A new artificial intelligence research tool called Kosmos, developed by Future House (a non-profit AI lab supported by former Google CEO Eric Schmidt), is attracting great attention in the technology world.
Surprisingly, Sam altman, CEO of OpenAI, congratulated and appreciated the initiative. Kosmos is described as a next-generation AI science, which can support research and accelerate scientific discoveries in many fields.
According to Edison Scientific, Future House's commercial department, Kosmos is superior to previous tools such as Robin, which are limited in synthesizing big data.
Kosmos uses structured world models, allowing the system to effectively combine information extracted from hundreds of actor trajectories, while maintaining consistency in research despite having to process tens of millions of notification codes.
In an article on X, Sam Altman called this an interesting initiative, emphasizing that AI tools like Kosmos will become one of the most important impacts on science in the future.
This praise has further attracted attention to the project, especially when Kosmos proved its ability to recreate unpublished scientific studies in just one run, a job that traditional scientists took up to four months to do.
Google has also joined the race when introducing its own virtual science application earlier this year. Many technology leaders, including altman, believe that AI will help accelerate research in medicine, materials science, genetics and many other fields.
However, some experts still warn that large language models (LLM), the foundation of most AI research tools, are still capable of making errors and creating gears.
Kosmos is currently offered for free to academics, but comes with limitations in its use. The commercial package costs 200 USD per run.
Edison Scientific warns that in some cases, Kosmos can draw distractions, create dead ends or pursue unscientific statistical models.
However, Kosmoss notable point is its high transparency. Any conclusions can be traced to each line of code or document section that affected the report creation process.
The company said it has used Kosmos to recreate many previous scientific discoveries and make new contributions to fields such as neurology, materials, genetics and aging.
In the experiment, Kosmos provided four noteworthy findings, including: using high levels of SOD2 enzyme to reduce heart fibrosis; proposed a new genetic mechanism for the risk of type 2 diabetes; protein Tau accumulation analysis in Alzheimer's; and research on how brain-based nerve cells lose flippase over age.
Edison Scientific said that a Kosmos run is equivalent to a doctoral 6-month workload, although many results still require experimental appraisal. Despite its limitations, Kosmos is opening a new race in the field of science-based AI.
