Podrobná bibliografia
| Názov: |
The Collapse of GPT. |
| Autori: |
Savage, Neil1 (AUTHOR) neil@stefan.com |
| Zdroj: |
Communications of the ACM. Jun2025, Vol. 68 Issue 6, p11-13. 3p. |
| Predmety: |
*ARTIFICIAL intelligence, *PERFORMANCE evaluation, *DATA analysis, CHATGPT, LANGUAGE models, ARTIFICIAL intelligence tests |
| Abstrakt: |
The article discusses the possibility that artificial intelligence (AI) systems might not perform at an optimal level in the future due to the inclusion of AI-generated information in the large language model (LLM) training data, and it mentions chatbots such as ChatGPT. According to the article, LLM model collapse is a statistical problem where the distribution of words or parts of words (tokens) does not match a human-produced natural distribution. The curation of synthetic data is assessed. |
| Databáza: |
Business Source Index |