OPT-IML vs ChatGPT Plugins

The paper titled "OPT-IML: Scaling Language Model Instruction Meta Learning through the Lens of Generalization" focuses on fine-tuning large pre-trained language models with a technique called instruction-tuning, which has been demonstrated to improve model performance on zero and few-shot generalization to unseen tasks. The main challenge addressed in the study is grasping the performance trade-offs due to different decisions made during instruction-tuning, such as task sampling strategies and fine-tuning objectives.

The authors introduce the OPT-IML Bench—a comprehensive benchmark comprising 2000 NLP tasks from 8 different benchmarks—and use it to evaluate the instruction tuning on OPT models of varying sizes. The resulting instruction-tuned models, OPT-IML 30B and 175B, exhibit significant improvements over vanilla OPT and are competitive with specialized models, further inspiring the release of the OPT-IML Bench framework for broader research use.

OpenAI follows an iterative deployment philosophy, and as part of this approach, it is gradually releasing plugins for ChatGPT. The purpose of this gradual release is to study the real-world use of the plugins, assess their impact, and identify any safety and alignment challenges that may arise. Getting these aspects right is essential for OpenAI to achieve its mission.

Since the launch of ChatGPT, users have expressed interest in having access to plugins, and many developers are experimenting with similar ideas. OpenAI is initially rolling out plugins to a small set of users and plans to gradually increase access as it learns more. After an alpha period, API users who want to integrate plugins into their products will also have access. OpenAI is excited to build a community that will help shape the future of the human-AI interaction paradigm.

Developers who have been invited off the waitlist can use OpenAI's documentation to build a plugin for ChatGPT. Once the plugin is created, it is listed in the prompt shown to the language model, along with documentation that instructs the model on how to use each one. The first set of plugins has been created by Expedia, FiscalNote, Instacart, KAYAK, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier.