Microsoft Executive Teases Innovative LLMs Surpassing OpenAI’s

Microsoft Executive Teases Innovative LLMs Surpassing OpenAI’s

Sign up for our daily and weekly newsletters to stay updated with the latest in AI and get exclusive content.

In a recent interview, Eric Boyd, the Microsoft exec overseeing the company’s AI platform, hinted that Microsoft’s AI services will soon offer additional large language models (LLMs) beyond the ones provided by OpenAI. This move aims to give customers more options. In his conversation with VentureBeat, the focus was on how ready enterprise companies are to adopt AI. Boyd’s comments came after Amazon AWS CEO Adam Selipsky suggested that companies don’t want to be tied to a single model provider.

When asked if Microsoft plans to offer more models outside of OpenAI, Boyd responded, “Stay tuned. There are definitely some things in the works.”

Microsoft has integrated OpenAI’s models across various products, including Bing, GitHub Copilot, and the Office CoPilots. Additionally, through its Azure Machine Learning platform, Microsoft provides access to other models, like those from Hugging Face. Although closed-source models like those from OpenAI tend to be more straightforward and come with better support, Amazon has been vocal about offering more choices, including models from Anthropic and Stability AI, among others.

Boyd emphasized Microsoft’s commitment to offering a variety of choices and ensuring their generative AI applications are safe to use. Companies that focus on areas where these models excel, such as text generation, can quickly progress.

Boyd discussed the readiness of companies for AI, mentioning that Microsoft has over 18,000 customers using their AI services, which cover industries from healthcare to finance and beyond. He also acknowledged that while OpenAI faced recent challenges, Microsoft remains confident in their partnership and continues to offer a broad selection of models, including open-source ones like Llama 2.

Regarding enterprise adoption of AI, Boyd noted that companies with a clear vision for specific applications, such as content creation, summarizing user reviews, or generating code, see the most success. Those with unrealistic expectations need to be guided towards practical uses of AI.

When asked about competitors like Amazon and Google and what sets Microsoft apart, Boyd highlighted the company’s early lead in the AI space and the robust set of tools they offer. Microsoft has integrated learnings from customer experiences into their products, making it easier for users to build their applications responsibly.

On the topic of safety, Boyd mentioned Microsoft’s collaborative approach with OpenAI in developing and deploying AI models and their efforts to reassure customers about the security and governance of these models. Microsoft’s partnerships with other companies, including Meta and NVIDIA, ensure a wide range of model choices for customers.

Boyd also addressed the issue of AI models occasionally delivering incorrect information, noting that Microsoft has developed techniques for fine-tuning and steering models to give better responses. They emphasize the co-pilot metaphor, where the model assists but doesn’t replace human judgment.

On the matter of explainability and interpretability of AI models, Boyd acknowledged it as a vital area of ongoing research. He expressed confidence in the advancements being made but recognized the complexity involved due to the models’ scale.

Boyd assured that Microsoft remains committed to both closed-source and open-source models, contributing to the development and optimization of the latter as well. He highlighted Microsoft’s long-standing focus on responsible AI practices, transparency, and governance, which are essential for customer confidence.

Addressing the company’s governance frameworks, Boyd explained that Microsoft works closely with enterprises to align with their standards while consistently adhering to their responsible AI principles. Boyd also gave examples of how Microsoft collaborates with industries like healthcare to meet regulatory standards.

Finally, Boyd touched on the potential for integrating reasoning capabilities into AI models and the ongoing research in combining modalities like vision and text. He was optimistic about the future but cautious about making specific predictions given the rapid pace of advancements.

Thank you, Eric Boyd, for the insightful discussion on Microsoft’s AI initiatives and future plans. Stay tuned for more updates as we continue to cover this exciting journey.