Jun 26, 2017, 4:59 PM ETPanel

AI: Reshaping the World As We Know It

Five panelists represent five critical AI dimensions: conversational interfaces (Rohit Prasad, Amazon), autonomous navigation (Leo Oran, Otto/Uber), venture capital/startups (Mark Gorenberg, Zeta Venture Partners), enterprise adoption (Paul Blaze, formerly PwC, now Trunk), and ethics (Francesca Rossi, IBM).
Moderator Siobhan Zillis (Bloomberg Beta) notes the fund has invested in 40 entrepreneurs solving machine intelligence problems.
Optimism Consensus: In a lightning round, all panelists expressed optimism about the next decade of AI, though some added caveats regarding media-driven "existential" fears.
Ownership Prediction: Regarding full autonomous vehicle ownership in 10 years, only Leo Oran answered "No," citing that the future is likely a "service" model rather than individual ownership.
Human-AI Relationships: One panelist predicted someone in their family would say "I love you" to a bot in 10 years; the consensus remains skeptical of deep emotional bonds but acknowledges the "companion" dynamic is already emerging.

Safety Priority: Over 1 million unnecessary deaths occur globally on roads annually; in the US, while 87% of fatalities happen on highways, trucks (1% of vehicles) are responsible for 25% of those fatalities.
Strategic Focus: Otto targets trucking first because the highway environment is "over-constrained" compared to cities (no pedestrians, fewer variables, limited routes).
Commercial Incentive: Autonomous trucks offer unlimited compute space, power supply, and strong efficiency incentives for logistics.
Augmentation vs. Automation: Oran advocates for full automation rather than augmentation; human-in-the-loop systems create over-trust risks where drivers fail to intervene when the system fails.
Deployment Reality: Otto successfully completed a 120-mile autonomous Budweiser run in Colorado, transporting 51,000 units without a driver.
Regulatory Hurdle: The primary barrier is not just safety metrics (e.g., being 2x or 10x safer than humans) but establishing "explainability" for deep learning "black box" decisions.
Acceptance Threshold: The industry does not require a significant delta in safety to deploy; if a system is demonstrably safer, it should be deployed regardless of total explainability.

User Experience Shift: Voice interfaces are replacing apps to solve the "scaling" problem of interacting with multiple devices (garage, music, lighting) without using hands/eyes.
Data Strategy: Alexa's success relies on "far-field" interactions and handling sparse data; models generalize well from limited data points to handle noisy environments (e.g., TVs, microwaves).
Diversity Requirement: To handle diverse accents and environmental noise, Amazon requires massive, diverse user datasets.
Research Initiative: The "Alexa Prize" is a $1 million university competition challenging students to build social bots capable of 20-minute conversations on trending topics.
Social Impact: Alexa is increasingly acting as a companion for the elderly, helping those with mobility issues control smart home devices and providing social interaction.

Market Structure: Large tech companies are building horizontal platforms (OS, compute), creating opportunities for startups to build vertical-specific applications on top of these scalable infrastructures.
Network Effects: Startups can leverage "minimal viable products" to crowdsource data, creating network effects that eventually attract enterprise customers (e.g., Inside Sales uses data to optimize lead closing by 30%).
Enterprise Adoption Gap: Large company adoption is binary; success requires a tolerance for failure, continuous experimentation, and converting unstructured data (e.g., maintenance logs) into structured variables.
Innovation Metric: A leading indicator of AI maturity is an organization knowing the exact number of predictive models in its production inventory; few companies currently know this.
Re-training Imperative: Corporations are expected to take responsibility for retraining workforces, potentially leveraging MOOCs and online platforms as the pace of change accelerates.

Trust Mechanism: AI must be trusted to make decisions more informative and ethical than humans, mitigating human cognitive biases and inconsistencies.
Value Alignment: Ethical principles are not universal; systems must adapt to local cultural norms, social norms, and specific professional codes (e.g., medical ethics).
Collaborative Governance: IBM, Amazon, Google, Microsoft, Facebook, and Apple have formed a collaborative ecosystem to address deployment challenges with stakeholders including policymakers and economists.
Safety vs. Explainability: There is a tension between achieving statistical safety and the "explainability" of deep learning; regulators may need to accept "black box" safety if forensic data shows consistent safety performance.
Existential Risk vs. Moral Imperative: Rossi argues the greater risk is not "AI overlords" but the failure to use AI to solve known, immediate problems like pandemics, resource scarcity, and traffic fatalities.

Terminology Shift: Paul Blaze predicts the word "artificial" will be dropped; the technology will simply be called "intelligence."
Platform Saturation: AI will become invisible infrastructure, pervasive in every component of business, moving from experimental to a "wheel of motion" where data superiority creates monopolies (e.g., Google search, Uber autonomy).
Algorithmic Evolution: Mark Gorenberg anticipates breakthroughs in reinforcement learning and new computer systems (e.g., quantum computing) that mimic human thinking.
Value Alignment: Francesca Rossi hopes value alignment, common sense reasoning, and ethical calibration will be significantly advanced within five years.
Urban Transformation: Autonomous vehicles will reduce parking needs by up to 80% in some cities, freeing land for urban agriculture, housing, and distributed energy systems.
Human-AI Symbiosis: The consensus is that the optimal future involves human creativity complementing machine intelligence (e.g., AlphaGo inspiring new human moves), rather than replacement.