See how companies are responsibly integrating AI into production environments. This invitation-only event in SF explores the intersection of technology and business.Find out how to participate here.
OpenAI strives to live up to the „open“ in its name.
While the company hasn't open sourced the new model, it did provide details this week about how it will do so. Approaching AI and technology problems (e.g. disinformation/deepfakes); Future Plans.
It was announced today that “model specsThis is a framework document designed to shape the behavior of AI models used within the OpenAI Application Programming Interface (API) and ChatGPT. Click here for web formopen until May 22nd.
As Sam Altman, Co-Founder and CEO of OpenAI I posted about it on X: „We'll take the time to listen, discuss, and adapt, but I think it would be very helpful to clarify whether something is a bug or a decision.“
Why does OpenAI release model specifications?
OpenAI says the release of this working document is part of its broader mission to ensure that AI technology operates in a way that is beneficial and safe for all users.
Of course, this is easier said than done, and doing so quickly falls into the realm of long-unresolved philosophical debates about technology, intelligent systems, computing, tools, and society more generally. It will be.
As OpenAI writes in its article, Blog post announcing model specifications:
“Even though a model is intended to be broadly informative and helpful to users, these intentions may conflict in practice. For example, a security company may use , you may want to generate phishing emails as synthetic data, but this same functionality can be harmful when used by fraudsters.
By sharing the first draft, OpenAI hopes to enable the public to engage in deeper conversations about the ethical and practical considerations involved in AI development.Users can submit comments via her OpenAI Model spec feedback form on website Next two weeks.
OpenAI then said that over the „next year,“ it will „share updates on changes to model specifications, responses to feedback, and progress on the research that shapes model behavior.“
In today's blog post announcing the Model Specification, OpenAI explains how exactly the Model Specification affects the behavior of AI models, and how some of the principles written in the Model Specification are called „system prompts“ or „preferences.“ It has not been disclosed whether this will be included in „Prompt“. It is used to fine-tune AI systems before making them available to the public, but it is safe to assume that they have a significant impact on AI systems.
In some ways, the model specs seem similar to its rivals. Humane AI’s “Constitutional” Approach to AI Developmentinitially a major differentiator, but the latter company did not widely emphasize this for some time.
AI behavior framework
Model specifications are organized around three main components: purpose, rules, and default behavior. These elements serve as the backbone that guides the interactions between AI models and human users, ensuring that AI models are not only effective but also compliant with ethical standards.
- the purpose: This document sets out broad and comprehensive principles aimed at assisting both developers and end users. These include maintaining his OpenAI commitment to helping users achieve their goals efficiently, considering the potential impact on various stakeholders, and reflecting positively on the community. This includes:
- rule: Model specifications establish clear rules to navigate the complexities of AI interactions. These require compliance with applicable laws, respect for intellectual property, protection of privacy, and strict prohibition on the creation of Not Safe for Work (NSFW) content.
- Default behavior: The guidelines emphasize the importance of assuming good intentions, asking clarifying questions when necessary, and cooperating as much as possible without overreaching. These defaults are designed to help balance the different needs of different users and use cases.
AI influencers and people like University of Pennsylvania Wharton Business School professor Ethan Mollick liken AI to a fictional world. „Three Laws of Robotics“ devised by science fiction writer Isaac Asimov It dates back to 1942.
Some people object to the current implementation of how OpenAI's model specification makes ChatGPT or other AI models work.asAuthor Andrew Curran points out about XOne example from OpenAI in the model specification shows a virtual „AI assistant“ retracting and not disputing a user's false claim that the Earth is flat.
Continuing efforts and development
OpenAI recognizes that model specifications are evolving documents. It is a dynamic framework that not only reflects the organization's current practices but also adapts based on ongoing research and community feedback.
This consultative approach aims to gather diverse perspectives from global stakeholders, including policy makers, trusted institutions, and sector experts, among others.
The feedback we receive plays an important role in refining our model specifications and shaping future AI model development.
OpenAI plans to keep the public updated on changes and insights gained from this feedback loop, reinforcing its commitment to responsible AI development.
Where do we go from here?
By clearly defining how AI models should behave in our model specifications and by continually seeking input from our global community, OpenAI is committed to improving AI, even in the face of lawsuits and litigation. We aim to foster an environment where people can thrive as positive forces in society. Criticism of training an artist's work without explicit consent.