Co-Intelligence

1. AI: The Alien Mind Among Us

We have invented technologies, from axes to helicopters, that boost our physical capabilities; and others, like spreadsheets, that automate complex tasks; but we have never built a generally applicable technology that can boost our intelligence.

A new form of intelligence. Large Language Models (LLMs) like ChatGPT represent a breakthrough in artificial intelligence, exhibiting capabilities that seem almost human. Unlike previous AI systems focused on narrow tasks, LLMs can engage in human-like conversation, generate creative content, and solve complex problems across various domains.

Rapid adoption and impact. The speed at which LLMs have been developed and adopted is unprecedented. ChatGPT reached 100 million users faster than any previous product in history. This rapid integration into daily life and work processes is already reshaping how we interact with technology and each other.

Ethical and societal implications. The emergence of AI that can mimic human intelligence raises profound questions about the nature of consciousness, creativity, and the future of work. As these systems become more advanced, society must grapple with issues of AI alignment, ensuring these powerful tools serve human interests rather than potentially harming them.

2. The Jagged Frontier: AI's Unpredictable Capabilities

There are always problems with giving the AI tests, because the answer key might be in its training data, effectively allowing it to cheat by knowing the answers in advance.

Uneven abilities. AI's capabilities are not uniformly distributed across tasks. Some tasks that humans find difficult, AI can perform with ease, while others that seem simple to us can be challenging for AI. This creates a "jagged frontier" of AI abilities that is difficult to predict or map.

Hallucinations and limitations. Despite their impressive capabilities, LLMs are prone to "hallucinations" - generating plausible but incorrect information. They don't truly understand or know facts, but rather predict likely sequences of words based on patterns in their training data.

Implications for assessment and use. The jagged frontier and potential for hallucinations make it challenging to accurately assess AI capabilities or rely on them for critical tasks without human oversight. Understanding these limitations is crucial for effectively integrating AI into workflows and decision-making processes.

3. Alignment: Ensuring AI Serves Humanity

Aligning an AI requires not just stopping a potential alien god but also considering these other impacts and the desire to build an AI that reflects humanity.

Existential risks. Some AI researchers worry about the potential for superintelligent AI to pose existential risks to humanity if not properly aligned with human values and goals. This "alignment problem" is a central concern in AI ethics and development.

Immediate ethical challenges. Beyond long-term risks, current AI systems present more immediate ethical challenges:

• Bias and fairness issues in AI decision-making
• Privacy concerns related to data used for training
• Potential for misuse in creating disinformation or manipulating people
• Impact on jobs and economic inequality

Multifaceted approach to alignment. Addressing AI alignment requires efforts on multiple fronts:

• Technical solutions to make AI systems more transparent and controllable
• Ethical guidelines and regulations for AI development and deployment
• Public education and engagement to shape the future of AI in society
• Ongoing research into AI safety and beneficial AI development

4. Four Principles for Working with AI

Always invite AI to the table.

Principle 1: Always invite AI to the table. Experiment with AI in various tasks to understand its capabilities and limitations. This exploration allows you to become an expert in applying AI to your specific domain.

Principle 2: Be the human in the loop. Maintain oversight and critical thinking when using AI. Human judgment remains crucial for verifying AI outputs and ensuring ethical use.

Principle 3: Treat AI like a person (but tell it what kind of person it is). Frame your interactions with AI by giving it a specific persona or role. This can lead to more tailored and effective responses.

Principle 4: Assume this is the worst AI you will ever use. AI capabilities are rapidly improving. Stay open to new developments and be prepared for AI to become increasingly capable over time.

5. AI as a Creative Partner: Augmenting Human Abilities

Given a machine that can make anything, we still default to what we know well.

Enhancing creativity. AI demonstrates remarkable creative abilities, often outperforming humans on standard creativity tests. However, its true power lies in augmenting human creativity rather than replacing it.

Democratizing creative expression. AI tools are enabling people with limited artistic skills to express themselves creatively in new ways. This democratization of creativity has the potential to unlock human potential on a vast scale.

Challenges and ethical considerations:

• Potential loss of originality as people rely too heavily on AI-generated content
• Copyright and attribution issues for AI-generated works
• Need to redefine creativity and authorship in the age of AI

6. The Workplace Revolution: AI as Coworker

Even if you could replace a doctor with an AI overnight, would patients be okay being seen by a machine? How would liability rules work? How would other health-care professionals adjust?

Task transformation. AI is reshaping work by automating or augmenting many tasks, particularly in knowledge-intensive fields. This doesn't necessarily mean job replacement, but rather a shift in the nature of work.

Productivity gains and skill leveling. Studies show that AI can significantly boost productivity, especially for lower-performing workers. This has the potential to reduce performance gaps within organizations.

Adaptation strategies:

• "Centaur" approach: Strategic division of labor between humans and AI
• "Cyborg" approach: Deep integration of AI into human workflows
• Focusing on uniquely human skills and expertise
• Reimagining organizational structures and processes to leverage AI capabilities

7. Reimagining Education in the AI Era

AIs can help us flourish by making it impossible to ignore the truth any longer: a lot of work is really boring and not particularly meaningful.

The homework apocalypse. AI's ability to complete assignments and write essays is forcing educators to rethink traditional assessment methods and the purpose of homework.

AI as tutor and learning companion. AI has the potential to provide personalized, adaptive learning experiences at scale, potentially addressing the "two sigma problem" in education.

Shifting focus of education:

• Emphasizing critical thinking and creativity over rote memorization
• Teaching AI literacy and effective collaboration with AI tools
• Preparing students for a rapidly changing job market
• Leveraging AI to create more engaging, interactive learning experiences

8. AI as Coach: Enhancing Expertise and Learning

The issue is that in order to learn to think critically, problem-solve, understand abstract concepts, reason through novel problems, and evaluate the AI's output, we need subject matter expertise.

Building foundational knowledge. Despite AI's vast knowledge base, human experts still need a strong foundation of facts and understanding to effectively work with and evaluate AI outputs.

Deliberate practice with AI. AI can enhance the process of deliberate practice by providing immediate feedback, personalized challenges, and continuous assessment.

Redefining expertise:

• Shifting from memorization to effective information synthesis and application
• Developing skills in AI collaboration and prompt engineering
• Cultivating uniquely human abilities like empathy, creativity, and complex problem-solving

9. Four Futures: Scenarios for AI's Impact on Society

There is no theoretical reason why this can't happen, but also no reason to suspect that it might. There are world experts on AI who argue both positions.

Scenario 1: As Good as It Gets. AI capabilities plateau, leading to incremental improvements but no major breakthroughs. Society adapts to current AI capabilities without radical disruption.

Scenario 2: Slow Growth. AI continues to improve at a steady, manageable pace. This allows for gradual adaptation of social, economic, and educational systems.

Scenario 3: Exponential Growth. Rapid AI advancement leads to dramatic societal changes, potentially including widespread job displacement and the need for new economic models.

Scenario 4: The Machine God. AI achieves human-level intelligence or beyond, fundamentally altering humanity's place in the world and raising existential questions about our future.

Preparing for an uncertain future:

• Developing flexible policies and regulations that can adapt to different AI scenarios
• Investing in AI safety research and alignment efforts
• Fostering public dialogue about the ethical implications of AI development
• Prioritizing human-AI collaboration and coexistence rather than replacement

Last updated: July 21, 2024