“We’re not worried about compute anymore”: The future of AI models

Key takeaways

The AI landscape is shifting towards utilizing multiple specialized models rather than relying on a single dominant model.
Efficient infrastructure is crucial for hosting generative AI applications due to their high resource demands.
Open source models allow organizations to fine-tune AI applications according to their proprietary data while maintaining control and security.
Ethical considerations are becoming increasingly important in AI development, emphasizing the need for transparency and responsible data usage.
The future of AI will likely focus on the entire system's performance rather than just the models themselves, with a growing emphasis on integrated AI strategies.

Summary

The Evolution of AI Models

In this episode, the conversation highlights a shift in the AI landscape where the focus is moving away from monolithic models to a plethora of specialized models. Jamie Deguerre from Together AI articulates the need for a more diversified approach to AI, emphasizing the benefits of harnessing multiple models to achieve optimal performance based on specific needs. The discussion also touches on the continuous advancements in AI research, especially the innovations coming from open source contributions.

The Importance of Infrastructure

The guests discuss how as the variety of models increases, so does the significance of the infrastructure housing these models. Jamie explains that hosting a generative AI application is extremely resource-intensive compared to traditional web applications, leading to the conclusion that building efficient infrastructure is critical for service providers. Together AI aims to deliver superior performance and cost-effectiveness, driven by a research-oriented approach to systems optimization.

Custom Models and Proprietary Data

As companies seek to leverage AI tailored to their specific needs, the discussion pivots to the advantages of using open source models with proprietary data. This allows companies to maintain greater control and security over their data, as opposed to depending on closed-source alternatives. Jamie elaborates on the importance of fine-tuning large pre-trained models with internal data to enhance the models’ performance for unique use cases.

Ethics in AI

The conversation also delves into the ethical implications of AI development. Jamie discusses the ongoing dialogue within Together AI about ensuring ethical practices in AI deployment and model training. This includes concerns about data ownership and the responsibility of AI developers to respect the sources and contributions of data. Ensuring transparency in model development is highlighted as a crucial step towards fostering trust in AI technologies.

Looking Ahead

As the podcast wraps up, Jamie reflects on the future of AI and the expected surge in the widespread adoption of generative AI technologies. There is a sense of optimism about the innovations yet to come and the evolving role of systems in AI applications, which could redefine how businesses interact with AI technologies. Jamie expresses enthusiasm towards building systems that exceed mere model performance to encompass holistic AI strategies.

Show notes

Together AI is a platform for building with open-source and specialized multimodal models. Check out their docs.

Connect with Jamie on LinkedIn.

Shoutout to user aryaxt who earned a Stellar Question badge by wondering about MySQL Data - Best way to implement paging?.

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.