Baiting AI [LIVE]

2026-05-01 · Source: Matthew Berman · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, extended

Summary

The content covers several key topics, beginning with technical difficulties during a live stream, including microphone crackling and video lag, which were eventually resolved. The discussion then shifted to Apple App Store scams, exemplified by the host's mother unknowingly purchasing fake ChatGPT applications due to misleading search results and similar logos. This highlighted issues with app store moderation and user experience, particularly for less tech-savvy individuals. The host also explored the "jagged nature" of AI intelligence, referencing Andre Karpathy's insights on why AI excels in verifiable domains like coding and math but struggles with nuanced common-sense questions, attributing this to verifiability and data mixture. Anthropic's controversial API usage restrictions and perceived "cult-like" company culture were also discussed, alongside a new model called Owl Alpha and Meta's plan to use employee computer activity for AI model training, raising privacy and ethical concerns.

Key takeaway

For CTOs and VPs of Engineering evaluating AI adoption, recognize that AI's "jagged" intelligence means it will excel in specific, verifiable tasks like code generation but may fail at common-sense reasoning. Prioritize AI applications where performance can be objectively measured and data is abundant, while also scrutinizing vendor policies for transparency and ethical data use, especially regarding API restrictions and employee monitoring.

Key insights

AI's intelligence is "jagged," excelling in verifiable domains due to data and reinforcement learning, while app stores struggle with scam prevention.

Principles

• AI performance correlates with data verifiability and feedback loop speed.
• App store search algorithms can be exploited by malicious actors.

Method

AI model training benefits from easily verifiable data and rapid feedback loops, as seen in coding tasks, leading to improved performance in those specific domains.

In practice

• Implement robust app store moderation to combat misleading applications.
• Consider the ethical implications of using employee data for AI training.

Topics

• Apple App Store Scams
• AI Jagged Intelligence
• Andre Karpathy Insights
• Anthropic API Restrictions
• Employee Monitoring

Best for: CTO, VP of Engineering/Data, AI Scientist, Director of AI/ML, Tech Journalist

Related on AIssential

• Meta rolls out new AI content enforcement systems while reducing reliance on third-party vendors — Advanced AI systems can significantly enhance content enforcement accuracy and efficiency while reducing human exposure to graphic content.
• Diffuse AI Control on Fuzzy Tasks — Diffuse AI Control mitigates AI sabotage on fuzzy tasks via an adversarial game framework to develop robust AI systems.
• What Apple’s AI Crackdown Got Right — Apple's removal of AI coding apps from its App Store signals distrust in AI-generated code.
• The cognitive demands of AI novelty — AI-accelerated software proliferation creates unsustainable cognitive demands, requiring sharper judgment amidst ambiguity.
• jagged intelligence - possibly a destination not a temporary detour — AI's "jagged intelligence" creates specialized capabilities, with the "harness layer" driving significant asymmetry.
• Claude Opus 4.7, Apple’s AI glasses and Allbirds AI pivot — New AI models offer enhanced agentic capabilities, while hardware and adoption trends reshape market strategies and societal risks.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Matthew Berman.