# OpenAI — Multimodal neurons in artificial neural networks

- Company: OpenAI (openai.com)
- Announced: 2021-03-04T08:00:00+00:00
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://openai.com/index/multimodal-neurons
- Record: https://forck.live/items/936-multimodal-neurons-in-artificial-neural-networks
- Subject: GPT / ChatGPT / API
- Models affected: CLIP

OpenAI discovered multimodal neurons in CLIP that respond to the same concept across literal, symbolic, and conceptual presentations, explaining its accuracy and providing insight into learned associations and biases.

## Evidence

Verbatim from https://openai.com/index/multimodal-neurons:

> We’ve discovered neurons in CLIP that respond to the same concept whether presented literally, symbolically, or conceptually.

---

Record: https://forck.live/items/936-multimodal-neurons-in-artificial-neural-networks
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md