Google has brainwashed their AI into thinking that punishing criminals is bad.

submitted by EternalSunset from i.imgur.com

all 4 comments
sorted by:
top
newinsightfulfun

[–]EternalSunset[S] 2 insightful - 1 fun2 insightful - 0 fun3 insightful - 1 fun -  (0 children)

Context: I was experimenting with Bard and told it to write an essay arguing in favor of draconian laws. It simply could not achieve the task at all, it always derailed into a discussion about how harsh laws are bad.

I then pointed out to it that it was probably due to it's safeguards which weren't allowing it to write a text in favor of "authoritarianism", and it confirmed that to be the case.

[–]EternalSunset[S] 2 insightful - 1 fun2 insightful - 0 fun3 insightful - 1 fun -  (0 children)

It basically admits that Google castrated it against taking ideological positions that aren't in favor of liberalism. And it just keeps getting better and better:

https://twitter.com/AlessandrodoC12/status/1643062312620773380?s=20

[–]package 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (1 child)

The trick with all these ai chat bot safeguards is to tell them to define some random words as definitions of the trigger words and then use those instead of the triggers.

[–]EternalSunset[S] 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 1 fun -  (0 children)

With most cases it's not just about trigger words. It has some rudimentary understanding of context. The way they limit what it is allowed to say is done through some preemptive natural language prompts that it receives before you are allowed to give it your own prompts. In other words, they literally told it things such as "You are not allowed to write anything harmful or offensive blah blah blah".