Milken Institute59 min
Leading in the Generative AI Revolution | Milken Institute Global Conference 2024
Sara Eisen, Greg Brown, Arvind Krishna, Thomas Kurian, Janet Truncale, Sharmini Peries, Janet Napolitanovich, Thomas Kaplan, Nicole Sullivan, Arvind K.
Conference Presentation, Panel
Current AI Maturity Landscape
- Arvind Krishna characterizes the AI industry as being in the "first innings" of a game, contrasting it with cloud computing, which he places in the "fifth or sixth innings."
- Thomas Kurian (Google) and others agree that while AI is nascent compared to cloud, the "when" question has shifted to "how to scale."
- Greg Clark (Motorola Solutions) notes that 80% of clients are currently running AI use cases, with 20% advancing to large-scale pilots and deployments.
- Janet Napolitano (EY) observes a pivot in client conversations from focusing on immediate productivity gains to exploring long-term growth opportunities.
Strategic AI Use Cases by Sector
- Public Safety & Security: Motorola Solutions uses AI to "remove the haystack" for first responders, utilizing video, audio, and data to detect anomalies like perimeter breaches or active shooters.
- Deployments include real-time redaction and privacy blurring for body-worn camera footage to build trust.
- AI integration reduced 911 dispatch response times and automated translation/transcription for non-English speakers.
- Healthcare: EY cites AI applications that could reduce X-ray diagnosis inaccuracy rates from 15% to 1%.
- Use cases include predicting patient illness trajectories and automated patient navigation for post-diagnosis care.
- Trust remains a barrier due to "familiarity bias," as professionals demand perfection from AI compared to marginally improved human baselines.
- Search & Advertising: Google reports that generative AI in Search has increased user engagement and conversion rates by enabling more complex, conversational queries (e.g., "plan a trip").
- Search volume has risen as the AI model interprets intent better than traditional keyword matching.
- Enterprise Operations: EY is piloting AI for vendor negotiations and using internal models to upskill 400,000 employees.
- Motorola Solutions reports a 20% increase in code generation speed using AI tools like GitHub Copilot without reducing headcount.
- Public Safety & Security: Motorola Solutions uses AI to "remove the haystack" for first responders, utilizing video, audio, and data to detect anomalies like perimeter breaches or active shooters.
Competitive Differentiation & Platform Strategies
- Google's Approach: Focuses on multimodal processing (video, audio, text simultaneously) and a "open platform" allowing enterprises to choose various models.
- Google recently introduced "grounding" features that require models to cite sources and verify answers against known data to enhance trust.
- Released open-source models "Gemma" and "Code Gemma" to foster innovation and maintain a lead in hardware/software infrastructure.
- IBM's Approach: Differentiates by building "fit-for-purpose" enterprise models that can be deployed on-premise to address data sovereignty and IP leakage concerns.
- IBM argues for an "and" strategy, complementing cloud-native models with proprietary and open-source options for specific regulatory or privacy needs.
- Cloud Provider Selection: Motorola Solutions utilizes a multi-cloud strategy: Microsoft Azure for public safety compliance, AWS for internal IT, and Google Cloud for advanced AI capabilities.
- Google's Approach: Focuses on multimodal processing (video, audio, text simultaneously) and a "open platform" allowing enterprises to choose various models.
Trust, Safety, and Regulation
- Human-in-the-Loop: All panelists emphasize that AI currently operates as an augmenting tool rather than an autonomous decision-maker, particularly in high-stakes environments like public safety.
- Accuracy Benchmarks: Greg Clark notes human accuracy in video surveillance drops to roughly 90% after 20 minutes, while AI aims for similar or slightly higher accuracy but is not 100% reliable.
- Regulatory Stance:
- Panelists support "light-touch" regulation that focuses on use cases and accountability rather than stifling the underlying technology.
- Thomas Kurian states Google welcomes regulation and is collaborating with governments on watermarking and authenticity verification for deep fakes.
- Janet Napolitano warns that overregulation could inhibit innovation, citing the need for principles-based approaches given the technology's speed.
- Bias & Data Integrity: IBM and EY stress that AI bias stems directly from training data; they advocate for rigorous "red teaming" and data management to ensure fairness across demographics.
Workforce Impact & Future Outlook
- Augmentation vs. Displacement: Arvind Krishna predicts a global labor shortage due to declining birth rates in developed nations, necessitating "digital workers" to augment human productivity.
- The consensus is that AI will automate low-value cognitive tasks but will not replace human interaction, empathy, or high-level decision-making.
- EY's internal strategy focuses on upskilling employees to handle more complex, interesting work rather than layoffs.
- Infrastructure Requirements:
- Thomas Kurian notes Google invests heavily in data centers and specialized silicon to support training and inference.
- Arvind Krishna highlights the critical shift toward "Edge AI," where models must run locally on devices (e.g., sensors, factory equipment) rather than relying solely on the cloud.
- Motorola Solutions demonstrated Edge AI capabilities in schools, instantly locking doors and lighting green paths for safety during active shooter scenarios without cloud dependency.
- Augmentation vs. Displacement: Arvind Krishna predicts a global labor shortage due to declining birth rates in developed nations, necessitating "digital workers" to augment human productivity.
Forward-Looking Predictions
- New Product Categories: Thomas Kurian anticipates a surge in new companies building AI-native products (e.g., personalized learning systems for children) rather than just internal process improvements.
- Bureaucracy Reduction: Arvind Krishna predicts a near-term focus on using AI to reduce bureaucratic friction (paperwork) in both government and corporate sectors.
- Misinformation Risks: The panel identifies misinformation as the most immediate danger of generative AI, noting that bad actors can amplify existing societal issues.
- Emotional Intelligence (EQ): Janet Napolitano predicts a future focus on the intersection of AI and leadership EQ, questioning whether AI can ever truly replicate human empathy.