Viewpoint
Efficiency gains will not eliminate the need for infrastructure
Efficiency gains will not eliminate the need for infrastructure
Altman expects better chips and algorithms to make AI much more efficient but argues that rapidly increasing demand for intelligence will still require substantially more infrastructure if AI is to remain inexpensive and abundant.
- Speaker
- Sam Altman
- Source timestamp
- 24:29
More from this interview
- Deep learning convinced Altman that advanced AI would become possible
- GPT-4 marked a turning point in expectations for AI
- AI could drive an unprecedented boom in entrepreneurship
- Persistent agents could become ongoing virtual collaborators
- Countries cannot afford to ignore AI adoption
- OpenAI should provide intelligence while users supply context
- AI belongs to a continuous exponential of technological progress
- Societies must adopt AI to build what comes next
- OpenAI aims for a pragmatic approach between AI extremes
- AI should preserve human agency and sovereign choice
- OpenAI does not want to capture all AI value
- AI can radically compress the startup development cycle
- Rejecting AI could resemble rejecting electricity
- AI may become an invisible part of everyday products
- Cybersecurity failures could significantly slow AI adoption
- Concentrated access to AI could worsen inequality
- AI demand could increase by orders of magnitude
- Abundant AI could democratize access to valuable services
- Concern about AI risks can improve technological safety
- Human ingenuity can mitigate many AI downsides
- AI benefits justify continued efforts to make it safer