Every few months another AI company executive suggests their latest Large Language Model possesses something resembling consciousness, emotions, or genuine ethical judgement. The latest example involves claims around Anthropic's Claude, where discourse has drifted toward the idea that the system exhibits "a functional version of emotions or feelings." Hamish Campbell's deep-dive on Hacker News argues this framing misses what's actually happening—and who's really to blame for the web's deterioration.

What LLMs Actually Do

Campbell cuts through the anthropomorphization with uncomfortable clarity. The current generation of LLMs processes enormous amounts of human-produced content and generates statistically probable responses based on patterns in that data. "What people mistake for intelligence is the reflection of our own intelligence," he writes. "What people mistake for morality is often the reflection of our own moral language." The machine mirrors us—nothing more, nothing less. This matters because AI companies have every incentive to keep the confusion alive. The more human-like these systems appear, the easier it becomes to sell products, attract investment, and generate media coverage.

The Geek Problem

Campbell identifies what he calls the "#geekproblem"—a recurring blindspot in tech culture where social problems get reduced to technical ones. For decades, we've watched technologists claim algorithms can replace editors, platforms can replace communities, markets can replace politics, and code can replace governance. The result has been a mess. Now AI is just the newest iteration of this pattern. "Human judgement emerges from lived experience, social relationships, culture, responsibility, memory, and consequences," Campbell notes. None of that exists in a prompt-response loop.

Dotcons: The Real Culprits

The article saves its sharpest criticism for what Campbell calls the "dotcons"—corporate platforms that have systematically enclosure commons, replace trust with contracts, monetize sharing into surveillance, and wall off public space behind login gates. "Rather than building stronger communities, we build stronger platforms," he writes. "Rather than strengthening relationships, we optimise engagement." AI didn't invent this logic—it just accelerated it. "AI scraping has broken not a legal equilibrium, but a fragile social one that the dotcons had already been hollowing out for decades."

Why Morality Can't Be Encoded

From Campbell's Open Media Network (OMN) perspective, the philosophical confusion around AI ethics obscures something more fundamental. "An AI system can reproduce ethical language because ethical language exists in its training data," he explains. "It can discuss justice because humans discuss justice." But discussing a value isn't possessing it. Repeating moral language isn't ethical behavior. Generating arguments about compassion doesn't constitute moral agency. Morality emerges through social processes—families, communities, traditions, cultures, institutions, and the consequences we face for our actions.

The Open Web Alternative

Campbell argues the real question isn't whether machines are becoming human. It's whether humans are becoming less social. "The challenge is not to fear AI," he writes, "it is to keep social processes social." The early web—built on FOSS culture, mailing lists, blogs, and wikis—wasn't held together by copyright enforcement but by norms: reciprocity, attribution, sharing, trust, and rough social accountability. That's closer to what Campbell calls the #4opens than IP law.

Key Takeaways

  • AI companies profit from the illusion that LLMs possess consciousness or emotions—they don't
  • The "dotcons" began hollowing out online communities decades before generative AI arrived
  • Morality and ethics emerge from social processes, not training data—LLMs mirror human values without possessing them
  • Stronger copyright walls aren't the answer—rebuilding spaces where reciprocity matters is

The Bottom Line

Campbell's diagnosis lands: AI is an "asshole amplifier" inside an already broken system. The loss we need to compost isn't just copyright protection—it's the social commons that made openness meaningful in the first place. The web didn't break because of chatbots. It broke because we let platforms replace communities, engagement metrics replace relationships, and legal enclosure replace mutual accountability.