The trip has helped us higher understand the perspectives of users, developers, and government leaders all over the world. With their feedback in mind, we’re placing additional deal with these areas:

Making our products more useful, impactful, and accessible.
The trip clarified our sense of what it takes for our products to be accessible and useful for users and developers all over the world. We’re working on changes to make it easier for people to guide our models toward responses that reflect a greater variety of individual needs and native cultures and contexts. We’re also working toward higher performance for languages aside from English, considering not only lab benchmarks, but in addition how accurately and efficiently our models perform within the real-world deployment scenarios that matter most to our developers. And we’re committed to continuing to make our pricing structure accessible for developers all over the world.

Further developing best practices for governing highly capable foundation models.
As the general public debate over recent AI laws and regulations continues, we’ll redouble our efforts to pilot and refine concrete governance practices specifically tailored to highly capable foundation models just like the ones that we produce. This includes critical safety work streams akin to pre-deployment safety evaluation and adversarial testing, and recent efforts to empower people to trace the provenance of AI-generated content. Such measures will, we imagine, be essential components of a governance ecosystem for AI, alongside long-established laws and sector-specific policies for some essential applications. We may also proceed to speculate in piloting broad-based public input approaches into our deployment decisions, adding in localization features to our systems, and fostering a global research community to expand and strengthen the evaluation of model capabilities and risks, including via external research on our AI systems and our cybersecurity grants program.

Working to unlock AI’s advantages.
We can be expanding our efforts to support broad AI literacy, which we heard is a priority for a lot of communities, in addition to investing in ways for creators, publishers and content producers to learn from these recent technologies so we are able to proceed to have a healthy digital ecosystem. Moreover, we’re constructing teams that may provide more support to organizations which might be exploring how you can use our tools for broadly helpful applications, and conducting research into and policy recommendations for the social and economic implications of the systems we construct.

We may have more to say within the weeks and months ahead in each of those areas. A warm thanks to everyone all over the world who shared their perspectives and experiences with us.