Hybrid Intelligence and Generative AI

Jul 7, 2024

Lecture Notes on Hybrid Intelligence and Generative AI

Introduction

  • The speaker had a bike accident and broke both arms, which provided an opportunity to explore text technologies.
  • Focus of the lecture: Hybrid Intelligence and Generative AI.

Generative AI in Use

Poll of Audience

  • 50% use generative AI in their code.
  • About a third use text-to-image technologies.
  • Few have used text-to-video technologies.

Auto-GPT Example

  • Auto-GPT can set up agents to create other agents, e.g., ordering a pizza through automated processes.

Impacts on Workflows

AI in the Military

  • AI can streamline and enhance military operations and decision-making.
  • Concerns about the efficiency and potential risks of AI in command systems.

Open Source and Future of AI

  • Open source AI models can be as effective as those from leading corporations.
  • Examples include the development of specialized GPT models (e.g., Bloomberg GPT).

Emerging Technologies

Text-to-Image Progression

  • Generative Adversarial Networks (GANs) were used five years ago.
  • Mid Journey and DALL-E 2 were prominent, now outdated, due to rapid advancements.
  • Stability AI can handle more complex tasks like integrating text and images.

Video and Voice Technologies

  • Deep fakes are becoming more sophisticated and can have significant real-world impacts.
  • Voice cloning and multilingual dubbing are becoming mainstream.
  • For instance, Hume AI can now dub videos in multiple languages seamlessly.

Chatbots and Workflow Integration

  • New tools like ChatGPT plugins enhance productivity (PDF readers, slide generators, etc.).
  • Chatbots can be rapidly prototyped using tools like Chatbase.
  • Example: A chatbot can be integrated into meetings to provide summaries and task updates.

Customer Relationship Management (CRM) Systems

Advanced AI in CRM

  • Salesforce and Adobe are using generative AI to automate and personalize customer interactions.
  • The aim is to create fully integrated solutions that handle all customer-related tasks.

Hybrid Intelligence

Concept

  • Hybrid Intelligence combines algorithmics, human-computer interaction, and management principles.
  • Aim: Create applications that guarantee employee upskilling and collaborative technology development.
  • IBM's CEO claims jobs won’t be lost to AI but to people using AI.

Practical Examples

  1. Autodesk: Collaborative development of AI tools with human input enhances product functionality and user satisfaction.
  2. Cab Drivers: Tasks heavily reliant on prediction (like London cabbie routes) are susceptible to automation.
  3. Radiologists: Although AI can handle image detection, the role of radiologists is multifaceted and not fully replaceable by AI.
  4. Meteorologists: Shift from prediction to decision-making and stakeholder communication due to advanced predictive algorithms.

Design Principles and Economic Models

  • Example of a hybrid intelligence approach in a mail-order service, enhancing operational efficiency and personalization.
  • Benefit Dependency Network: Differentiates between low-road (efficiency) and high-road (upskilling) AI applications.

Human-Centered AI vs. Hybrid Intelligence

  • Human-centered AI focuses on combining high automation with high user control.
  • Hybrid intelligence is broader, aiming to reimagine workflows and enhance organizational efficiency.
  • Systems should be designed for user personalization and co-development, not just adoption.

Final Thoughts

Future of Hybrid Intelligence

  • Continuous upskilling and collaboration are essential for future technology development.
  • Emphasis on creating a safe space for employee input and iterative development processes.
  • Proposal for establishing principles and certification for hybrid intelligence branding.

Call to Action

  • The speaker encourages reaching out to collaborate on developing and globalizing hybrid intelligence principles.