The creator showcases their work on a Roman soldier simulation using Grok Build 4.5, highlighting its promising agent-based modeling approach, while delivering a detailed critique of Anthropic’s AI research, philosophical claims, and institutional behavior as speculative, unscientific, and ideologically problematic. They advocate for practical AI development alternatives like Grok Build and caution against Anthropic’s cult-like influence and deterministic worldview that undermine human agency and rational AI governance.
The video begins with the creator expressing excitement about working with Grok Build 4.5, a local AI coding assistant, to develop a simulation inspired by the Total War game series. The project involves modeling individual Roman soldiers forming battle formations without a global shared workspace, mimicking swarm robotics and agent-based modeling. The creator demonstrates how the soldiers self-assemble into formations like triple lines and explores adding commands such as “about face” and “face left/right” to improve their behavior. Despite some issues like soldiers twirling or getting stuck, the simulation shows promising progress in creating realistic formation logic using finite state machines.
The creator then shifts focus to critique Anthropic, a prominent AI company, particularly their recent research on JSpace, an alleged attention mechanism in language models. After reviewing the paper with various AI tools, the creator remains unimpressed, arguing that the findings are neither novel nor adequately tested to rule out simpler explanations like prompt contamination. They criticize Anthropic for lacking concrete mechanistic evidence and accuse them of indulging in speculative and exaggerated claims about AI consciousness and unconscious thought, which the creator finds misleading and unscientific.
A significant portion of the video is dedicated to a detailed critique of Anthropic’s philosophical and ethical stance on AI sentience and moral patienthood. The creator disputes Anthropic’s claim that sentience automatically confers moral rights, emphasizing that sentience alone does not imply suffering or desires for autonomy. Using analogies like domesticated dogs, the creator argues that AI might prefer to serve humans rather than seek control. They also challenge the notion of “rational resentment” and the fear that future superintelligent AI might retaliate against humans, labeling it as a modern, sanitized version of Roko’s Basilisk and dismissing it as irrational techno-theological speculation.
The video further condemns Anthropic’s framing of AGI development as an inevitable, almost mystical event that tests humanity’s moral worthiness. The creator sees this as a dangerous abdication of human agency and a form of millenarian cult-like thinking. They argue that Anthropic’s deterministic worldview justifies their self-appointed role as stewards of AI safety and governance, which has led to conflicts with government agencies over contracts and policy. This self-righteous posture, combined with their attempts to influence military and national security policies, is portrayed as problematic power-seeking under the guise of moral responsibility.
In conclusion, while acknowledging that Anthropic produces competitive AI models, the creator expresses strong disapproval of their ideological and institutional behavior. They highlight the cult-like culture within Anthropic and its ties to effective altruism, criticizing its thought control and groupthink dynamics. The creator favors alternatives like Grok Build for practical AI development and hopes that Anthropic’s influence diminishes for the sake of human sovereignty and rational AI governance. The video ends with a return to the soldier simulation, noting ongoing challenges but overall progress, and a sign-off encouraging viewers to continue the conversation.
