OpenAI chose a specific path by focusing on building large language models and generative AI systems at an enormous scale. However, before this, the trend in the AI research community leaned towards creating smaller, efficient AI systems. Researchers explored how minimal datasets and computational resources could still produce powerful AI models. There were fascinating studies showing that just a few hundred images could train highly effective AI systems, or that AI systems could be trained and run on mobile devices with minimal hardware. In contrast, OpenAI's approach involves hundreds of thousands of computer chips to train a single model, consuming vast amounts of energy comparable to powering entire cities. If we separate AI progress from the scaling paradigm, it’s clear that innovation doesn’t have to come at such a high cost. Unfortunately, most AI experts today work for major corporations, creating a conflict of interest similar to if climate scientists were funded by oil and gas companies—prioritizing corporate benefits scientific integrity. Additionally, I’ve encountered Angus Hansen, who provides sharp insights into the increasingly strained economic relations between the U.S. and the UK. His work is compelling, yet it seems politicians fail to grasp the severity of the situation. This ties back to the broader conversation about AI consumers and creators, highlighting the importance of independent perspectives in shaping ethical advancements.
