▲ 617 ▼ These people are writing policies based off information they’re getting from ChatGPT (lemmy.world) submitted 1 day ago by cannedtuna@lemmy.world to c/microblogmemes@lemmy.world 85 comments fedilink hide all child comments
[–] RememberTheApollo_@lemmy.world 27 points 19 hours ago (1 child) What a typical republican. Who cares if it‘s 100% correct, but it makes me feel good about what I hear. Feels before reals, a snowflake in his safe space. permalink fedilink source hideshow 2 child comments replies: [–] terranoid@lemmy.cafe 21 points 17 hours ago* (1 child) If this was a software engineer discussing how to architect something, I'd just nod and agree, it's not 100% correct but whatever, it's like a rubber ducky method on steroids. HOWEVER this is a fucking politician who writes laws. From a national security perspective, this is 100% fucked. There is huge gain to be had out of monitoring their accounts alone, but this is beyond that. It means monitoring and controlling their thought processes. All Sam Altman has to do is give a special prompt or fine tuned model to politicians and have it lean towards policies he agrees with. You can puppet these dudes by controlling their internal monologue with LLM if the end up relying on it for this... It's totally fucked. It's as if it was the 70s and one of your US senators was going to a psychologist who was secretly a USSR agent. It's that level of fucked. permalink fedilink source parent hideshow 2 child comments replies: [–] RememberTheApollo_@lemmy.world 9 points 17 hours ago* (1 child) The LLM models are already designed to be compliant and not turn people off to their use by correcting them or making them feel bad about their opinions by telling them they’re wrong. IOW, pushback will be minimal if it exists at all so people stay engaged. Grok has already been tuned to not contradict right wing views or at least not criticize them so quickly as it had previously famously done. I haven’t heard about the others, but I’m sure they all bend the knee similarly to their wealthy overlords. permalink fedilink source parent hideshow 2 child comments replies: [–] Katana314@lemmy.world 1 point 7 hours ago This kind of thing causes me to put a lot more trust in an AI answer on the occasions it tells me I’m wrong. (A lot more meaning: More than none) permalink fedilink source parent
[–] terranoid@lemmy.cafe 21 points 17 hours ago* (1 child) If this was a software engineer discussing how to architect something, I'd just nod and agree, it's not 100% correct but whatever, it's like a rubber ducky method on steroids. HOWEVER this is a fucking politician who writes laws. From a national security perspective, this is 100% fucked. There is huge gain to be had out of monitoring their accounts alone, but this is beyond that. It means monitoring and controlling their thought processes. All Sam Altman has to do is give a special prompt or fine tuned model to politicians and have it lean towards policies he agrees with. You can puppet these dudes by controlling their internal monologue with LLM if the end up relying on it for this... It's totally fucked. It's as if it was the 70s and one of your US senators was going to a psychologist who was secretly a USSR agent. It's that level of fucked. permalink fedilink source parent hideshow 2 child comments replies: [–] RememberTheApollo_@lemmy.world 9 points 17 hours ago* (1 child) The LLM models are already designed to be compliant and not turn people off to their use by correcting them or making them feel bad about their opinions by telling them they’re wrong. IOW, pushback will be minimal if it exists at all so people stay engaged. Grok has already been tuned to not contradict right wing views or at least not criticize them so quickly as it had previously famously done. I haven’t heard about the others, but I’m sure they all bend the knee similarly to their wealthy overlords. permalink fedilink source parent hideshow 2 child comments replies: [–] Katana314@lemmy.world 1 point 7 hours ago This kind of thing causes me to put a lot more trust in an AI answer on the occasions it tells me I’m wrong. (A lot more meaning: More than none) permalink fedilink source parent
[–] RememberTheApollo_@lemmy.world 9 points 17 hours ago* (1 child) The LLM models are already designed to be compliant and not turn people off to their use by correcting them or making them feel bad about their opinions by telling them they’re wrong. IOW, pushback will be minimal if it exists at all so people stay engaged. Grok has already been tuned to not contradict right wing views or at least not criticize them so quickly as it had previously famously done. I haven’t heard about the others, but I’m sure they all bend the knee similarly to their wealthy overlords. permalink fedilink source parent hideshow 2 child comments replies: [–] Katana314@lemmy.world 1 point 7 hours ago This kind of thing causes me to put a lot more trust in an AI answer on the occasions it tells me I’m wrong. (A lot more meaning: More than none) permalink fedilink source parent
[–] Katana314@lemmy.world 1 point 7 hours ago This kind of thing causes me to put a lot more trust in an AI answer on the occasions it tells me I’m wrong. (A lot more meaning: More than none) permalink fedilink source parent