Anthropic has opened its own laboratory for biological experiments

Anthropic has officially launched a research laboratory specializing in conducting physical biological experiments. The facility is designed to test hypotheses generated by the company's models in real laboratory conditions.

The main task of the division is to validate the data that AI models use to analyze biological threats and develop medicines. The laboratory is equipped with equipment for automated DNA synthesis and biochemical analysis, which allows reducing the "prediction - verification" cycle from several weeks to several days.

As part of the laboratory's work, researchers are testing the ability of neural networks to predict protein interactions and the effectiveness of molecular compounds. The results obtained during the experiments are directly integrated into the process of retraining models, which allows adjusting the weights of neural networks based on empirical data, rather than just synthetic datasets.

Investments in physical infrastructure are aimed at reducing the risks associated with the use of AI in biological research. The company plans to use the results of the laboratory's work to create safety protocols that prevent the generation of hazardous pathogens or toxic compounds when using models for scientific purposes.

This step marks the transition from purely software-based research to a hybrid development model, where AI systems undergo mandatory verification in a "wet" laboratory. Such integration allows the company to more accurately assess the capabilities of its models in the field of biotechnology and medicine.