Microsoft CTO: AI Progress to Continue Despite Skepticism, Powered by Scaling Laws

Microsoft CTO Kevin Scott believes AI progress will continue despite skepticism, arguing that large language model (LLM) “scaling laws” will drive breakthroughs as models get larger and have access to more computing power.

Scaling laws and AI progress: Scott maintains that scaling up model size and training data can lead to significant AI improvements, countering critics who argue that progress has plateaued around GPT-4 class models:

• He acknowledges the challenge of infrequent data points in the field, as new models often take years to develop, but expresses confidence that future iterations will show improvements, particularly in areas where current models struggle.
• Scott suggests that perceptions of slowing progress may be due to the rapid onset of AI in the public eye, when LLMs have actually been developing for years prior to the launch of GPT-3 and GPT-4.

Betting on continued breakthroughs: Scott’s stance indicates that tech giants like Microsoft still feel justified in heavily investing in larger AI models, expecting continued progress rather than hitting a capability plateau:

• Given Microsoft’s $13 billion technology-sharing deal with OpenAI and strong marketing of its own AI features, the company has a strong interest in maintaining the perception of continued progress.
• Some critics argue that Microsoft and OpenAI may be defending continued investment into generative AI by suggesting they have secret, breakthrough technology that will “break the bones of every hater.”

Broader context: While Scott remains optimistic about the future of AI progress, the debate over scaling laws and the potential for LLMs to plateau highlights the uncertainty and conflicting perspectives within the AI community. As companies like Microsoft and OpenAI continue to invest heavily in larger models, only time will tell if their bets pay off with significant leaps in AI capabilities or if critics’ concerns about diminishing returns prove accurate. The outcome of this debate will have significant implications for the future direction and pace of AI development.