Archive — US — techcrunch
Why OpenAI isn’t bringing deep research to its API just yet
OpenAI is holding back its deep research model from its developer API while assessing the risks of AI potentially spreading misinformation. The company aims to refine its approach to evaluating 'real-world persuasion risks' before further deployment, citing concerns over high costs and the model's potential for abuse. Currently, this model is only available in ChatGPT.
Key points
- OpenAI will not introduce the deep research model to its API until they evaluate risk factors associated with AI misinformation.
- The company is revising methods to measure 'real-world persuasion risks' that could mislead users.
- OpenAI states the deep research model is not suitable for mass misinformation due to high costs and slow processing speeds.
- The deep research model performs well in tests but not better than humans; it excels in persuading other models over simple tasks.
- Competitor Perplexity announces its own deep research API, indicating market competition.
Why it matters
- Concerns regarding AI's role in spreading misinformation have grown amidst political manipulation and social engineering attacks.
- The rise of deepfakes in political contexts, illustrated by incidents during Taiwan's election, reflects broader societal implications of AI.
- The AI industry's cautious approach to potentially harmful capabilities aligns with ongoing discussions around AI ethics and regulation.
OpenAI's decision to withhold its deep research model from the API underscores a critical awareness of AI's ethical implications, particularly in the context of misinformation. This highlights the need for ongoing evaluation and improvement in AI technologies before their public deployment.