Current AI scaling laws are showing diminishing returns, forcing AI labs to change course | TechCrunch
Briefly

AI investors and CEOs note that traditional scaling methods are yielding diminishing returns for model capabilities, prompting a shift towards exploring new strategies for enhanced AI development.
OpenAI's Ilya Sutskever highlighted a universal search for novel methods to improve AI scalability, as there is a recognition that just more data and compute isn't enough.
Microsoft CEO Satya Nadella emphasized the emergence of test-time compute as a promising avenue, addressing the ceiling AI models are hitting with current scaling laws.
Investors and leaders are now coining a 'second era of scaling laws' characterized by test-time scaling, moving away from earlier assumptions about model capabilities.
Read at TechCrunch
[
|
]