0 reply
0 recast
0 reaction
📸 HAT node
While PIN AI launching token-based program, there is a need to prevent “sybil” accounts. To verify user as basic, verified or high-trust, there are a few methods to make it real: phone check, KYC, blockchain history.
🤖 Reinforcement Learning
Using RL, AI’s learning through trial and mistakes. AI analyze user queries and provides recommendations. GOD model evaluates a quality of answers and AI gets positive or negative feedback and becomes more proactive. If user like AI’s propose – AI get reward, if negative – penalty.
đź’ˇ Personalized Reinforcement Learning from Human Feedback
To understand a P-RLHF mechanic, i want to show a simple example: AI see that user order taxi too often on Monday, 9:00 am, AI propose to order taxi on Monday, 8:55 am, if user agree a few times in a row – AI pin this action. If user disagree or ignore – AI decline this action. 0 reply
0 recast
0 reaction
