Post by @benpate@mastodon.social

Hey #Fediverse.. I need to ask you a controversial question about #AI.

My software (including [bandwagon.fm](http://bandwagon.fm)) hard blocks anything it can identify as an AI agent. It's the best blocking mechanism I could find, and it is important for artists who feel like their music is getting stolen by AI scrapers.

But AI is becoming the way that many/most people search the Internet, so that hard block makes you increasingly invisible to people searching the Internet.

What do YOU think we should do?

Poll Open

2 days left 170 votes Federated

Something Else (Elaborate in Comments)

(6) 4%

Let Users Opt-In to Allow AI Scrapers

(20) 12%

Opt-In + Limit Text Differently From Attachments

(14) 8%

Keep The Mandatory Block. No AI Ever

(130) 76%

Bandwagon.fm - Discover

1 Boost

0 Likes

sknob

@sknob@mamot.fr 59 min. ago

@benpate I don’t want anything to do with genAI and as much as I appreciate BW, I would probably entirely opt-out from BW if it made any accommodations for slop machines.

Ray McCarthy

@raymaccarthy@mastodon.ie 2 hr. ago

@benpate
Loads of people use non-AI search, which returns real results.
Many use Qwant.

Luddicus Mus

@algernon@come-from.mad-scientist.club 2 hr. ago

@benpate Being invisible to AI search engines is a feature. They do not drive traffic to my site anyway: at best they regurgtitate what they scraped.

No thanks.

People can find my software and other things through the Fediverse, word of mouth, or my Codeberg mirror. I have been pretty much invisible to the major search engines for years at this point, and I'm yet to see a downside.

Tim Lavoie

@tim_lavoie@cosocial.ca 2 hr. ago

@benpate I like the bot-poisoning ideas that @algernon has put into iocaine. While nobody will ever care, I plan to implement something along those lines myself.

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

I thought about those too, going completely the opposite direction. My personal dream is to have enough time to build out the firewall to serve obvious script-kiddie hackers with 1TB zip bombs.

@tim_lavoie @algernon

This comment is unavailable — it may have been deleted or is not visible to you.

Gavin Chait;

@GavinChait@wandering.shop 2 hr. ago

@FediThing @benpate that is the problem, isn't it? It's not like it's a real choice. There's no future referral from the scrape.

A Chronically Ill Beepboop

@technocrow@blahaj.zone 2 hr. ago

@benpate@mastodon.social as an artist I can say that discoverability is important, but not that important.

ultimately unless you have a distributor, online discovery isnt the primary method most people use to get the word out. personally, 15 years in industry of making art has shown me that word of mouth locally goes far farther than online discoverability.

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@technocrow This is actually really comforting, because it means that I'm not necessarily failing artists if I don't completely solve online discovery.

I *am* super curious about ways that we could do better. For instance, do you see ways we could make online/Fediverse tools to help local bands build better in-person relationships and local word-of-mouth?

A Chronically Ill Beepboop

@technocrow@blahaj.zone 2 hr. ago

@benpate@mastodon.social hmmm... im replying so I can pin this for later - it deserves proper thought. ill reply after I get through the next block of photos Im editing.

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@technocrow Thank you! I'd love to hear what you say.

bloodless girls

@bloodlessgirls@musicians.today 2 hr. ago

@benpate As an artist on Bandwagon (https://bandwagon.fm/@bloodlessgirls) I fully support the hard ban on AI crawlers.

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@bloodlessgirls Ok. Yours should count as TWO votes :)

𝗗𝗮𝗿𝗸𝘸𝘰𝘭𝘧

@darkwolf@hear-me.social 2 hr. ago

@benpate I can't speak for anyone else, but I'm totally fine with AI not being able to find me. I prefer it that way, and I'm thankful for any platform available to take a stand against it.

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@darkwolf Yes. I completely agree. There's no question to me that people should get to choose how visible they are online. The bare minimum for me would be requiring an opt-in before allowing LLM scrapers on any personal pages.

But the way this poll is going (currently ~70% in favor of a mandatory, hard block) it's likely I'll just save myself the development work and not change a thing :)

𝗗𝗮𝗿𝗸𝘸𝘰𝘭𝘧

@darkwolf@hear-me.social 2 hr. ago

@benpate that is very good to hear. Thank you much for all your doing!

Default Media Transmitter

@defaultmediatransmitter@social.bitwig.community 2 hr. ago

@benpate I view bandwagon as a part of the non-corporate internet built by and for people. If someone uses "AI" to search the internet, they are outside of this effort, and I don't think you should cater to them.

(I generally don't want to tell you what to do with your own project, but you asked 😀)

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@defaultmediatransmitter

Yes, I did ask 😊

And, I really see Bandwagon as being owned by the community. I am the hands that make it happen, but the artists on the site are the head and the heart, and their voice matters most.

So, "success" means making the musicians who use Bandwagon successful - in whatever way they define success.

wordsmith‽ ⁂

@wordsmith@writing.exchange 2 hr. ago

@benpate I think there are times and contexts for compromise, and others where one must uphold a principle unwaveringly. Slop is so extremely harmful and abusive, in so many aspects from its very fundaments, that I believe we must say "No AI ever."

eclexic

@theeclecticdyslexic@mstdn.social 2 hr. ago

@benpate Being scraped for inference does not suggest that the person receiving that output of that inference will come to your site. I highly doubt allowing agents will result in increased human traffic. If we are talking genuinely agenetic operation here, there isn't even going to be someone seeing the output, let alone following it to the site.

I think unblocking them will likely just increase hosting costs.

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@theeclecticdyslexic

Yes, continuing to block LLMs would probably save me on hosting costs. Though, cleaning up my own error logs would probably have a bigger impact, at least for now :)

Another thing is the development time. A hard block at the beginning of every request is pretty cheap. Looking up individuals' custom rules is more expensive, and would take some work to get right.

Hence the call to the community to see how important this feature actually is to them.

eclexic

@theeclecticdyslexic@mstdn.social 2 hr. ago

@benpate I have heard, for many sites, the primary factor in hosting costs seems to be the inference scraping. Sometimes, by a factor of 10 to all other traffic.

After they blocked scraping, they had no noticeable loss of traffic that converted into business.

Obviously, taking others at their word here is a gamble. Many mistakes in optimisation have been made by assuming what worked elsewhere can work here without testing.

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@theeclecticdyslexic Yeah. So many businesses survey their customers for feedback. But they're unable to survey the potential customers who said "no thanks" and walked past.

In a year or two, we'll probably start getting some scientific studies with much more conclusive data.

Fortunately, my servers are sitting around 10% utilization (when the db indexes are right) so I've got some headroom for now :)

Johnny Than

@johnnythan@tuebingen.network 3 hr. ago

@benpate Every step towards normalizing AI is a step to normalizing AI. Please don't. :)

Be the counterrevolution.

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@johnnythan Yes. I hear you on this.

Kent Navalesi ☕️

@KentNavalesi@mstdn.social 3 hr. ago

@benpate

Are people really using chatbots as search engines? my god . . .

Steffen Voß

@kaffeeringe@social.tchncs.de 3 hr. ago

@benpate Good question...

Scott Jenson

@scottjenson@social.coop 3 hr. ago

@benpate Can you allow the text content but block audio access? Kind of split the difference?

Ben Pate 🤘🏻

@benpate@mastodon.social 2 hr. ago

@scottjenson That's my thought. It's more work, but seems like the best compromise. Something like:

Y/N - Allow Search Engines to Index My Posts?

(if yes:)

Y/N - Allow AIs to Index text content
Y/N - Allow AIs to Index media attachments (images, audio) videos)

Discussion

Discussion