Hey Superintelligence Fam
AI is no longer just spotting cancer - it’s designing new treatments. Google and Yale’s recent breakthrough is a signal: artificial intelligence may soon play doctor, researcher, and innovator—all in one.
From robots that walk to models that reason across 96K tokens, AI’s future feels supercharged. But even as it builds wealth and tools, fixing AI’s own “jagged intelligence” remains its biggest challenge.
Can AI Cure Cancer? Google’s Groundbreaking Research Hints at a Transformative Future : Google Research and DeepMind, working with Yale School of Medicine, have developed an AI-model that not just analyzes but proposes new cancer therapies, validated in human cell tests - a leap from diagnostics to discovery.
JPMorgan: AI Stocks Inject $5 Trillion Into U.S. Household Wealth, But Concentration Risks Loom : According to JPMorgan Chase analysis, 30 key AI-related companies boosted U.S. household wealth by about $5 trillion in the last year, driving consumer spending - but also highlighting heavy concentration risk in the tech-market.
Jagged Intelligence: DeepMind CEO Demis Hassabis Sounds the Alarm on AI’s Most Dangerous Flaw - Demis Hassabis, CEO of DeepMind, warns that today’s AI systems suffer from “jagged intelligence” - excelling at complex tasks yet failing trivial ones - and cautions this inconsistency must be fixed before we reach true general intelligence.
Google Veo 3.1 : Google’s Gemini-powered video generator creates cinematic, long-form shots from text or images, with native audio, fine-grained editing controls, style consistency, and professional outputs for creators.
Figure 03 : A next-generation humanoid robot combining dexterous manipulation, mobile autonomy, and multimodal AI for general-purpose assistance in homes and workplaces, learning from demonstrations and natural language.
Claude Haiku 4.5 : Anthropic’s compact, fast model delivers strong reasoning and coding performance with low latency, improved tool use, vision, and safety - ideal for responsive apps and robust workflows.
Inoculation Prompting (IP) : A clever fine-tuning strategy where models are deliberately prompted to make mistakes during training - drastically reducing reward hacking, sycophancy, and toxicity while preserving core task accuracy.
Reasoning over Longer Horizons via RL : By chaining simpler tasks and training with outcome-only rewards under a length curriculum, this method boosts long-horizon reasoning accuracy up to 2.9× and transfers to harder benchmarks.
The Markovian Thinker : A breakthrough RL framework that chunks long reasoning traces into fixed-size states, slashing memory costs and enabling 96K-token-long thinking with linear compute and improved performance.
Get A foundational Understanding of LLMs
oin our free 7-day email series and dive into the essentials of Large Language Models—covering core concepts, model design, training workflows, real-world applications, and ethical best practices.
Bonus: On the final day, you’ll receive a comprehensive Intro to LLMs eBook - absolutely free.
Sign up now and take your AI expertise to the next level!
UNESCO launched a Global AI Ethics Observatory and cross-border expert network. At GITEX Global, experts warned that generative AI “hallucinations” demand stronger liability frameworks. Meanwhile, Boston Consulting Group highlighted widespread gaps in workforce training, underscoring that ethical AI requires both governance and human-centric capacity building.
Thank you for tuning in to this week’s edition of Superintelligence Newsletter! Stay connected for more groundbreaking insights and updates on the latest in AI and superintelligence.
For more in-depth articles and expert perspectives, visit our website | Have feedback? Provide feedback.
To explore sponsorship opportunities : Explore Here
Stay curious, stay informed, and keep pushing the boundaries of what’s possible!
Until Next Time!
Superintelligence Team.