OpenAI is providing broader access to its new text-to-image generator, DALL-E 3, by making it available on ChatGPT for Plus and Enterprise customers. This expands on Microsoft’s recent integration into Bing.
OpenAI says it has implemented more rigorous security mitigations to address issues that have plagued AI generative image models. These include publishing copyrighted content, problematic social biases, and misrepresentations.
The company cites research on its website about the enhanced security capabilities of DALL-E 3. This includes limiting imitation of the styles of living artists, reducing inappropriate depictions of public figures, and improving representation in generated images.
Provenance Detection Tool Can Identify DALL-E 3 Images
Additionally, OpenAI has developed an internal “provenance classifier” tool that it claims is 99% accurate in detecting images created by DALL-E 3. This could help identify misuse of your online content .
Previous text-to-image conversion models, such as DALL-E 2, Midjourney, and Stable Diffusion, have faced backlash for harmful image errors despite attempts at safeguarding. Users have found ways to get around blocking words and phrases.
Larger, more cautious release follows Microsoft integration
This expanded but limited release for select users follows Microsoft’s rapid addition of DALL-E 3 to Bing search and imaging. Despite the announced protections, some worrying images were leaked.
OpenAI appears to be taking a more cautious approach by restricting access initially. However, the real test will come in monitoring the security of the technology on a larger scale.
The risks of generative AI remain very difficult to eliminate. But OpenAI’s commitments and transparency around security research are welcome steps amid the rollout. The key to this powerful technology is constant vigilance, not complacency.
Subscribe to our latest newsletter
To read our exclusive content, register now. $5/Monthly, $50/Yearly
Categories: Technology
Source: vtt.edu.vn