Articles Collectés

873

Total

864

Non traités

9

Traités

Non traitée 24/08/2022 09:00
Our approach to alignment research

RSS: OpenAI News

We are improving our AI systems’ ability to learn from human feedback and to assist humans at evaluating AI. Our goal is to build a sufficiently aligned AI system that can help us solve all other al...

Non traitée 10/08/2022 09:00
New and improved content moderation tooling

RSS: OpenAI News

We are introducing a new and improved content moderation tool. The Moderation endpoint improves upon our previous content filter, and is available for free today to OpenAI API developers.

Non traitée 28/07/2022 09:00
Non traitée 25/07/2022 09:00
Non traitée 20/07/2022 09:00
DALL·E now available in beta

RSS: OpenAI News

We’ll invite 1 million people from our waitlist over the coming weeks. Users can create with DALL·E using free credits that refill every month, and buy additional credits in 115-generation incremen...

Non traitée 18/07/2022 09:00
Reducing bias and improving safety in DALL·E 2

RSS: OpenAI News

Today, we are implementing a new technique so that DALL·E generates images of people that more accurately reflect the diversity of the world’s population.

Non traitée 14/07/2022 09:00
DALL·E 2: Extending creativity

RSS: OpenAI News

As part of our DALL·E 2 research preview, more than 3,000 artists from more than 118 countries have incorporated DALL·E into their creative workflows. The artists in our early access group have help...

Non traitée 28/06/2022 09:00
DALL·E 2 pre-training mitigations

RSS: OpenAI News

In order to share the magic of DALL·E 2 with a broad audience, we needed to reduce the risks associated with powerful image generation models. To this end, we put various guardrails in place to p...

Non traitée 23/06/2022 09:00
Learning to play Minecraft with Video PreTraining

RSS: OpenAI News

We trained a neural network to play Minecraft by Video PreTraining (VPT) on a massive unlabeled video dataset of human Minecraft play, while using only a small amount of labeled contractor data. With ...

Non traitée 17/06/2022 09:00
Non traitée 13/06/2022 09:00
AI-written critiques help humans notice flaws

RSS: OpenAI News

We trained “critique-writing” models to describe flaws in summaries. Human evaluators find flaws in summaries much more often when shown our model’s critiques. Larger models are better at self-c...

Non traitée 09/06/2022 09:00
Techniques for training large neural networks

RSS: OpenAI News

Large neural networks are at the core of many recent advances in AI, but training them is a difficult engineering and research challenge which requires orchestrating a cluster of GPUs to perform a sin...

Non traitée 02/06/2022 09:00
Best practices for deploying language models

RSS: OpenAI News

Cohere, OpenAI, and AI21 Labs have developed a preliminary set of best practices applicable to any organization developing or deploying large language models.

Non traitée 28/05/2022 09:00
Non traitée 24/05/2022 09:00
Powering next generation applications with OpenAI Codex

RSS: OpenAI News

Codex is now powering 70 different applications across a variety of use cases through the OpenAI API.

Non traitée 18/05/2022 09:00
DALL·E 2 research preview update

RSS: OpenAI News

Early users have created over 3 million images to date and helped us improve our safety processes. We’re excited to begin adding up to 1,000 new users from our waitlist each week. 

Non traitée 05/05/2022 09:00
OpenAI leadership team update

RSS: OpenAI News

We’re happy to announce several executive role changes that reflect our recent progress and will ensure continued momentum toward our next major milestones.

Non traitée 13/04/2022 09:00
Non traitée 13/04/2022 09:00
Measuring Goodhart’s law

RSS: OpenAI News

Goodhart’s law famously says: “When a measure becomes a target, it ceases to be a good measure.” Although originally from economics, it’s something we have to grapple with at OpenAI when figu...

Non traitée 15/03/2022 08:00
New GPT-3 capabilities: Edit & insert

RSS: OpenAI News

We’ve released new versions of GPT-3 and Codex which can edit or insert content into existing text, rather than just completing existing text.