Executive Brief: Nov 30

Innovation Insights 

1. LLM in creative work: The role of collaboration modality and user expertise (Management Science)
An experiment comparing two collaboration modalities—using LLMs as “ghostwriters” versus “sounding boards”—revealed that LLMs as sounding boards significantly improve ad quality for nonexperts, while using them as ghostwriters can be detrimental to experts. The study found that the ghostwriter approach leads to an anchoring effect, resulting in lower-quality ads, whereas the sounding board approach helps nonexperts produce content similar to that of experts, thereby narrowing the performance gap.

2. AI and the R&D revolution (Financial Times)
An entrepreneurial culture within R&D teams is crucial for fostering transformational innovation. Clean and well-managed data is essential for AI systems, enabling personalized product development and efficient market targeting. Generative AI acts as a co-founder or brainstorming buddy, assisting in brainstorming, market research, and product testing without replacing human roles. Technological tools like CAD-CAM software, digital twins, and 3D printing are revolutionizing the R&D process, making it more efficient and accessible. AI’s impact on the pharmaceutical industry is profound, potentially accelerating drug development by four years and reducing costs by up to 45%. 

3. Large language models surpass human experts in predicting neuroscience results (Nature Human Behaviour)
Large language models trained on scientific literature can synthesize vast amounts of research to predict novel results, potentially outperforming human experts. The BrainBench benchmark was created to evaluate this, and findings show that LLMs, particularly BrainGPT tuned on neuroscience literature, surpass experts in predicting experimental outcomes. High-confidence predictions from LLMs are more likely to be correct, indicating a future where LLMs assist in making scientific discoveries across various fields.

4. Train your brain to work creatively with GenAI (Harvard Business Review)
Most people prompt generative AI based on their existing paradigms and linear thinking, similar to how they use search engines like Google. However, AI’s true potential is unlocked when it is used to create unique outcomes that wouldn’t be possible without human-machine collaboration, requiring a shift in how we think about AI’s capabilities and expected results. The article provides 12 exercises to help expand thinking around prompting GenAI, encouraging leaders to cultivate a culture where achieving the “impossible” becomes a regular challenge and accomplishment.

5. Building AI trust: The key role of explainability (McKinsey)
Explainable AI enhances transparency by clarifying how AI models function, enabling operational risk mitigation, regulatory compliance, continuous improvement, and user confidence. It supports stakeholder-specific needs through global and local explanations and requires a cross-functional, human-centered approach integrated into the AI development lifecycle. By aligning XAI practices with ethical and regulatory standards, organizations can foster adoption, improve performance, and unlock AI’s full potential for value creation while ensuring responsible and human-centric use.