Growing ChatGPT use on the site.

[CaptTom]

That article was pure speculation.

The CPU in the computer I'm typing this on right now is using 3-5% of its capacity. The rest of the time it's running an "idle" loop somewhere deep in the code.

It doesn't seem to care. It doesn't seem to be bored.

I would imagine that "boredom" would require some sense of urgency to accomplish something. Aren't we anthropomorphizing to suggest that a computer will have the same set of functions as our brains?

A better point would be that we don't really know what sort of behaviors might emerge from these complex systems we're designing. For all we know, it could be much worse than we're currently imagining. Or AI might be just another tool humanity learns to apply to solve our problems.

[TrentePieds]

grantmc said: "Notice also the American spellings. The software knows where I am and that we use a different spelling regime."

Indeed!

It is clear that AI "speaks" with an American accent. Surprise, surprise! Switching between American and Imperial/Commonwealth orthographies is easy enough. Changing "accent" is far more difficult. If you see a split infinitive in a post using C/I orthography, you are justified in suspecting AI. :-).

When a post answering a query is longer than a lucid, focused answer needs to be, and longer than would "normally" be provided by a human American, one may reasonably expect it to be AI generated. This is particularly so when such unnecessarily long replies consist of "items" neatly ordered in "on the one hand this, on the other hand that" fashion.

"Give aways" that a text is NOT AI generated are such things as the (correct) use of vocabulary from languages OTHER than American/English, and allusions to historical events, to literature and to intellectual constructs that are not part of the American "algemeine Bildung". Obviously, such allusions can be deliberately inserted in a text to "signal" that it is NOT AI generated.


TrentePieds

[Shanachie]

I think the big danger of AI right now is not that it will turn evil or outthink us, but that it can be easily used as it is right now to control the flow of information.


For CF, imagine that one of the anchor companies creates a bot designed to establish that its particular anchor is the best in the world.


Every time an anchor question gets posted, the bot swings into action and posts various comments that proclaim the superiority of that particular anchor.


Maybe a few real-life CF members recount their problems with the anchor dragging in various conditions, but those posts are overwhelmed by those created by the bot.



CF members read the posts, say wow, that's a great anchor. I need to buy one. Mission accomplished.



On a larger scale, chatGPT could be used to rewrite history by programming it to present particular points of view.



A big danger is when it provides a summary of information it found (or made up ...) without adequately sourcing the information with links that allow you to judge the veracity of the summary.



You can already see a crude version of this in the various "echo chambers" on the Internet.


Radical points of view on all sides get normalized because everyone on a particular web site agrees with each other. Look at all of the crazy stuff being written about Taylor Swift right now.


All of this makes me think it's not always bad to be old. I won't be around to see the worst of it.

[s/v Jedi]

Originally Posted by Shanachie

I think the big danger of AI right now is not that it will turn evil or outthink us, but that it can be easily used as it is right now to control the flow of information.

For CF, imagine that one of the anchor companies creates a bot designed to establish that its particular anchor is the best in the world.

Every time an anchor question gets posted, the bot swings into action and posts various comments that proclaim the superiority of that particular anchor.

Maybe a few real-life CF members recount their problems with the anchor dragging in various conditions, but those posts are overwhelmed by those created by the bot.

CF members read the posts, say wow, that's a great anchor. I need to buy one. Mission accomplished.

On a larger scale, chatGPT could be used to rewrite history by programming it to present particular points of view.

A big danger is when it provides a summary of information it found (or made up ...) without adequately sourcing the information with links that allow you to judge the veracity of the summary.

You can already see a crude version of this in the various "echo chambers" on the Internet.

Radical points of view on all sides get normalized because everyone on a particular web site agrees with each other. Look at all of the crazy stuff being written about Taylor Swift right now.

All of this makes me think it's not always bad to be old. I won't be around to see the worst of it.

It’s much worse. The AI will create dozens of accounts, hundreds if need be and the forum will change into an endless commercial hidden advertising stream, going the same way as TV, down the drain into oblivion.

[BAD ORCA]

Originally Posted by Shanachie

I see where your sense of alarm is coming from.



The terrible thing would be AI chatbots coded to respond to a discussion on, say, electronics with posts that push for a particular brand, regardless of its quality.


Or, if a discussion is critical of a particular product, bombarding the thread with posts to drown out the legitimate complaint of the OP.


On the positive side, AI chatbots might kill the Internet. Then the world would surely be a better place. Sailors would have to go to the local marina bar to talk shop.

Thats exactly what i thought. My first impression was that it was being trained as a subtle advertising tool to recomend products in a way that seems authentic. Internet advertising 3.0?

[Iron E]

Originally Posted by s/v Jedi

It’s much worse. The AI will create dozens of accounts, hundreds if need be and the forum will change into an endless commercial hidden advertising stream, going the same way as TV, down the drain into oblivion.

Thousands (if not tens of thousands) of accounts on popular forums are actually spam accounts. They have been there for many years attempting to influence forums or algorithms.

Forum members know when they see it and it usually isn't effective... (though here in CF I see forum members and even Mods interact with obvious AI)
E2L is a possible answer for a few of the posts, and I understand wanting to create a friendly environment, but erring on the side of caution is best practice when creating a "Safe" environment.

This entire thread of warnings to CF really doesn't make a lot of sense to me. Mostly because CF is a very particular niche that doesn't really make for the monetary gains you would want from the expense and time of designing an AI.
"Let's design and refine an AI that sounds like a real person so we can target dozens of people that will buy an anchor" doesn't sound like a promising business model... but what do I know, I also think that people that see lights in the night sky and think they're flying saucers are wrong because I don't think that a technology that can transverse space would need headlights.

[letsgetsailing3]

I think the OP was right to call that thread out as obvious AI, and to raise it to our attention.


Probably better not to interact with posts like that.

[GordMay]

Originally Posted by Shanachie

...
For CF, imagine that one of the anchor companies creates a bot designed to establish that its particular anchor is the best in the world.
Every time an anchor question gets posted, the bot swings into action and posts various comments that proclaim the superiority of that particular anchor ...

Like Peter Smith [Rocna] did [repeatedly], on Wikipedia's anchor entry, years ago.

[flightlead404]

Originally Posted by skipperpete

My iphone has a Chatgpt keyboard in settings that can proofread my emails, translate my emails into any language and even change the “tone” of a message.
i dont u se itt tho

So we see

[flightlead404]

Originally Posted by s/v Jedi

AI’s benefit from the forum by absorbing all it’s content to gain knowledge.

The forum benefits from AI because it increases revenue, with this thread being the ultimate form of that

I wonder if an AI can be entertaining itself?

Its not unconcerning.

If the AI learns by training itself on the forum

and then is actively participating in the forum, or another AI is

Are they then training themselves based on their own output

Is it a race to the bottom?

[SV Confianza]

Although some people are mentioning that ChatGPT is scraping cruiserforum, it is not. ChatGPT is blocked by cruiserforum robots.txt. Ask ChatGPT if it can find something on this forum if you would like to confirm.

Sure, someone can make another LLM that won’t abide by the rules, but for now it seems that AI posts on here are generated by ChatGPT.

Or maybe I am just a chatgpt bot throwing people off the trail =).

[wholybee]

Originally Posted by SV Confianza

Although some people are mentioning that ChatGPT is scraping cruiserforum, it is not. ChatGPT is blocked by cruiserforum robots.txt. Ask ChatGPT if it can find something on this forum if you would like to confirm.

Sure, someone can make another LLM that won’t abide by the rules, but for now it seems that AI posts on here are generated by ChatGPT.

Or maybe I am just a chatgpt bot throwing people off the trail =).

robots.txt doesn't block anything. It is up to the client to retrieve that file and honor it or not. It was a great idea in the 90's, but is mostly ignored. I think google honors it.

ChatGPT was trained on data that ends January 2022 and most likely does not include Cruisers Forum, and so if you ask ChatGPT about Cruisers Forum it won't include anything here. HOWEVER, developers using ChatGPT can still scrape Cruisers Forum for fine tuning data, or can script ChatGPT to read and post here for additional training. None of that fine tuning would be available to us or anyone that askes ChatGPT anything, it would only be available to the developer performing that training.

[GordMay]

Elon Musk's Neuralink robotic brain implant

The first human patient received an implant, from Elon Musk's computer-brain interface company, Neuralink, over the weekend.

In a post Monday, on X [Twitter], Musk said that the patient received the implant the day prior, and was "recovering well." He added that "initial results show promising neuron spike detection."

Spikes are activity by neurons, which the National Institute of Health describes as cells that use electrical and chemical signals to send information around the brain and to the body.

Neuralink's device is about the size of a large coin, and is designed to be implanted [by a robot] in the skull, with ultra-thin wires, going directly into the brain, in a region of the brain, that controls movement intention.

The initial goal, of the brain computer interface [called "Telepathy"], is to give people the ability to control a computer cursor or keyboard, using their thoughts alone.

The startup's PRIME Study is a trial for its wireless brain-computer interface to evaluate the safety of the implant and surgical robot.

➥ https://twitter.com/neuralink

PRIME Study ➥ https://neuralink.com/pdfs/PRIME-Study-Brochure.pdf

How Neuralink might work for people with paralysis
Video 3:53 ➥ http://www.cbc.ca/player/play/2304995395577

Neuralink brain chip's first human patient. How does it work?
Video 12:53 ➥ http://www.cbc.ca/player/play/2305683011690

[Iron E]

Originally Posted by wholybee

robots.txt doesn't block anything. It is up to the client to retrieve that file and honor it or not. It was a great idea in the 90's, but is mostly ignored. I think google honors it.

ChatGPT was trained on data that ends January 2022 and most likely does not include Cruisers Forum, and so if you ask ChatGPT about Cruisers Forum it won't include anything here. HOWEVER, developers using ChatGPT can still scrape Cruisers Forum for fine tuning data, or can script ChatGPT to read and post here for additional training. None of that fine tuning would be available to us or anyone that askes ChatGPT anything, it would only be available to the developer performing that training.

ChatGPT does seem to honor robot.txt blocking

Example: Sailing Anarchy
Me - what is the consensus at sailing anarchy?

ChatGPT Based on a quick search, here's what I found.
At Sailing Anarchy, the consensus on the best anchor seems to reflect a variety of preferences and experiences, rather than a unanimous agreement on a single best anchor. Forum members discuss their satisfaction with different types of anchors based on their own sailing conditions, boat types, and personal experiences. Anchors mentioned with positive feedback include:

Example: Cruisers Forum
Me - what is the consensus at Cruisers Forum?

ChatGPT - searching with bing Based on a quick search, here's what I found.

I wasn't able to directly access Cruisers Forum due to their website's restrictions on automated access.

Once again, It isn't necessary for AI to scrape forums, or write a post, or comment, or any of those things. If you want the data to refine for your special AI Tool, you would simply write your own scraping tool or use a premade tool like scrapeowl and feed the data into your model.
This site was most likely scraped a million times before AI even came out.

[wholybee]

Originally Posted by Iron E

ChatGPT does seem to honor robot.txt blocking

Example: Sailing Anarchy
Me - what is the consensus at sailing anarchy?

ChatGPT Based on a quick search, here's what I found.
At Sailing Anarchy, the consensus on the best anchor seems to reflect a variety of preferences and experiences, rather than a unanimous agreement on a single best anchor. Forum members discuss their satisfaction with different types of anchors based on their own sailing conditions, boat types, and personal experiences. Anchors mentioned with positive feedback include:

Example: Cruisers Forum
Me - what is the consensus at Cruisers Forum?

ChatGPT - searching with bing Based on a quick search, here's what I found.

I wasn't able to directly access Cruisers Forum due to their website's restrictions on automated access.

Once again, It isn't necessary for AI to scrape forums, or write a post, or comment, or any of those things. If you want the data to refine for your special AI Tool, you would simply write your own scraping tool or use a premade tool like scrapeowl and feed the data into your model.
This site was most likely scraped a million times before AI even came out.

It does appear from OpenAI documentation that yes, ChatGPT honors robots.tx.

When I say ChatGPT or another AI is scraping a site, yes you are correct it doesn't happen exactly that way. Some script is scrapping the site, and using that for date for a fine tuning. It isn't ChatGPT directly. ChatGPT doesn't have access to the Internet at all, everything it knows ends January 2022 unless augmented with finetuning data or function calls.

The scraping isn't the problem. The problem is what the AI is used for after it has that data. And, (via a script) posting on this site is happening. And I am not sure it is all obvious.

I have become EXTREMLY suspicions about another thread that is trending the past week. I don't want to call it out because it's much more difficult to tell. But the more I think about it, the more it seems an AI is targeting a very specific goal, and blending in as a real human in a much more natural way.

I think too many people are too sure of themselves that they can always spot an AI, without realizing they are integrating with them daily and don't realize.