The Thinking Game Film – Google DeepMind documentary

AlphaFold: Optimization vs “Thinking”

• Several commenters stress that AlphaFold is sophisticated curve-fitting/optimization, not “thinking” or general intelligence.
• Concern that marketing and the film’s framing encourage the public to conflate pattern-matching with understanding or agency.
• Others argue that human brains are also computational/optimization systems, so drawing a hard line between optimization and intelligence may be arbitrary.

Scientific Impact and Limits of AlphaFold

• Strong agreement that AlphaFold is one of deep learning’s most genuinely beneficial outcomes.
• Domain experts point out that it solves static structure prediction, not the full “protein folding problem,” and that this does not by itself revolutionize drug discovery.
• AlphaFold can output plausible-looking but physically impossible structures; it optimizes for “looks like a protein,” not “obeys chemistry.”
• Predicting clinical trial success or full human pharmacology is described as vastly harder and likely beyond feasible simulation.

Documentary Tone, Hype, and “AI‑Washing”

• Some found the film inspiring and emotionally powerful, especially the AlphaFold segments.
• Others see it as a polished puff piece or “AI‑washing” for DeepMind/Google and particularly for its CEO, with leader-centric storytelling and limited critical scrutiny.
• The everyday multimodal phone demo and conversational chess-book scene are viewed by some as cheesy, outdated, or potentially misleading given later reports of staged product demos.
• Viewers note the lack of deep technical detail and the near-omission of transformers/LLMs, despite their centrality to current AI.

Hassabis, Open Data, and Motives

• The decision to release AlphaFold predictions openly is praised as genuinely altruistic by some, and as strategically savvy (Nobel ambitions, low direct commercial value, long‑term advantage) by others.
• There is recognition that later models (e.g., for designing new proteins) are more tightly controlled and monetized via pharma partnerships.

AI Use Cases: Entertainment, Science, and Warfare

• Split views on generative media: some see “AI cats and Ghibli art” as wasteful distraction; others argue entertainment drives hardware/algorithm progress and will eventually yield high‑quality, meaningful art.
• Many see the real long‑term value in science and engineering: chemistry, weather, fusion, climate, and broader “AI for science” (physics‑informed nets, operator learning, etc.).
• Several caution that the strongest near‑term applications may be military (targeting, disinformation, “digital fog of war”).

Economic and Societal Concerns

• Debate over whether AI for science can sustain itself economically or is mostly a subsidized public good.
• Worries that AI, like prior technologies, will be used primarily to concentrate power and extract value rather than broadly reduce human toil.
• Others counter with examples like the free AlphaFold database as evidence that not all outcomes are purely exploitative.