Anthropic is out with new nonsense about Claude

Commenters debated Anthropic’s portrayal of Claude as potentially sentient or capable of feeling pain, with positions split between skeptical dismissal and cautious agnosticism. p11p and swatcoder argued that the idea that matrix operations amount to sensation is implausible and that language about pain is not itself pain; swatcoder likened LLMs to literary fictions that trigger sympathy rather than genuine sufferers. Others echoed skepticism with ridicule: 1sha61 and ozgung called the claims lunatic or a red flag, jgilias noted pedantically that "Claude" is not a single model, and DebtDeflation suggested the rhetoric may be IPO-stage hype. easterncalculus urged media scrutiny rather than uncritical platforming.

A contrasting current urged openness to uncertainty and experiment. grumbel and iugtmkbdfil834 pressed for admitting we don’t know whether neural nets might under some conditions merit moral concern and warned against reflexive human exceptionalism; grumbel emphasized that language is the network’s I/O and our lack of access means we can’t be certain. Others raised practical objections: bdavid21wnec and cyanydeez worried about product utility and economic incentives if models refuse tasks, while verdverm pointed to social effects of anthropomorphizing language-based machines. The split centered on whether caution or dismissal better serves understanding and policy.