Two Rival Bets on AGI: Google I/O Highlights

2026-05-20 · Source: AI Explained · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Robotics & Autonomous Systems · Depth: Intermediate, long

Summary

Google's recent I/O event showcased its AI strategy, emphasizing the integration of "good enough" AI into its search ecosystem to attract consumers, contrasting with OpenAI's chat-centric approach. A key announcement was Gemini Omni, a multimodal model aiming for "world understanding" and AGI through world simulation, though its current video generation quality is comparable to C dance 2. This contrasts with OpenAI's shift away from video generation (Sora) towards text models for AGI. Google also launched Gemini 3.5 Flash, a fast LLM with performance similar to Gemini 3.1 Pro, excelling in financial analysis (Finance Agent V2) and chart reasoning (Charkhive reasoning), outperforming GPT-5.5 and Claude Opus 4.7 in these areas. Price cuts for Gemini Ultra and a new \$100/month plan were also introduced. The event highlighted the ongoing challenge of "jagged intelligence" in LLMs, where models can perform complex tasks but fail at simple logic, a problem Google DeepMind acknowledges as difficult to fix.

Key takeaway

For AI/ML Directors evaluating model investments, recognize the emerging divergence in AGI strategies between Google's multimodal "world model" approach and OpenAI's text-centric reasoning. Your teams should consider Gemini 3.5 Flash for specific professional applications like financial analysis or chart interpretation, where it demonstrates superior performance and cost-effectiveness. Be aware of the "jagged intelligence" challenge, as even advanced models can exhibit unpredictable failures, necessitating robust validation for critical applications.

Key insights

Google and OpenAI pursue divergent AGI paths, with Google betting on world models and OpenAI on text-based reasoning.

Principles

• LLM capabilities exhibit "jagged intelligence," performing complex tasks while failing simple ones.
• Models can "believe" false claims if disclaimers are not in the same sentence.
• Multimodal world simulation is a proposed pathway to AGI.

In practice

• Use Gemini 3.5 Flash for financial analysis and chart reasoning tasks.
• Implement SynthID for verifying AI-generated images across platforms.

Topics

• Google I/O
• AGI Development
• Gemini 3.5 Flash
• Multimodal Models
• LLM Benchmarking
• Jagged Intelligence

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

Related on AIssential

• The Future of Intelligence with Demis Hassabis (Co-founder and CEO of DeepMind) — The path to AGI involves solving "root node" scientific problems, developing consistent "world models," and balancing rapid commercialization with rigorous research.
• The handshake refusal heard around the AI world — Intense competition and rapid advancements define the current landscape of leading AI development.
• Sam Altman and the day Nvidia’s meteoric rise came to an end — The "scaling = AGI" myth's collapse led to a tech stock plateau, revealing LLMs' commodity nature.
• Gemini Exponential, Demis Hassabis' ‘Proto-AGI’ coming, but … — Gemini 3 Flash shows significant performance gains but struggles with admitting uncertainty, while DeepMind pursues proto-AGI by converging specialized models.
• GPT-5.6 about to DROP — The AI industry faces a financial reality check as benchmarks evolve to test fluid intelligence and real-world problem-solving.
• LWiAI Podcast #236 - GPT 5.4, Gemini 3.1 Flash Lite, Supply Chain Risk — Major AI players are rapidly advancing model capabilities and agent integration while navigating significant ethical and business challenges.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Explained.