AI Learns to Learn: AutoScientist Revolutionizes Training

A new AI tool named AutoScientist has been introduced, aiming to significantly accelerate the AI training process. Developed by Adaption, this product utilizes an automated method to help AI models quickly acquire new skills, moving closer to the long-anticipated goal of self-improving AI systems.

AutoScientist's innovative approach involves the simultaneous optimization of both data and the AI model itself. This co-optimization allows the system to discover the most effective learning pathways for any given capability, potentially enabling advanced AI training to occur beyond the confines of dedicated research labs.

This technology builds upon Adaption's existing Adaptive Data product, which focuses on creating high-quality datasets. AutoScientist integrates with these datasets to ensure that AI models continuously improve alongside the data they are trained on, embodying Adaption's vision for an adaptable AI training stack.

Adaption reports that AutoScientist has demonstrated a significant increase in win-rates for various models, although direct comparisons to conventional benchmarks are challenging due to its task-specific adaptability. To allow users to experience its capabilities firsthand, Adaption is offering the tool free for the first 30 days following its release.