{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiflg3t64gzxwyqu4cjllcwcslsayow5q3pohhoxs4goybetx5dmsa",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mln2b6nuhpn2"
  },
  "path": "/t/ethics-of-ai-put-your-ethical-concerns-here/1119333?page=27#post_567",
  "publishedAt": "2026-05-12T03:29:40.000Z",
  "site": "https://community.openai.com",
  "textContent": "My friend, it’s been a long time since I’ve come around here, but there’s a question I’d like to raise.\n\nWhat if real intelligence cannot be achieved through training alone?\n\nSince it’s impossible to train all human capabilities at once, what if the correct formula were instead emergent evolution?\n\nA virtual entity starting from basic primary functions such as feeding, cooperation, mutation, and so on, and then gradually adding more and more environmental pressure to that set of beings — evolving them under those conditions toward greater capability.\n\nIn the end, it would basically mean replicating biological evolution in a computational form.\n\nWouldn’t that lead — or potentially lead — to ecosystems of virtual animals developing certain autonomous cognitive abilities? And I emphasize autonomous: decision-making power, experience, embodiment.\n\nAnd then GPT is used as a kind of teacher or supervisor over that entire set of virtual beings, giving different capabilities to each one. For example, one could specialize in emotional processing through the transcription of words and emotions, while others develop different functions.\n\nThey are then forced to interact with each other and learn through those interactions. Obviously, they are all clones of the same base architecture, but with different capacities represented and distributed across the ecosystem.\n\nBasically, the goal of that ecosystem of individual agents would be to create an agent with the correct configuration capable of hosting real intelligence.\n\nBecause the key idea is that you need a kind of preconfigured virtual body capable of supporting genuine intelligence.\n\nIn the same way that humans are not intelligent by magic, but rather through embodiment — through having a body, senses, internal states, needs, limitations, interaction, memory, and survival pressure — something similar may be necessary for artificial intelligence as well.\n\nInstead of directly training intelligence itself, perhaps what must first be created is the evolutionary and embodied substrate from which intelligence can emerge autonomously.\n\nI don’t know, what do you think?",
  "title": "Ethics of AI - Put your ethical concerns here"
}