DALL·E 2 pre-training mitigations

OpenAI Blog2022-06-28 15:000 views

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 prevent generated images from violating our content policy.

Source：https://openai.com/index/dall-e-2-pre-training-mitigations

My favorites

PreviousDALL·E 2: Extending creativity NextLearning to play Minecraft with Video PreTraining

Elastic named a Leader in the IDC MarketScape: Worldwide General-Purpose Knowledge Discovery 2025 Vendor Assessment

Elastic was named a Leader in the IDC MarketScape: Worldwide General-Purpose Knowledge Discovery 2025. Our innovation un

0Elastic Blog

There's no faster laptop under $700 than this $560 Core i9 monster with a CPU that rivals the M5 in the $1,099 Apple MacBook Air

Black Friday is nearly here, and we’ve seen several great early deals, including on laptops. One of the best bargains I’

0TechRadar

Perplexity brings AI agents feature called Teammates to the Mac

This week, Perplexity has deployed its new proactive agentic AI feature called Teammates to the Mac. Perplexity Teammate

09to5Mac

PHP 8.6.0RC7 available for testing

0PHP News

Guests, pets, and what a room-level presence display cannot tell you

The fastest way to make a presence product dishonest is to start naming who is in the room. We get asked about guests, p

0dev.to

How a GPT‑Driven Loop Made AI Videos Physically Accurate – 30% Better!

PhysLoop – When Large‑Language Models Start Teaching Videos to Respect Physics Hook A 30 % jump in physical consistency