Microsoft’s Legal Safeguards, Azure AI Studio’s Empowerment, and OpenAI’s GPT-4 Turbo Breakthroughs
In the rapidly evolving world of artificial intelligence, Microsoft and OpenAI are at the forefront, pushing the boundaries and shaping the future of AI applications. From legal safeguards for generative AI users to empowering developers with AI copilots and the latest breakthroughs in GPT-4 Turbo, these developments are transforming the AI landscape. Let’s delve deeper into each of these groundbreaking stories.
Microsoft’s recent announcement at Ignite showcases the company’s commitment to protecting its commercial customers from copyright infringement lawsuits related to generative AI, specifically the Azure OpenAI Service. This fully managed service incorporates OpenAI models like ChatGPT and DALL-E 3, adding governance layers to ensure responsible AI usage.
Generative AI models, such as ChatGPT and DALL-E 3, are trained on vast datasets comprising e-books, art pieces, emails, songs, and more. Microsoft’s expanded policy promises to defend and compensate customers for any adverse judgments arising from copyright infringement lawsuits tied to the use of Azure OpenAI Service or its generated outputs. However, this protection comes with a set of conditions. To be eligible for these safeguards, subscribers must implement “technical measures” and adhere to specific documentation to mitigate the risk of generating infringing content using OpenAI’s models. While Microsoft declined to provide specifics on these measures, it signals a proactive approach to responsible AI usage.
Moreover, it remains unclear whether these protections extend to Azure OpenAI Service products in preview, such as GPT-4 Turbo with Vision. Microsoft clarified that the policy applies to all products in paid preview and encompasses Microsoft’s, but not a customer’s, training data. This move aligns with Microsoft’s September announcement to cover legal damages for customers using certain AI products if they face copyright infringement lawsuits. Customers must use the built-in “guardrails and content filters” to retain coverage, indicating a broader industry trend towards AI accountability.
Interestingly, OpenAI recently announced a similar initiative, pledging to cover legal costs for customers facing lawsuits over IP claims related to work generated by OpenAI tools. Microsoft’s expanded policy for Azure OpenAI Service seems to be an extension of this shared commitment to user protection.
At Microsoft’s annual Build conference, the company unveiled Azure AI Studio, a revolutionary capability within the Azure OpenAI Service. This empowers developers to combine OpenAI’s models, like GPT-4 and ChatGPT, with their own data—whether text or images—to build personalized AI copilots. A “copilot” in this context refers to a chatbot app that utilizes AI, typically text or image-generating AI, to assist with various tasks, such as writing a sales pitch or generating images for presentations. While Microsoft has previously created copilot apps like Bing Chat, Azure AI Studio takes it a step further by allowing developers to ground these AI models on their data securely.
The process within Azure AI Studio involves selecting a generative AI model, providing a “meta-prompt” or a base description of the copilot’s role, and integrating cloud-based storage for context-aware conversations. The platform supports the addition of plug-ins to extend copilots’ functionality, giving them access to third-party data and services. Azure AI Studio addresses a crucial need for developers to leverage OpenAI’s models on their data while adhering to organizational policies and access rights. This move is particularly significant as Microsoft reports serving over 4,500 companies, including notable names like Coursera, Grammarly, Volvo, and IKEA, through the Azure OpenAI Service.
Responding to customer concerns about potential intellectual property (IP) infringement claims associated with generative AI outputs, Microsoft introduced the Copilot Copyright Commitment. This commitment extends Microsoft’s existing intellectual property indemnification support to commercial Copilot services, ensuring that the company will defend customers and cover any adverse judgments or settlements resulting from copyright infringement lawsuits.
This new commitment comes with specific conditions, emphasizing the responsible use of Copilot services. Customers must use the built-in guardrails and content filters to reduce the likelihood of generating infringing content. While Microsoft charges commercial customers for Copilots, the commitment reflects a philosophy of standing behind users when legal issues arise, aligning with the company’s two-decade history of defending customers against patent claims.
Microsoft acknowledges the sensitivity of authors’ concerns and believes that the responsibility to address them lies with the company rather than the customers. By incorporating filters and technologies to reduce the likelihood of Copilots returning infringing content, Microsoft aims to strike a balance between AI advancement and the protection of creators’ rights.
The Copilot Copyright Commitment covers paid versions of Microsoft commercial Copilot services, including Microsoft 365 Copilot and GitHub Copilot. The commitment ensures that customers can use Copilots and the output they generate without worrying about copyright claims, provided they adhere to the specified conditions.
At its first-ever developer conference, OpenAI unveiled GPT-4 Turbo, an enhanced version of its flagship text-generating AI model, GPT-4. Promising to be both “more powerful” and less expensive, GPT-4 Turbo comes in two versions: one strictly for text analysis and a second version that understands the context of both text and images. OpenAI adopts a pricing strategy of $0.01 per 1,000 input tokens and $0.03 per 1,000 output tokens, offering a 3x reduction in input token prices and a 2x reduction in output token prices compared to GPT-4. The text-analyzing model is available in preview via an API, with both versions set to become generally available in the coming weeks. GPT-4 Turbo introduces several improvements over its predecessor. It boasts a more recent knowledge base, with a cut-off date in April 2023 compared to GPT-4’s knowledge cut-off in September 2021. This ensures more accurate responses to questions about recent events.
The model also features an expanded context window, measuring 128,000 tokens—four times the size of GPT-4’s and the largest among commercially available models. A larger context window allows the model to retain more information, avoiding the issue of “forgetting” recent conversations.
Additionally, GPT-4 Turbo introduces a new “JSON mode,” ensuring valid JSON responses. This is particularly useful for web apps transmitting data, maintaining consistency in data interchange formats. OpenAI’s commitment to providing a performant model is evident in its focus on tasks requiring careful instruction following and improved function parameter returns. OpenAI recognizes the challenges in fine-tuning GPT-4 and is launching an experimental access program with increased oversight and guidance due to technical complexities. Despite the challenges, OpenAI is doubling the tokens-per-minute rate limit for paying GPT-4 customers, offering more flexibility for developers.
In conclusion, these recent developments from Microsoft and OpenAI highlight the industry’s commitment to responsible AI development, user protection, and innovation. From legal safeguards to customizable AI copilots and breakthroughs in text-generating models, these initiatives contribute to a broader conversation about the delicate balance between technological innovation and ethical considerations.
As we navigate the intricate terrain of AI, it’s imperative to recognize that these initiatives are part of a broader journey. The future promises further collaboration, refinement of best practices, and the continuous evolution of AI technologies. Industry leaders, developers, and users alike are integral participants in shaping an AI landscape that not only thrives on innovation but also upholds ethical standards. In closing, the stories of Microsoft’s legal safeguards, Azure AI Studio’s empowerment, and OpenAI’s GPT-4 Turbo breakthroughs converge to paint a picture of a dynamic and responsible AI future. This is a future where the transformative power of AI is harnessed with a keen awareness of ethical considerations, ensuring that as AI progresses, it does so in harmony with the values that underpin a conscientious technological era. For all my daily news and tips on AI, Emerging technologies at the intersection of humans, just sign up for my FREE newsletter at www.robotpigeon.beehiiv.com