Comments by "clray123" (@clray123) on "Tina Huang" channel.

  1. You know you're in trouble when random youtubers start shitting on your product instead of hyping it up. Apparently AI companies are not paying out enough bribes yet.
    1
  2. Policy and regulation is not "really encouraging", it is the most authoritarian approach of them all. Like what CCP is doing in China. You are not a CCP agent, are you?
    1
  3. As far as open source role goes, I think you are about 1 year late to the party.
    1
  4. You correctly recognize that AI companies use safety as a pretense to obtain monopoly (by regulatory capture). Yet in the same video you act as if the "safety" argument really mattered, i.e. if the public was in danger. There is no danger. The only danger is the regulatory capture and the big corps cooperating with oppressive governments who also like to pretend they are acting "for your good and safety". You should not pick sides in this argument because as many misleading arguments by arbitrageurs it is constructed so that whichever "side" you pick, you lose. It is based on the false dichotomy fallacy, which is very often employed by the powerful to mislead their victims into behaving as they would prefer them to. Like in any compenently executed con, the goal of those who run it is for you to exclaim "oh, I see through your tricks" - and then, having been distracted and emboldened by this, fall for the actual confidence trick. So I would advise you to reconsider and stop helping these confidence scammers by further misleading the public about it. The whole safety argument is a scam. It originated from arm-chair Internet philosophers imagining AIs getting out of control and pursuing their own goals to the detriment of humanity. It was popularized around the time AlphaGo and computer playing AIs were developed, and developers observed the phenomenon of "reward hacking" in models trained with reinforcement learning. Not much has come out of it since. LLMs do not even use reinforcement learning all that much, and certainly not in the sense of a reward function which obtains a better "score" through cheating.
    1