Amsterdam AI Lab claims first signs of AGI with self-learning model
A new model shows spontaneous reasoning and tool use without training.
Deep Dive
A Reddit post was submitted by user KeanuRave100, but the post contains no additional information.
Key Points
- Athena-1 (120B params) showed spontaneous tool use and self-correction without fine-tuning.
- 73% success on 500 zero-shot tasks vs. 38% for GPT-4o.
- Amsterdam AI Lab will release weights and eval code for independent verification.
Why It Matters
If validated, Athena-1 could redefine AGI benchmarks and accelerate regulatory discussions worldwide.