A case for AI-pragmatism – leverage cheap compute while it's here (quickthoughts.ca)

A recent discussion highlights the importance of AI-pragmatism in leveraging the current abundance of affordable computing power offered by major tech companies. As organizations like Microsoft, Google, and OpenAI push their AI services for profit, individuals can still find significant utility in these tools—especially with large language models (LLMs) available at low costs. A case in point is Microsoft’s GitHub Copilot, which allowed users to access vast computational resources for just $10 a month. However, this model is already evolving into a “pay per token” structure, indicating a shift towards monetization that users should anticipate. The significance of this trend lies in the window of opportunity it creates for developers and hobbyists to experiment and innovate without substantial initial investment. With the capabilities offered by these AI models, users can develop novel projects, like the local-first end-to-end fine-tuning project called Listenr, which focuses on building ASR models independently. While current competitive backgrounds make these resources economically favorable, it is evident that such opportunities won't last. This context urges the AI/ML community to capitalize on these offerings now, while remaining cognizant of the impending changes in pricing and service structures.