Avoiding Disinformation

#Draft

Neither Reddit, nor ChatGPT are vulnerable to disinformation.

Reddit - comment chain manipulation - reply and block, with or without moderator assistance - dangerous because it initially gives the impression that it is immune to manipulation

this works because 1) if a reply to a comment is deleted before it received a reply that isn't deleted too, it won't show up as a deleted reply when someone browses the thread; and it won't seem deleted to the user too, so they won't know that the manipulation is at play; and 2) if someone replies and blocks you, you won't be able to reply to them, and this will make it seem like you didn't have a better reply, in case of political discussions - e.g. if someone asks you for proof and blocks you so that you can't reply with it.. - and if you edit the original comment, if it then gets deleted with moderator assistance in suppression of opinion, it'll be as if you didn't reply at all. They will try to frame you if they can, and if they fail, then they will remove your opinion. Framing is better than suppression in destroying adversaries.

This is also bad, because this happens at the subreddit level, and Reddit moderators cannot be fully blamed for what happens.

ChatGPT - regurgitates misinformation in training data, and does not ever self-correct unless you force it to in each conversation

Lemmy - much more dangerous in that it gives dangerous people access to a platform. Many Lemmy instances have no privacy policies. They retain usernames even if the reply is deleted. It would be dangerous to participate in an extremist space in Lemmy, only to later realize that you could be doxxed by some link somewhere. You won't be able to delete your account and escape easily. Additionally the same comment chain manipulation of Reddit applies, but it could even be worse.

Trust should be the foundation of all platforms and communications. You cannot trust an LLM that is not sentient nor has an ethics of its own, unless you trust the platform providing the LLM fully and also that the information fed to it is actually accurate. An example I found is when a Sanskrit term important to Advaita Vedanta was translated wrongly by many Western writers, and as a result, ChatGPT continually misinformed everyone who tried to study it using it, leading them to feel that Advaita Vedanta was like Neo-Advaita (Advaita Vedanta stripped of Vedanta and nuances).

Because of this, I'd rather trust Zucky any day more than Fediverse. Also, be wary when participating in communities where the community admin has control over who can respond.

Comments

Post a Comment

Popular posts from this blog

Open Sourcing Video Games

#Draft - Clients can already be reverse engineered to make cheats - Any detection should be at the server side - Problem 1: Viewport width can be adjusted by players Solution: In that case, anticipate that and design the game around it. - Problem 2: Use of AI assisted macros Avoiding AI assisted keypresses is not always possible. Solution: The best thing to do would be to report and ban hackers. And people should hack less because there should be no inherent value in your scores, it's just a fun game. It will also keep the players from getting addicted.
Read more

Don't steelman a fascist (Wisdom of the Year)

Steelmanning is a method used by super honest debaters, whereby they state the opponent's position better than themselves, before chopping it down. This is as opposed to strawmanning, whereby one makes a distorted version of the opponent's argument, or a strawman, and breaks it down. The goal is to break the opponent's argument in such a way that no part of the argument will survive the attack, as the edge cases are also included when you steelman the argument. However, when you steelman a fascist, you're playing right into their trap. A fascist wants to gain popularity, and they want you to strengthen their argument as well as possible. However, once you have done your part, they won't let you deal the breaking blow. They will silence you by whatever means necessary. A fascist is wicked. But the worst part is that, you never know you're steelmanning a fascist until you've steelmanned them. So if you ever encounter them, don't try to correct your mistake...
Read more

Why is negative times negative positive?

While I was philosophizing, I forgot how this could happen XD! So I had to rethink how it works. So multiplication can be thought of as repeated addition. This even true for non-integer real numbers, just that sometimes you are adding it only by a non-integer proportion sometimes. Addition is commutative, so you can explain adding A B times also as adding B A times. With negative numbers, when we take B from A and if B is greater than A, we can say that the result is a negative number with the value of taking A from B, so that the borrowed amount is represented as the difference between the larger and smaller number. Since we have established the origin of negative numbers through subtraction, we can add a positive number to a negative number to bring it closer to the positive realm. Adding a postive number reduces the borrow from the negative number, until it reaches zero, and then the familiar addition of applies. Since we have established subtraction as removal earlier, we can use t...
Read more