Cafe Announcements 📣 Updated! Cafe rules and policies regarding AI and bots.

[Pete Thomas]

As you may all know, further to a customer satisfaction survey (member vote), we gave the saxbot an involuntary retirement package. It (the Saxbot) seemed like a good idea at the time but was an experiment that we deem to have failed. Our policy on AI will mostly be based on what Cafe members think though I believe there will not always be a consensus.

Basically we are aiming at a policy of transparency. All AI generated content (including images) must be declared as such, the only exception being the use of AI for language translation of posts.

In order to aid compliance with the UK Online Safety Act we use NSFW content detection via Sightengine which uses AI. Occasionally this may return false positives as it may not always distinguish the difference between art and life. Note that Sightengine has no rights over any of the content or data submitted for detection beyond the moderation service it provides. If you have any concerns feel free to ask.

 

[Pete Thomas]

We have now clarified the rules around AI and posts that contain AI generated text.

Forums (and information websites in general) are possibly endangered by recent developments in search engines that provide AI answers as a priority over links to information sites and discussion forums. Where people used to ask a question at Google and being directed to sites that may help, they often given an immediate answer via AI (Gemini?)

This may obviously harm websites in general which rely very much on SEO and being found to answer queries. ( I'll be making another post specifically about Cafe survival).

For the above reasons we believe people who do come to the cafe do so primarily to engage with humans so we now actively discourage people from posting AI answers or generating posts using AI.

We did our experiment with the Cafe Saxbot and found that in general people were against it hence this policy.

That isn't to say we forbid all AI. It can be useful especially for our non-native English speaking members who we want to encourage so AI for post translation is fine.

Also if you want to discuss AI or LLMs in general it may obviously be appropriate to include a short quote for illustrative purposes. The newly clarified rules around this stipulate making it very clear what is AI, e.g. use the spoiler block for AI content.

What we think would be against this policy:

• Answering someone's question by getting an AI response and posting it as an answer. Again, we believe people want to hear other peoples' actual experience and learned knowledge.
• Posting an AI generated article., or substantial part thereof

Obviously this is all relatively new, so we may develop the policy as we go along, and also in the interest of doffing our cap towards democracy, we want to hear your views on this.

 

[Grumps]

For the above reasons we believe people who do come to the cafe do so primarily to engage with humans so we now actively discourage people from posting AI answers or generating posts using AI.

Someone on a watch forum asked a question seeking examples of watches that fit his specific criteria. I'd done the same thing in the past and was offered all sorts of examples and photos that were most helpful to me in this regard. Plus, the discussions and insight. All of it much appreciated. So what happens the other day when someone asked this sort of question? The FIRST response was someone telling him that they just did an AI search and found his three watches for him... and like why was he bothering people with something he could do so very simply with AI.

Yeah... I tore him a new one.

 

[Pete Thomas]

and like why was he bothering people with something he could do so very simply with AI.

There is no doubt that AI can give helpful or correct information, but that is a great example of why we think it isn't great for the forum.

 

[Grumps]

Yeah, who'd have thought it.
Grumps is with Humanity.

 

[Digger2054]

I'm torn on AI.

I have customers send me visuals and designs that AI has generated, oblivious to the fact it would need to be recreated in a format I can use.

Some contain lots of images, but you can't ever get the original source file, so it's then impossible to recreate fully. Some aren't even using the correct vehicle.

Ai = Headache.

On the other hand, I did a job the other week, after we finished I asked where they found us, as it wasn't a typical job I do. "Chat GPT"

The customer proceeded to show me the whole conversation, a map of local companies, their details, a draft email, advice on what to expect, what materials and their pro's and cons, even budgets.

The customer had shared my quote with it for feedback, and it gave positive feedback.

That was why they picked me - because I'd done everything that the AI had suggested.

Ai = good

I've shared the same data with both Gemini & Chat GPT, and both have different perspectives and answers, both are 'correct' but just like humans there's an element of opinion rather than fact to what the AI responds.

Ai = A bit worrying if people take what ever output they get as gospell.

It blows my mind, and frustrates me. It's a tool that needs to be used appropriately.

 

[Grumps]

It's a tool that needs to be used appropriately.

It is a weapon to be used to control us.

 

[Targa]

People will come to love their oppression, to adore the technologies that undo their capacities to think
- Aldous Huxley (1894-1963).

 

[Grumps]

Speaking of Huxley, legal weed is the new Soma... for a less brave world, I imagine.

 

[PigSquealer]

IMHO AI information should include citations used for review. Without the citations there’s no telling the date or accuracy of the stated reply. Especially on subjects like values or historical significance. Fleabay is only one source for instance. That’s a national or international value reply. Regional markets are entirely different. IE local prices realized. History, yeah five digit saxophone metal is better in tone. Especially with Roo pads and a gold plated Link NY made in Florida

We just purchased a new car $5k under what net listed prices are. It pays to be human !

 

[MikeMorrell]

There are a lot of potentially harmful applications of AI. And many potential applications are as yet unknown.

Surprisingly, I've used 'AI' (Google Gemini and Microsoft Co-pilot) a couple of times over the past couple of months to develop a 'proposal' to a voluntary organization for which I'm a volunteer. I didn't just 'copy and paste' the AI results in my 'proposal' but I went meticulously through each AI-generated text, correcting, modifying and personalizing the text where necessary.

But I'm amazed at how fast 'Google Gemini' gave me the 'facts' that I needed and how fast Microsoft Co-pilot produced a great 'management summary' of my original proposal!

 

[kamak]

As someone who works in IT, AI can do alot. However, I do not use it personally. I use it minimally in my job, and mostly because people keep asking about it.

AI will replace many white collar jobs. If it is done on a computer, AI can likely do it faster and better than you can right now. If the current model cannot, the next probably will. AI is now being used to make the next model as well. It is using itself to make itself. New models can create an entire functioning application and write thousands of lines of code and do it well. It can also do UAT testing.

I find that AI can be useful for gathering large amount of data and summing it up pretty quickly and pretty well. But it isnt perfect. Taking what it says as truth is highly not recommended. There are biases built in as well for those who make this software, and we'd do well to remember that.