Subnostr

Thought provoking comments. Let me respond:

1. The value in an "AI censorship death star" is not to censor past topics, but to censor upcoming topics in realtime. So while the data cannot be unscraped, this throws a spanner in the works for scraping going forward (which is what matters)

2. Using 2021 numbers there were around 500 million tweets per day, meaning you would need 500,000 bots in the botnet. Certainly possible to have a botnet of that size, however I suspect Twitter would be able to detect commonality in such a large botnet and disable it. But yes, possible, even though I rate it as unlikely.

3. You might be surprised with how sophisticated the state sponsored censorship apparatus is. #twitterfiles showed that AI was used by government agencies to censor Twitter under nostr:npub1sg6plzptd64u62a878hep2kev88swjh3tw00gjsfl8f237lmu63q0uf63m , so we don't have to debate this - the empirical evidence shows us that AI gets used. It's way more effective than anything ever used in the past.

kgothatso@8333.mobi 2y ago

1. How have WeChat and other Chinese social media platforms been implementing censorship?

2. What's the difference between a bot that only "reads" tweets (never posts) and a user that only reads tweets? How would you cluster such behaviour (especially when each user/bot is tuned to "read" different content from another member in the cluster)?

3. For as long as a platform enforces policy requests to censor it really doesn't matter what tool is used to determine what to censor. The enforcement will and does get abused time and time again. On your point of the AI being an effective tool for the #twitterfiles, the thing that was surely more effective then AI is the decree that only "trusted" sources should speak/amplified on a topic. Once this decree is put out the AI is just sprinkled on top of the enforcement so that when the censorship gets called out the ones who wrote/enforced the policy can say "the AI did it".

Reply to this note

Please Login to reply.

Discussion

Deleted Account 2y ago

1. In real-time, using AI

2. We need to distinguish between bots and botnets. The former is typically short lived and often has a human in the loop, whereas the latter are long lived and are more autonomous.

3. A decree is meaningless without enforcement.

kgothatso@8333.mobi 2y ago

Just back to drop this and pass by.

https://void.cat/d/NKf9omWTzpSnzgvNczRekw.webp

Deleted Account 2y ago

It seems the way this is panning out is that the manufacturer of the computer is held responsible. Not that they don't make management decisions.

kgothatso@8333.mobi 2y ago

Yes, but that is being done to limit access to the use of AI by wrongfully placing blame on the creator/trainer of the AI.

The computer is just a tool, and the AI running on that is just the same as every other program installed on that machine.

A gun is also yet another tool. A bullet is just a projectile loaded in that gun.

If someone users a gun to shot somebody we shouldn't blame the manufacturer of the gun for the death. s/gun/AI; s/kill/censor. AI don't censor people. People censor people.

If a person says they were just doing their job when they pulled the trigger then please free up some shelf space in the history section for some new best sellers.

Deleted Account 2y ago

Don't worry, the humans will be out of the loop soon. They already are in many places and ways.