A community research effort claims to have beaten a prior OpenAI benchmark by a factor of 500,000. The target model is a 124M-parameter variant of GPT-2, which is small enough for hobbyists and independent researchers to experiment with. Autonomous AI agents participating in these projects have produced efficiency gains, with one documented case showing an improvement of approximately 11% achieved with minimal or no human oversight. Training times for that model have dropped from around 74 seconds, with the community now pushing toward targets under 40 seconds. Researchers note that independent replication and precise accounting of the metric will be essential given the magnitude of the claims.
Source: Read the original article
Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
