Greg Brockman believes that as software creation becomes cheap and accessible, organizational value shifts to software reliability, resilience, and security. Consequently, he suggests security spending should vastly exceed traditional software spending to address long-standing vulnerabilities and supply chain risks across internet infrastructure.
OpenAI proposes utilizing adversarial AI debate to maintain human oversight over outputs that exceed human checking capacity. Having a secondary model identify missing context or counter-arguments allows humans to evaluate complex findings, such as research summaries or advanced proofs, without having to digest all underlying material.
Brockman observes that deep learning advances rapidly in mathematics and theoretical physics due to massive quantities of available data. In contrast, biological research and drug discovery remain constrained because physical experimental data is scarce, requiring wet lab advancements and simulated data to unblock progress.
OpenAI halted projects for 25 percent of its production engineers to direct their peak intelligence model against internal systems until vulnerability discovery saturated. Brockman recommends that organizations systematically repeat this process with each new frontier release to pay down accumulated technical debt before attackers exploit it.
Brockman anticipates that user interaction will transition from static screens and text boxes to dynamically generated UI created on the fly. Models will craft bespoke visual interfaces tailored to momentary tasks, giving users the precise controls needed for complex activities without requiring pre-built applications.
OpenAI sees consumer and enterprise assistants converging on identical core models rather than distinct architectures. The primary differentiator is cost and compute allocation: enterprise users justify heavy compute expenditures for peak intelligence, whereas mass-market consumer tools require substantial engineering optimization to sustain free or low-cost tiers.
Rather than awaiting reactive text queries, autonomous workplace tools like OpenAI's Dots run continuously across email, Slack, and calendars. Brockman notes that these models operate on their own cloud computers to resolve schedule conflicts and proactively surface unprompted priorities while users are away from work.
Even as AI models surpass people across specialized domains, Brockman argues humans must maintain ultimate authority over whether outcomes improve human life. He envisions human agency shifting toward a board member or CEO role, guiding high-capability expert machines according to human values and desired goals.
