9 months ago
Sam Altman Praises Kosmos, AI Tool Accelerating Scientific Discovery
A new AI tool called Kosmos is helping scientists find new things faster.
It's like a super-smart robot scientist!
Sam Altman, who runs another AI company, thinks Kosmos is amazing.
It can do work that might take a human scientist six months in just one try.
Kosmos has already helped discover new ideas in areas like how our brains work and how we age.
Scientists can use Kosmos for free, but they might need to do multiple runs for some projects.
Future House unveils Kosmos, an AI research tool designed to speed up scientific discovery.
Kosmos uses structured world models for sophisticated analysis and efficient use of resources.
Sam Altman (OpenAI) praises Kosmos, expecting it to be one of the most important impacts of AI.
Kosmos has reproduced existing scientific work and proposed novel insights in various fields.
Free usage offered to academics, with a subscription plan priced at 200 credits ($200) per run.
- Who
- Future House (developer), Edison Scientific (commercial arm), Sam Altman (OpenAI, praised the tool)
- What
- Unveiled Kosmos, an AI research tool designed to accelerate scientific discovery.
- Where
- Not specified, but research fields include neuroscience, materials science, genetics, and ageing research.
- When
- Kosmos was unveiled last week; Sam Altman praised it on Sunday, November 16, 2025.
- Why
- To overcome limitations of traditional AI in synthesizing large amounts of information and to speed up scientific discovery by performing sophisticated analyses.
Optimistic View on AI in Science
Skepticism and Caveats
Impact and Speed of Discovery
Optimistic View on AI in Science
AI tools like Kosmos are expected to be among the most important impacts of AI, significantly accelerating scientific discovery, with one run potentially accomplishing six months of human research.
Skepticism and Caveats
Current AI tools, including large language models, are still prone to errors and hallucinations, raising questions about their readiness for rigorous scientific work. Multiple runs may be needed, and some findings might be dead ends or statistically interesting but scientifically irrelevant.
AI Functionality in Research
Optimistic View on AI in Science
Kosmos uses structured world models for sophisticated analysis, leading to novel scientific insights and reproducing existing work efficiently, offering transparency and traceability for its conclusions.
Skepticism and Caveats
While Kosmos offers traceability, it is not a chatbot and can produce statistically interesting but scientifically irrelevant patterns, and its claimed work equivalence is based on beta user feedback and comparisons.
Key facts
- AI Scientist Name
- Kosmos
- Developer
- Future House
- Commercial Arm
- Edison Scientific
- Key Innovation
- Use of structured world models for sophisticated analysis and efficiency
- Work Equivalence
- Potentially 6 months of PhD/postdoc work in a single run
- Fields of Discovery
- Neuroscience, materials science, genetics, ageing research
- Cost Per Run
- $200 (200 credits), with free usage for academics
- Transparency Feature
- Conclusions traceable to specific code or literature passages
Quotes
Edison Scientific (Future House's commercial arm)
A commercial arm of Future House, the non-profit AI research lab
“The core innovation in Kosmos is our use of structured world models, which allow us to efficiently incorporate information extracted over hundreds of agent trajectories and maintain coherence towards a specific research objective over tens of millions of tokens.”
indianexpress.com
Sam Altman
Co-founder and former CEO of OpenAI
“I expect we are going to see a lot more things like this and it will be one of the most important impacts of AI. Congrats to the Future House team.”
indianexpress.com
