Lecture, Conference Presentation, Presentation
The Promise of AI
- AI is predicted to become as ubiquitous in software as the relational database over the past four decades, permeating every critical application.
- Autonomous driving is expected to become sufficiently cheap to enable self-driving vehicles for logistics, including beer delivery from breweries, autonomous shopping carts moving to customers, and drones following users during exercise.
- Unmanned aerial vehicles will eventually operate in unmapped environments to map buildings in real-time, identify occupants, and deliver supplies like blood to distribution centers in Western Rwanda via parachute to bypass hazardous roads.
- Current AI image classification algorithms have already surpassed human accuracy with error rates below 5%, enabling real-time object recognition in consumer apps like Pinterest's live camera.
- AI-driven sorting technology will allow individuals to grade produce with high precision, while agriculture will shift to variable-rate farming that treats individual crops like lettuce heads rather than entire fields.
- Physical retail will utilize cameras and robots for automatic checkout, product recognition, planogram verification, and inventory management, mirroring Amazon Go's model.
- Future AI research will advance from single-image classification to reconstructing narratives from picture sequences, enabling search engines to answer complex queries and allowing retroactive generation of cookbook instructions from video tutorials.
- Generative AI capabilities will expand to create photorealistic images and music from text or sketches, select scenes for movie trailers, and partially generate software by remixing existing code samples.
- The cost of prediction will decrease significantly, facilitating new applications in video selection, behavioral biometric authentication to replace passwords, and automated customer service routing that identifies callers without manual verification.
- Advanced diagnostic and monitoring tools will include inexpensive liquid biopsies for early cancer detection, wearable sensors to predict abnormal cardiac events, and algorithms predicting suicide risk with 80% accuracy based on electronic health records over a two-to-three-year horizon.
- Operational efficiency gains are projected through AI optimization of traffic flow, soccer strategies, compiler code, and data center electricity consumption, the latter potentially saving 20-25% of energy costs.
- Logistics optimization will shave 8% off grocery delivery times, while natural language understanding will triple communication speed over phone typing and enable AI to summarize multiple documents or generate 10% of mobile email replies.
- Specific societal applications include removing bias from job descriptions, automating legal document categorization, and implementing emotional design for computer systems that simulate human emotions.
- Real-time language translation via earpieces is expected to convert spoken languages within one to two seconds, and AI will analyze crisis counseling texts to identify effective intervention strategies.
- AI tools are anticipated to improve daily and increasingly become open source, eventually supported by an "AI cookbook" to guide developers on available resources.