Five New Thinking Styles for Working With Thinking Machines

• Instead of designing step-by-step solutions, engineers now train AI systems by exploring infinite possibilities until one works. This isn’t about finding universal truths; it’s about achieving specific outcomes.
• This paradigm shift—from theorizing to engineering—could fundamentally reshape how we approach problems, pushing us to embrace more flexible, results-oriented thinking styles.

• In a pre-AI world, success depended on isolating and mastering the essence of a problem—defining clear, static rules to address challenges.
• Software required pinpointing the core user and their problem; marketing plans relied on fixed theses and strategies.
• This reductionist approach assumed that breaking problems into their simplest elements would yield the best solutions.

• Predictive AI thrives on sequences—mapping input patterns to likely outcomes (e.g., identifying churn in SaaS businesses).
• Generative AI, on the other hand, takes sequences further by creating entirely new data or solutions that were not predefined (e.g., AI-generated content like GPT-4 writing marketing copy).

This shift shows generative AI as a partner in design, imagining potential paths that predictive models cannot discern.

• In a post-AI world, however, essence gives way to sequence. The focus shifts from identifying universal truths to analyzing patterns of behavior or chains of events.
• For example, a SaaS business once relied on rigid rules to predict churn—like tracking login frequency or payment data.
• Now, AI models examine vast sequences of user behavior, uncovering nuanced patterns no human could predict or explicitly define.

• This shift is profound: AI doesn't "understand" the problem as humans would but detects intricate correlations across millions of variables.
• By feeding historical data into machine learning models, businesses can predict outcomes with staggering precision, bypassing traditional rule-setting altogether.
• The surprising implication? AI redefines problem-solving as probabilistic rather than deterministic.
• Instead of designing solutions around abstractions, we now let machines find actionable insights buried in complexity—a radically different approach to understanding and shaping the world.

• In a post-AI world, the focus shifts from crafting explicit rules to identifying patterns through examples—a transformation that challenges traditional problem-solving.
• Rather than painstakingly coding systems or defining principles, AI enables us to translate abstract preferences and complex behaviors into actionable frameworks.
• For instance, instead of dictating how software should function, you can provide a collection of UI inspirations, and AI will discern the underlying patterns to build it. 

• Similarly, in creative work, AI can replicate brand voice and style by analyzing curated examples, bypassing the need for rigid systems. 
• This approach turns "taste" into a teachable input, empowering teams to scale and adapt their output fluidly.
• The surprising twist is that clarity now hinges on high-quality examples, not logical frameworks—shifting the emphasis from explaining to showing.
• This pattern-first mindset reframes intellectual and creative work, pushing us to think less about rules and more about resonance.

While predictive AI excels at recognizing patterns to classify or forecast (e.g., fraud detection in financial systems), generative AI transcends pattern detection, producing creative or functional outputs such as MidJourney’s art or AlphaFold's protein structures.

Generative AI isn’t confined to finding rules but proposes new paradigms by simulating possibilities.

• Traditionally, software required reducing tasks to explicit rules—a framework that worked well for structured systems like customer relationship management.
• However, many complex tasks, like optical character recognition (OCR), resist such reduction. Deep learning now enables software to develop "intuition," effectively handling tasks once deemed unprogrammable. 
• This breakthrough extends beyond OCR into realms thought untouchable by software, like evaluating startup pitches or diagnosing medical conditions.

• These tasks rely on nuanced, ineffable judgments that AI can now replicate and transfer.
• In a post-AI world, intuition becomes not just accessible but deployable, breaking free from human exclusivity and revolutionizing fields reliant on tacit expertise.
• This shift challenges the traditional belief that only processes can be mechanized, opening doors to new, intuitive possibilities for technology and human collaboration.

• Predictive AI often supports process refinement (e.g., optimizing logistics with data insights).
• Generative AI builds on intuition, creating novel possibilities—like OpenAI’s DALL-E generating advertising concepts from abstract prompts.

This ability to work beyond logical constraints introduces a form of AI-driven creativity, blending engineering precision with artistic imagination.

• Previously, creativity demanded precision and control, akin to sculpting—a process where every detail was manually crafted, with each step entirely reliant on the creator’s intent.
• Coding, for instance, required meticulous shaping of logic and structure, piece by piece.
• Now, AI tools like Cursor reframe the process as gardening. Instead of chiseling ideas into existence, creators establish conditions for ideas to flourish.
• Prompting AI models replaces manual labor, transforming the act of creation into one of guidance and curation. 

• The focus shifts from direct intervention to fostering an environment where growth can happen organically.
• This evolution challenges traditional notions of authorship, emphasizing collaboration over singular ownership.
• As the creative process becomes more intuitive and adaptive, the role of the creator morphs into one of partnership with intelligent tools—an unpredictable and deeply collaborative journey.

• In predictive workflows, AI operates like a sculptor refining solutions from raw data.
• Generative AI, akin to gardening, nurtures multiple creative "plants" through iterative processes, such as ChatGPT brainstorming strategies or Runway ML generating movie scenes.

This cooperative approach emphasizes user input to grow tailored results dynamically.

• Western thought fixates on explanations, seeking rules to control and demystify success, whether in individuals like Mark Zuckerberg or in scientific phenomena.
• Yet explanations often fail to capture reality's complexity. Zuckerberg’s success, for instance, stems less from explicit principles and more from intuitive decision-making he likely can’t fully articulate.
• The post-AI era flips this dynamic. Predictions, not explanations, are becoming paramount.
• AI models, trained on vast data, encapsulate intuition without needing human comprehension of underlying processes. 

• This transition is reshaping science: the 2024 Nobel Prizes in physics and chemistry went to computer scientists optimizing predictive architectures—not traditional theorists crafting better explanations.
• This shift transforms science into engineering, emphasizing creation over understanding.
• The question is no longer "What is this?" but "How can I predict this?"
• This pivot could redefine progress, transcending Enlightenment rationalism and opening new paradigms for navigating complexity and understanding ourselves.

• The shift from explanations to predictions defines predictive AI, yet generative AI challenges this by creating data-driven insights that go beyond explanation or prediction into transformation.
• For example, instead of explaining human language patterns (predictive AI), generative AI enables translation tools like DeepL to simulate the nuanced generation of idiomatic text.

Generative AI challenges the engineering-centric narrative. It acts both as a tool for creation and as a collaborator, simulating forms of "intuition" that mimic human innovation.

For example, AlphaCode solves competitive programming problems, not by strictly predicting outcomes but by generating potential solutions that are tested and refined.

This contrasts with predictive AI’s deterministic bent, offering an emergent layer of creativity.

While predictive AI aligns well with engineering goals (e.g., Tesla’s autonomous vehicle prediction models), generative AI introduces a vital layer of possibility in problem-solving:

• Science: Generative AI like Meta's ESMFold predicts protein functions and imagines hypothetical mutations for therapeutic discovery.
• Creativity: Tools like Adobe Firefly offer design professionals intuitive platforms for creating personalized visuals without predefined outputs.
• Education: Khanmigo AI shifts from grading patterns to generating tailored learning pathways, making education a personalized journey.

1. How do predictive and generative paradigms reshape the notion of "expertise" in creative fields?
2. Could generative AI models be prone to amplifying biases more than predictive ones due to emergent properties?
3. How might society redefine intellectual property rights in a world increasingly driven by generative AI outputs?