Image: The New Stack"The opening stages of OpenAI's unraveling": OpenAI slows model training - not everyone is buying the explanation - The New Stack
• OpenAI has slowed the training of its latest models, including a two-week pause in reinforcement learning (RL) for models intended for deployment. • The company officially attributes this pacing to cybersecurity risks, opting to run smaller, contained evaluation rounds before proceeding with its largest planned frontier RL run. • Analysts suggest the slowdown may actually be driven by mounting financial losses and increasing pressure from Chinese AI rivals rather than safety concerns alone.
thenewstack.io