From: redpointai
The advent of AI introduces a shift in what is considered a scarce factor in society. Historically, intelligence has been the critical scarcity, but in a future with ubiquitous and free AI, agency is likely to become the scarce factor of production [00:49:16]. This refers to knowing the right questions to ask and which projects to pursue, problems that AI is not expected to solve easily for humans [00:49:44]. Humans will need to develop this kind of agency to work effectively with AI [00:50:02].
AI’s Impact on Productivity and Work
Despite significant advancements, AI’s impact on productivity statistics has been minimal, primarily seen in capital expenditures for data centers rather than direct productivity gains [00:31:55]. This is because AI primarily automates tasks, whereas a job is often composed of many tasks, some of which are difficult to automate [00:33:22]. For example, in programming, AI first optimizes “boilerplate” code, leaving the more abstract “giving direction” part as the last hurdle [00:33:36].
Underexplored Applications
A promising but underexplored application for AI is in “boring” tasks that require infinite patience rather than infinite intelligence [00:34:06]. For instance, an AI could tirelessly compare shopping prices for procurement, a task that smart humans find mundane [00:34:22]. This concept is likened to the role of consultants, who get smart people to work on dull problems [00:35:18]. Initial productivity studies show AI significantly benefits lower-performing individuals, enabling them to produce code or solutions when they understand the goal but lack the specific skills to execute [00:35:55]. This is seen as a hopeful development [00:36:34].
Implications for Social Sciences and Policy
AI is expected to transform social sciences research and policymaking [00:54:47]. In business, product management often functions as an experimental social science, using methods like A/B testing to understand human reactions [00:55:20]. AI could enable the creation of “fake users” that react like real ones, allowing A/B tests to be conducted without going to production, and even deep interviews with these simulated users [00:55:40]. The general principle is to explore replacing any task one would ask a human to do with AI, especially tasks that could be done a hundred times by an AI but only painfully once by a human [00:56:14].
The Nature of AGI and Future Progress
The concept of AGI is challenging to define, and there may not be a single “moment” of its achievement [00:47:06]. Instead, it is seen as a fractal progression where more and more tasks become automated [00:47:17]. It’s predicted that the arrival of AGI might feel “banal,” with people still experiencing mundane daily life despite self-driving cars and AI armies [00:47:25].
Solving reasoning was considered the last fundamental challenge needed to scale to human-level intelligence, alongside pre-training and multi-modality [00:47:59]. Now, the primary remaining challenge is scaling, which is a complex systems, hardware, optimization, and data problem [00:48:17]. This scaling is hard, but in some ways, the progression towards a more integrated and capable AI future is seen as “predestined” [00:48:47].