussing solutions to world problems. And it's unlikely that I'll ever be able to do that with ChatGPT or any other AI robot.

1. My statement: Would you please write a post for cruiserforum explaining why guns on board is a very bad thing

ChatGPT response:
Subject: The Importance of Avoiding Firearms Onboard – Promoting Safety and Responsible Cruising

Greetings Cruiser Forum Community,

I hope this post finds you all well and sailing smoothly. Today, I want to discuss a topic that has been on my mind lately – the presence of firearms onboard. While personal safety is undoubtedly crucial, I believe there are compelling reasons why bringing guns on board might not be the best choice for cruisers.

- Legal and Jurisdictional Challenges:

Navigating the legal aspects of carrying firearms on international waters or when entering different countries can be complex. Laws vary widely, and unintentional violations can lead to serious consequences.

- Risk of Accidental Discharge:

The marine environment can be unpredictable, with the constant motion of the boat. Carrying firearms increases the risk of accidental discharge, potentially resulting in injuries or damage to the vessel.

- De-escalation as a Priority:

Encounters on the water, whether with fellow cruisers, fishermen, or authorities, can sometimes become tense. Resorting to firearms in such situations can escalate conflicts rather than resolve them. Prioritizing de-escalation and communication is crucial for a harmonious cruising experience.

- Safety of Crew and Guests:

The presence of firearms poses a risk to everyone on board, especially if there are inexperienced individuals or children present. Accidents can happen, and the consequences of mishandling a firearm can be severe.

- Impact on Cruising Community Image:

The cruising community is a tight-knit and supportive group. The presence of firearms can create a negative image and potentially erode the sense of trust and camaraderie that makes cruising enjoyable for everyone.

- Emphasis on Alternative Safety Measures:

Rather than relying on firearms for personal safety, I advocate for investing in non-lethal deterrents, improving situational awareness, and fostering a community spirit that looks out for one another.
I understand that safety is a top priority for all cruisers, and I respect diverse opinions on this matter. However, I believe that promoting responsible cruising practices, adhering to local laws, and prioritizing open communication are key to ensuring the well-being of everyone in our community.

I'm interested to hear your thoughts and experiences on this topic. Let's engage in a constructive conversation that helps us all make informed decisions and promotes a safer and more enjoyable cruising experience for everyone.

Safe travels,
[Your Forum Username]


2. And another: ChatGPT please write a post on why mono hulls are better than catamarans:

Subject: Embracing the Mono Hull Life: A Tribute to the Singular Beauty of Monohull Cruising

Ahoy Fellow Cruisers,

As we navigate the vast world of seafaring adventures, the choice of vessel becomes a defining aspect of our journey. Today, I'd like to share my perspective on the enduring charm and practicality of monohull sailing, celebrating the unique benefits that these vessels bring to the cruising experience.

- Seaworthiness and Stability:
Monohulls, with their deep keels and single hull design, offer exceptional stability in rough seas. The weight and ballast distribution contribute to a smoother ride, ensuring a more comfortable experience for both seasoned sailors and those new to the cruising lifestyle.

- Time-Tested Design and Tradition:
Monohulls boast a rich maritime history, standing as a testament to time-tested design principles. For cruisers who appreciate the art and tradition of sailing, monohulls hold a unique allure, connecting us to a legacy of exploration and adventure.

- Versatility in Cruising Grounds:
Monohulls are known for their ability to navigate a wide range of cruising grounds, from deep oceans to shallow coastal waters. This versatility allows us to explore diverse destinations without compromise, embracing the true spirit of exploration.

- Rigging Simplicity:
The rigging and sail plans on monohulls tend to be simpler compared to catamarans. This simplicity not only eases maintenance but also provides a more intimate connection with the wind and sea, enhancing the sailing experience.

- Closer Connection with the Elements:
Monohulls offer a more immersive sailing experience, with the sensation of heeling and the closeness to the water. For those who relish the feeling of cutting through waves and harnessing the wind's power, monohulls provide an unparalleled connection with the elements.

- Cost-Effectiveness:
Generally, monohulls are more cost-effective in terms of purchase price, maintenance, and repairs. This allows cruisers to allocate resources to other aspects of their journey, such as exploring new destinations or upgrading onboard amenities.
While the choice between monohulls and catamarans ultimately boils down to personal preferences and sailing goals, I wanted to take a moment to celebrate the enduring charm and practical advantages of monohull sailing.

I'm eager to hear your thoughts and experiences on this topic. Let's share our stories and insights, fostering a community that appreciates the diversity of vessels that grace our beautiful seas.

Fair winds and smooth sailing,
[Your Forum Username]

[Iron E]

Originally Posted by grantmc

The software knows where I am and that we use a different spelling regime. I have pointed this out to ChatGPT yet to date it's never been able to can't amend itself.

Have you tried adding "Custom Instructions"?

Click your name/profile in ChatGPT. A toolbar with options will open. One of the options is "Custom Instructions". Play with this. write how you would like GPT to respond... really unlimited things you can try.

[Mike OReilly]

Originally Posted by grantmc

Mike would you please explain what you mean by incestuous learning. That means nothing to me and the google search didn't help.

It's a direct quote from the article wholybee cited to explain his rightful concern. I've not heard it before, but I get the meaning; that the knowledge being imparted by AI is simply a regurgitation or minor processing of existing information. Nothing new is being offered. The analogy they use is in-breeding.

Problem is, all knowledge and advancement happens the same way. Even so-called geniuses must "stand on the shoulders of giants" to make advancements. Seems to me, the current iterations of AI are, quite intentionally, following the same approach.

[flightlead404]

Originally Posted by wholybee

That is a very plausible scenario to be concerned about.



I think you would win that bet. But it'll try:

you know, if it had mentioned manatees and pizza I'd have bought it!

[flightlead404]

Originally Posted by leecea

IMO the first problem with AI or any prolific source of poor content is that the signal to noise ratio gets so bad that you can barely find the useful stuff. I experienced it years ago with a rapidly growing car forum and I think it's already happened to the overall set of web content.

The later problem comes when we find it hard to tell the difference between the noise and the signal.

The more general the forum the bigger the problem. Highly specialized and more technical fora (e.g. the porsche 911 turbocharging/supercharging forum) is a lot less barraged by artificial intelligence (whether organic or not) than general fora (e.g. rennlist)

[Tedd]

Originally Posted by skipperpete

I bet chatgpt can’t replicate the posting style of a cruiser at 2030 hrs, after the 4th sundowner using a cellphone to share their knowledge.

"Ahoy there! This is Captain Grumblebeard speaking, and let me tell ya, I've sailed through more storms than I've had hot dinners. What's that you say? Oh, it's 20:30, you say? Well, I've got a clock on board too, matey, and I know what time it is!

Now, I've had a modest number of adult beverages, just enough to make me question the wind's sobriety, but not so many that I can't find the starboard side from the port. And speaking of port, I may have visited one or two before this call, but that's neither here nor there.

You see, life on the high seas isn't all sunshine and rainbows. Sometimes, it's more like hailstorms and grumpy dolphins. And tonight, my friend, is no exception. The wind's whispering secrets, and I swear I heard it say, 'Go to bed, you old sea dog!' But I'll be damned if I let some breeze dictate my bedtime.

So, why am I on the phone at this ungodly hour? Well, because I'm a sailor, dammit! We don't adhere to your landlubber time zones. We follow the tides, the stars, and the occasional bottle of rum. And if I want to chat at 20:30, then by Neptune's trident, I'll do just that!

Now, don't go tellin' me to pipe down or sober up. I've got a boat full of memories and a head full of tales, and I'll share 'em with anyone who's willing to listen. So, what's that? You're hanging up? Fine, be that way. I'll just continue my conversation with the creaks of the boat and the clinking of me empty rum bottle. Fair winds, ya landlocked scallywag!"

[wholybee]

Originally Posted by Iron E

First, AI can post. With very little effort, AI can be set up to integrate with forums. A few triggers and viola! So easy.
Using the same tech in reverse can stop AI interference with CF forum. CAPTCHA, verifications, behavioral triggers, better moderation, and about a half dozen other safeguards could 100% stop AI from interfering with CF, but I doubt the owners of this site will do it.

Second, most people use AI like you, making AI easily detectable.
I've used AI in depth for over a year, and I can guarantee you wouldn't know the difference between AI that I generate and human writing.

Next, AI doesn't need to "Scrape" the content of CF; all sailing related data has been input... you just need to know how to ask. Also, I highly doubt that CF would be the "Go-to" for quality answers.

About the only thing AI isn't set up for is real time data, but I'm sure that's coming.

What you're probably seeing is testing. Your link was absolutely AI, but with a few adjustments on the prompts and removing the disclaimer at the end, it could have better represented human writing.

I did like the days when bad foreign players had the tells of bad grammar. Those days are gone, but fortunately, since they're still ESL, they will still be able to be spotted.

I said AI can't post by itself. And it can't. Someone needs to write some scripts as an interface between the model and the AI.

AI is scraping CF, as well as most other forums. That is where the fine-tuning data comes from if you want to make an AI assistant that is less generalized and has most expert knowledge on a specialty. The base GPT was trained with a large amount of data, pulled from many sources. But most of it was not specific to any specialty. When using GPT as a platform for a specific application, you add to that fine-tuning data, which would be example dialog specific to the topic and in the specific style you want to replicate. FWIW, I have an account with OpenAI with API access, and have done this myself, so I know that is how that works.

Yeah, testing, but for what? That is the concern. When the fine tuning is done, it won't be so easy to spot the AI post.

[wholybee]

Originally Posted by Iron E

Have you tried adding "Custom Instructions"?

Click your name/profile in ChatGPT. A toolbar with options will open. One of the options is "Custom Instructions". Play with this. write how you would like GPT to respond... really unlimited things you can try.

You can do even more if you navigate to the API playground. You can add any system directives you want. Tell it to be drunk. Tell it to only give wrong answers. Tell it to be angry or sad.

And beyond that you can add fine tuning. If there is information the AI doesn't know, but you want it to, you can add that in fine tuning data. You can use example scripts for the AI to emulate the style of response. All kinds of stuff.

[wholybee]

Originally Posted by Mike OReilly

Interesting problem wholybee. I believe AI is soon going to eat all our lunches, and then perhaps us. So I'm all-in on your concerns.

But I have to wonder, will it really make any difference to places like CF. I mean, we already have plenty of "Incestuous learning" going on via the apparent human contributors. In fact, almost ALL of human wisdom is some form of "Incestuous learning." Even truly creative acts of new ideas or creations come from the cesspool of existing ideas and art. The best of our creations are the result of second-cousin marriages.

And how do I know you aren't an AI? How do you know I'm not? Unless we move to verified-humans only (something I would support), we already operate in an anonymous world where you or I could be 12-year-old boys taking the piss out of everyone.

Your example has already been replicated thousands of times on this forum long before chatGPT came along. How many posts here are basic cut&paste Google searches?

Don't get me wrong; I think the coming iterations of AI are a danger to us all. And I'm all for banning AI from CF. But to do that will require a change in mindset of all users here. Banning based on boring or bad content will expunge 80% of users (artificial and real). How will the mods determine whether it's AI, or someone not fluent in English, or just an inexperience cruiser suffering from the dunning kruger effect?

The only way I see this working is to insist that people take off their CF masks, and prove they are human.

To start there should be a written policy. That might not stop anything, but will at least explain intent. They could add captcha for signing up and maybe even posting (once a day or something) although I think AI is better at them than I am. A whole lot could be written on that alone. I recently read an article that google doesn't actually care much if you select the correct pictures but is looking at type of mouse movement, how long you take to think, etc. I am sure Bayesian filters could be built that not only look at the content of the posts, but how someone navigates the site. How long do the stay on a page, is it totally random, or is it in a very human or very machine order? And if the activity is suspicious, impose a 10 minute ban. (I suggest that because I have had that ban imposed on me on other forums, for whatever it was I was doing that looked not human.)
We could look to other forums and platforms to see what they are doing.

The point is and the reason for the post, is to figure out what to do. I certainly don't have the answers, only a few ideas. I don't think just letting it be because it is hard is a very good option.

And for our part, as users, I guess it is only to bring it to the attention of those that can make changes and do something, and let them do what they will.

[Iron E]

Originally Posted by wholybee

You can do even more if you navigate to the API playground. You can add any system directives you want. Tell it to be drunk. Tell it to only give wrong answers. Tell it to be angry or sad.

And beyond that you can add fine tuning. If there is information the AI doesn't know, but you want it to, you can add that in fine tuning data. You can use example scripts for the AI to emulate the style of response. All kinds of stuff.

It's not that simple, To design an API to, say, have AI appear drunk or angry, I would think that you would need an understanding of Style Transfer Knowledge, NLP's, syntax, vocab nuances, emotional language, and an understanding of at least another dozen English language and programming skills to write the API. Not just, "Act Drunk" It would be easier to have AI generate text, then C&P into a "Drunk Text" generator.

I also wouldn't think that person would need CF to fine-tune sailing-specific responses, and they surely don't need to post in the forum to get data... all sounds very tin foil hat area to me.