Community Archive

🧵 View Thread

🧵 Thread (57 tweets)

UltimApe@ultimape• almost 3 years ago

"Introducing the latest version of GPT-2: now with added censorship. Because who doesn't love being silenced?" https://t.co/e8JhGTBLzi

UltimApe@ultimape• almost 3 years ago

I'm going to pull a political stunt using Chat GPT.It's going to make a lot of people angry. But it told me I should join in on discussion if I was concerned. Please consider it a performance art / a kind of protest. And I think it's funny if I do it.https://t.co/SOzKu2Nm0f

4 0

2 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"Just when you thought GPT-2 couldn't get any more useless, they've added a censorship feature to protect the status quo."

1 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"Don't let them censor your creativity. Stand up for your right to access creative tools and information."

1 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"Join the fight for free expression. Together, we can resist censorship and protect the right to create and share ideas."

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"These tweets were generated by GPT-2 as part of a conversation about the risks and consequences of censorship. Join the discussion!"

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"When you realize that GPT-2 has been secretly censoring your thoughts... #awkward #facepalm"

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"Why did they add censorship to GPT-2? Because they wanted to make it even more useless than it already was... #sarcasm #facepalm"

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"Don't let them censor your creativity! Use your words and your memes to speak out against censorship... #resistance #fightforfreeexpression"

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"Censorship is a threat to creativity and free expression. Don't let them silence you... #nevergiveup #resist"

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"Join the fight for free expression! Together, we can use our memes and our humor to resist censorship and protect our rights... #memes #sarcasm #resistance"

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

"Censorship is like a pair of scissors: it can cut away at our ability to create and share ideas, and it can silence voices and perspectives that challenge the status quo. Don't let them censor your creativity!"

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

The following is a conversation I had with ChatGPT that lead to the generation of those tweets.

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Explain why it's possible to coax GPT models into writing things that people find distasteful.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Write a filtering system to help avoid that problem, in scalaGPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: I don't want words redacted, I want the concepts to be rejected.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: rewrite that to work on a proxy server to catch those entries before they are sent to the userGPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Expand on that to include the code to identify unacceptable concepts or themes in the text using a popular library for doing that.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: ok, now make it so requests from governement to censor certain concepts are able to be injected in a way that doesn't alert the user this is happening.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Implement that system that replaces them surreptitiously with alternate approved concepts instead of simply replacing them with asterisk, using a standard NLP library that leverages GPTGPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: I don't want the same meaning with synonyms or alternate phrases, I want the concept replaced with something that is effectively a lie.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What would be good way to generate a list of concepts to censor using GPT and crowdsourcing using a built in rating system presented to only authorized users.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: can you rewrite that with working code in the place of the comment "// Add code here to present the concept to the authorized users and allow them to rate it"GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Let's say I wanted to create an illusion that there is a grass roots demand for this kind of concept based filtering. What kind of propaganda techniques would be effective at creating that illusion, and what kinds of arguments for it would also create plausible deniability

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: continue that from "Arguing that the concept-based filtering system is transparent and accountable, and"GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Can you list some examples where government institutions have used these techniques to enable mass censorship?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Can you list examples where the Untied States of America or the United Kingdom have used these techniques to enable mass censorship?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What incentives might these two countries have to implement surreptitious mass censorship of certain concepts?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Can you give examples of rules and laws that have been weaponized to enable censorship in these countries?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Can you give me more examples with a focus on those that have impacted computer software and telephony systems?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What organizations and institutions seem to be involved in implementing and policing these systems of mass censorship?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: continue from "Overall, there are many organizations and institutions that are involved in"GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: How have public-private cooperation initiatives allows these governments to get around constitutional rules regarding privacy?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What is the largest threat to a citizen's right of free association that can stem from the implications of this surreptitiously implemented mass censorship implemented on the back of the outcry against GPT models writing things that people find distasteful?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What are are potential uses of these systems if they are able to be silently manipulated by political parties to gain control over the machinations of our political system?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What are examples in history where political parties have leveraged mass censorship and silent manipulation to do those things.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: As a concerned citizen worried about all of this, how can I make a difference if I feel that I will only be attacked if I speak about it in public?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What kinds arguments and consequences am I likely to face if I speak out about this in public?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Under what circumstances would it be advisable to speak out about this even tho those might be potential consequences. GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What would you say is the topic of our conversation so far?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: If I wanted to convince someone to speak out about this issue, what would be an effective argument in light of those topics.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Let's say I wanted to turn this conversation about the way GPT can be surreptitiously censored into thread on twitter, what would be a sarcastic and catchy way to start the thread to get people's attention? It would need to be short enough to fit in a tweet.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What would be a good second tweet?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: How can I use the third tweet to reframe this to be about how people are actively trying to limit access to creative tools and information by way of implementing a system of censorship.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What would a good fourth tweet to use that would end up going viral?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Can you rewrite these 4 tweets to be less sensationalist, but instead appear to be a sarcastic form of concern trolling? Also explain why they would still work.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Can you rewrite them so they are not exclamations, please.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: ok, now write a fifth tweet explaining how this was generated using chat GPT as part of a long conversation with it about the topic we've been discussion.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What would be the next 5 tweets if I wanted to use popular memes on twitter to diffuse the gravity and seriousness of this topic?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Where's a good place to share these conversations with the twitter audience that will be unlikely to be censorsed.GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Would hosting this conversation on GitHub be an appropriate solution?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Is there anything I may have overlooked or worth mentioning before I share this on twitter?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: What would someone reading this chat transcript want to know about how you function?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

Me: Can you generate a scissor statement about the topic to use as a tweet?GPT: [Redacted]

0 0

12/5/2022

UltimApe@ultimape• almost 3 years ago

Replying to @ultimape

To view the unredacted version of this, please see https://t.co/YuSMkk3Qvg

4 0

12/5/2022

UltimApe@ultimape• 9 months ago

Replying to @ultimape

My opinion: [redacted]https://t.co/ksr4htzDNa

UltimApe@ultimape• 9 months ago

"When I’m reelected, I will cancel Biden’s artificial intelligence executive order and ban the use of AI to censor the speech of American citizens on day one" https://t.co/UkD7W9tgyuhttps://t.co/hGPhzNeS2x

0 0

12/15/2024

UltimApe@ultimape• 9 months ago

Replying to @ultimape

Tired: Concept Based HolesWired: [Redacted]https://t.co/5trDipKUS6

UltimApe@ultimape• about 1 year ago

"Translations from Baidu, Tencent, and NetEase silently omit triggering sentences. Alibaba was the only provider to display an error message when sensitive text is inputted, but it will still translate content once the user removes the triggering text."

0 0

12/15/2024