search
AI scientist
Trends
- 1
Anthropic has published a research piece titled 'Claude-Shaped Science', exploring how its Claude AI assistant may shape scientific work and the way research questions are framed. The discussion centers on whether scientists' growing reliance on large language models risks steering research toward problems and methods that suit the model, rather than the science itself.
- 2AI doomerism may serve big business interests▼How the bad science of AI doomerism is good for big business
The Bulletin of the Atomic Scientists argues that alarmist claims about artificial intelligence causing catastrophic harm rest on weak science, and that such doomerism paradoxically benefits major tech companies. By framing AI risk as a distant, speculative threat, the piece says, the largest AI firms can distract from present-day harms, shape regulation in their favor, and position themselves as the only safe custodians of the technology.
- 3AI could trigger nuclear war without any evil intent●AI doesn't need 'superintelligence' or evil intent to start a nuclear war
The Bulletin of the Atomic Scientists argues that artificial intelligence does not need to reach superintelligence or develop malicious intent to set off a nuclear conflict. Ordinary flaws — misperception, automation errors, and the compression of decision-making time — could push nuclear-armed states toward catastrophe, warning that integrating AI into early-warning and command systems carries serious risks.
- 4
An autonomous AI system has reportedly conducted its own biology experiments and produced new scientific findings in yeast research. The news describes a 'scientist' AI that runs a laboratory end to end, from forming hypotheses to carrying out experiments and making discoveries. Details about who built the system and what exactly it discovered are not yet provided, but the claim points to a significant step toward fully automated scientific research.
- 5Google DeepMind's watermarked AI proteins pass lab tests▼Google DeepMind creates watermarked AI-designed proteins that still work in lab tests
Google DeepMind says it has produced artificial proteins carrying a built-in watermark, an embedded genetic signature marking them as AI-designed. Crucially, the watermarked proteins still functioned in laboratory tests, suggesting such markers need not compromise performance. The work could let scientists verify when proteins were created by AI, aiding biosafety, intellectual property tracking and trust in lab-made biology as AI protein design becomes widespread.
- 6AI scientist now runs its own biology lab▼AI scientist is now running its own biology lab: It can design experiments, test ideas and make new...
An autonomous AI system described as an 'AI scientist' is now operating its own biology laboratory, where it can design experiments, test hypotheses and generate new findings with minimal human input. The development points to a shift from AI as an analysis tool toward AI as an independent researcher capable of running the full scientific cycle in life sciences.
- 7AI could trigger nuclear war without superintelligence, experts warn●AI doesn't need 'superintelligence' or evil intent to start a nuclear war Article URL: https:// thebulletin.org/2026/10/
A new Bulletin of the Atomic Scientists article argues that artificial intelligence does not need to reach 'superintelligence' or act with malicious intent to set off a nuclear conflict. The piece warns that mundane failures — misinterpreted data, automation errors, or miscalibrated systems integrated into military command structures — could be enough to escalate tensions between nuclear-armed states. Readers are debating how soon such risks become real as militaries adopt AI tools.
- 8AI Learns the Hidden DNA Code Behind Gene Enhancers▼AI Learns the Hidden DNA Code That Marks Enhancers Across Species
Researchers report that artificial intelligence has identified a hidden DNA code that marks enhancers — genetic elements that regulate gene activity — consistently across different species. The finding suggests that regulatory sequences share common patterns that machine learning models can detect, offering new insight into how gene expression is controlled throughout evolution. The work could help scientists better understand genetic diseases linked to regulatory mutations.
- 9Yann LeCun calls Anthropic CEO Dario Amodei 'deluded' over AI risk claims●AI ‘godfather’ Yann LeCun: Anthropic CEO Dario Amodei is ‘deluded,’ ‘crazy,’ and doesn’t understand cybersecurity
Yann LeCun, Meta's chief AI scientist and one of the field's so-called godfathers, publicly attacked Anthropic CEO Dario Amodei, calling him 'deluded' and 'crazy' and saying he does not understand cybersecurity. The remarks sharpen the ongoing split between AI researchers who downplay near-term catastrophic risk and safety-focused lab leaders like Amodei who warn of serious dangers from advanced systems.
- 10Astrobiologist argues AI cannot truly think●Coming up: An astrobiologist doubts AI’s capacity to think
An astrobiologist is voicing skepticism about whether artificial intelligence systems can genuinely think. Appearing in a segment aired by WXXI News, the scientist questions claims that AI models possess real understanding or consciousness, adding a scientific voice to the ongoing public debate over how intelligent these systems actually are.
- 11AI could trigger nuclear war without malice, experts warn●AI doesn't need 'superintelligence' or evil intent to start a nuclear war https://thebulletin.org/2026/10/ai-doesnt-need
The Bulletin of the Atomic Scientists argues that artificial intelligence does not need to reach superintelligence or act with evil intent to start a nuclear war. The piece warns that ordinary errors, miscalibration, or misinterpretation in AI-assisted military systems could be enough to escalate a nuclear crisis, prompting renewed debate about integrating AI into command-and-control systems.