A new study suggests that ChatGPT’s abilities decline with age
Researchers from Stanford and UC Berkeley conducted a study on OpenAI’s chatbot, ChatGPT, and found that its abilities have worsened over the past few months.
They asked different versions of ChatGPT to solve math problems, answer questions, write code, and do spatial reasoning.
Image by AI
The latest model, ChatGPT-4, showed a significant decrease in accuracy compared to its performance in March. It had trouble identifying prime numbers and generating new lines of code. Additionally, its responses to sensitive questions became more concise, refusing to answer without providing explanations as it did before.
The researchers suggest continuous monitoring of AI models to ensure their quality remains consistent. OpenAI is also planning to create a team to manage risks related to superintelligent AI systems in the future.
ChatGPT experiences 10% traffic drop in June post-initial user rush
ZK proofs can revolutionize the internet, beyond just Web3, according to Aleo executive