Trajectory raises $45 million to automate artificial intelligence model training

The artificial intelligence startup Trajectory has raised $45 million in funding, TechCrunch reported. The one-year-old company is now valued at $300 million. [1]

According to TechCrunch, current artificial intelligence models from creators like OpenAI and Anthropic, as well as open source alternatives, fail to learn from their mistakes without additional training. Companies in specialized fields such as law and finance usually fine-tune these models by hiring internal research teams or using costly consultants. [1]

TechCrunch reported that Trajectory was founded by three Stanford University classmates who previously worked at Google DeepMind and Apple. The startup aims to automate the fine-tuning process, enabling models to learn directly from feedback and continuously improve. [1]

In the broader field of artificial intelligence development, recursive self-improvement describes a process where a system repeatedly designs a more capable successor. Projects involving recursive self-improvement remain ongoing, though human work is still required and can create a bottleneck. [2]