A rational conversation on where AI is actually going | Benedict Evans

Benedict Evans discusses the profound impact of AI, comparing it to the internet and mobile revolutions. He emphasizes that we are in an early, uncertain phase similar to 1997 in the internet era, where much of what AI will change is still undefined. The conversation highlights the vast distribution of AI adoption, from tech insiders to everyday users, and the ongoing and unpredictable nature of AI development and application.

Evans explains that we are at a '1997 moment' for AI—early days with much potential and unknown outcomes. He stresses the importance of understanding what tasks and jobs are affected by AI, warning against overly simplifying assumptions about job automation. He describes the rapid capacity of AI to alter industries like accounting, legal services, and software development, and discusses the difficulty in predicting when full transformation will occur.

A surprising trend is discussed: AI labs and companies are heavily investing in consulting and professional services to implement AI solutions, contrary to expectations of cost-cutting. Evans argues that, as software becomes more commoditized, distribution channels and brand become crucial differentiators. He compares this to historical shifts in technology, where dominant players leverage distribution to maintain their position, shaping the future landscape of AI and application deployment.

The episode covers where value will accrue in the AI stack, emphasizing that foundational models face intense competition and may see margins squeezed. Conversely, application layers and integrated ecosystems are likely to generate more value. Evans discusses the ongoing 'platform wars' among Google, Meta, Apple, and OpenAI, highlighting that the true economic and competitive power lies in distribution and integration rather than mere model dominance.

Evans addresses growing public skepticism, backlash, and concerns about AI’s impact on jobs, privacy, and security. He references misconceptions about the environmental impact of data centers and emphasizes the lack of comprehensive data on AI’s actual effects. The conversation touches on deepfake dangers, algorithmic biases, and societal challenges, urging a balanced perspective and cautious optimism.

Advice is given on how individuals and parents should approach AI’s rise. Evans recommends engaging actively with AI tools, understanding their capabilities and limitations, and focusing on skills that are hard to automate. He notes that the future of work is unpredictable but encourages a mindset of adaptability, curiosity, and continuous learning. Parenting strategies are also discussed, emphasizing resilience and awareness over specific job predictions.

The conversation concludes with reflections on the historical cycles of technological disruption, reinforcing that change is constant and manageable if approached thoughtfully. Evans advocates for understanding the broader implications of AI, focusing on what new opportunities it creates rather than fears of obsolescence, and encourages proactive experimentation and learning.