Comment on ChatGPT spends 'tens of millions of dollars' on people saying 'please' and 'thank you', but Sam Altman says it's worth it

• CileTheSane@lemmy.ca ⁨4⁩ ⁨days⁩ ago

  If my tool sucks, I will tell it so

  So thanking your tools: dangerous on a humanity level scale

  Telling your tool it sucks: Normal behaviour

  source
• sugar_in_your_tea@sh.itjust.works ⁨4⁩ ⁨days⁩ ago

  Exactly!

  I’m a parent, and I set a good example by being incredibly respectful to people, whether it’s the cashier at the grocery store, their teacher at school, or a police officer. I show the same respect because I’m talking to a person.

  When I’m talking to a machine, I’m direct without any respect because the goal is to clearly indicate intent. “Alexa play <song>” or “Hey Google, what’s <query>?” They’re tools, and there is zero value in being polite to a machine, it just adds more chances for the machine to misinterpret me.

  Kids are capable of understanding that you act differently in different situations. They’re super respectful to their teachers, they don’t bother with that w/ their peers, and us as parents are somewhere in between. I don’t want my kids to associate AI/LLMs more with their teachers than their pencils. They’re tools, and their purpose is to be used efficiently.

  source
• nickwitha_k@lemmy.sdf.org ⁨3⁩ ⁨days⁩ ago

  why are you arguing that at me?

  Rationally and in vacuum, anthropomorphizing tools and animals is kinda silly and sometimes dangerous. But human brains don’t work do well at context separation and rationality. They are very noisy and prone to conceptual cross-talk.

  The reason that this is important is that, as useless as LLMs are at nearly everything they are billed as, they are really good at fooling our brains into thinking that they possess consciousness (there’s plenty even on Lemmy that ascribe levels of intelligence to them that are impossible with the technology). Just like knowledge and awareness don’t grant immunity to propaganda, our unconscious processes will do their own thing. Humans are social animals and our brains are adapted to act as such, resulting in behaviors that run the gamut from wonderfully bizzare (keeping pets that don’t “work”) to dangerous (attempting to pet bears or keep chimps as “family”).

  Things that are perceived by our brains, consciously or unconsciously, are stored with associations to other similar things. So the danger here that I was trying to highlight is that being abusive to a tool, like an LLM, that can trick our brains into associating it with conscious beings, is that that acceptability of abusive behavior towards other people can be indirectly reinforced.

  Basically, like I said before, one can unintentionally train themselves into practicing antisocial behaviors.

  You do have a good point though that people believing that ChatGPT is a being that they can confide in, etc is very harmful and, itself, likely to lead to antisocial behaviors.

  that is fucking stupid behavior

  It is human behavior. Humans are irrational as fuck, even the most rational of us. It’s best to plan accordingly.

  source