The Evolution of AI: From Cloud to Pocket
Introduction
When ChatGPT was launched in November 2023, it was only accessible via the cloud-rendering-service-for-faster-easier-projects/” title=”AWS Launches Cloud Rendering Service for Faster, Easier Projects”>cloud due to its massive underlying model. Today, a similarly capable AI program can run on a MacBook Air without even warming it up. This significant reduction in size demonstrates the rapid advancements in AI technology.
The Shrinking AI Models
The Phi family of AI models highlights the potential for modern AI to be improved through selective training. Sébastien Bubeck, a researcher at Microsoft, explains that these models were designed to test whether focusing on specific training data could enhance an AI system’s capabilities.
Training Large Language Models
Large language models like OpenAI’s GPT-4 or Google’s Gemini, which power chatbots and other services, are typically trained on vast amounts of text from books, websites, and other sources. Despite legal concerns, increasing the volume of text and computational power used for training has unlocked new capabilities for these models.
A New Approach to AI Development
Bubeck suggests that instead of competing with OpenAI and Google by building ever-larger cloud AI models, there might be a different approach. By concentrating on reducing the size of AI to fit into portable devices, companies could offer powerful AI solutions that are more accessible to users.
Rather than go toe-to-toe with OpenAI and Google in building ever more enormous cloud AI models, it might think different by focusing on shrinking AI down to fit into its customers’ pockets.
Conclusion
The journey from cloud-based AI to portable, efficient models marks a significant milestone in AI development. As researchers like Bubeck continue to explore selective training and model reduction, the future of AI looks increasingly promising and accessible.
Will Knight via Microsoft
6 Comments
Could pocket-sized AI models mean the end of privacy?
Pocket-sized AI models might seem convenient, but are we just ignoring the potential ethical dilemmas?
The idea of pocket-sized AI models sounds like sci-fi turned reality!
Pocket-sized AI models: just a marketing buzzword or game-changer?
Trailblazing or ticking time bomb, pocket-sized AI models: what say you?
Pocket-sized AI models could redefine portable computing, but with what trade-offs?