Sam Altman & Reid Hoffman on AI’s Future: Business, Multimodal Models, Society

1. Commercial Opportunities of Large Models

Reid Hoffman: Many large models are offered via APIs—what commercial opportunities can they create?

Sam Altman: Language models are already valuable for copywriting and education, and in a few years they will be powerful enough to rival Google’s trillion‑dollar search business. Their adoption will reshape everyday life.

AI tools such as Stability.ai’s open‑source Stable Diffusion, Microsoft‑OpenAI Copilot, OpenAI’s GPT‑3, DALL‑E and the popular ChatGPT demonstrate how AI can turn ideas into reality and become mature, widely used products.

Altman: In five years we will likely stop using Transformer models, as newer architectures will replace them.

2. The “AI+” Era: AI for Science and the Metaverse

Reid Hoffman: What role can AI models play in scientific research?

Altman: AI is used in two ways: directly as scientific tools (e.g., AlphaFold for protein‑structure prediction) and as productivity enhancers that help scientists discover new directions or write code (e.g., Copilot). Both will accelerate scientific progress.

He also warned that self‑improving AI scientists carry both promise and risk, especially regarding alignment and unintended code changes.

Reid Hoffman: How will AI and the metaverse intersect?

Altman: The metaverse will become a new software container, like a smartphone, while AI is the underlying technological revolution. The key question is “How will the metaverse integrate into the new AI world?” rather than the reverse.

3. Future Directions of AI

Reid Hoffman: What are the main directions for AI in the coming years?

Altman: First, language models will continue to outpace current expectations. Even though compute and data are bottlenecks, algorithmic improvements still hold huge potential.

Second, multimodal models will evolve beyond text‑image conversion to seamless transformation across all modalities.

Third, models will acquire continuous‑learning capabilities, moving away from static, once‑trained states.

If these three trends materialize, they will unlock countless new applications and drive a genuine technological leap.

Altman also emphasized that AI will dramatically lower the marginal cost of intelligence and energy, reshaping the cost structure of all industries. He warned that while costs approach zero, they will never be truly zero, and those who can invest heavily in compute will gain unprecedented power.

Regarding societal impact, Altman highlighted the need for new social contracts to ensure fair wealth distribution, universal access to AGI, and governance of AGI’s capabilities.

He noted OpenAI’s ongoing universal basic income (UBI) experiment, which aims to provide every U.S. citizen with $13,500 per year once AI‑driven productivity creates sufficient wealth.

Finally, Altman defined AGI as a system that can collaborate with humans on any task—learning new domains like medicine or coding as needed—emphasizing that AGI’s value lies in its meta‑learning ability rather than any single skill.

He concluded that deep human motivations—social interaction, curiosity, competition—will remain unchanged for millennia, even as AI permeates every aspect of life.