This video features Wharton School professor Ethan Mollick, author of "Co-Intelligence," illuminating in an accessible and practical way how humans and artificial intelligence (AI) influence each other and reshape work, society, and the future. It clearly proposes insights on the current state of AI development, experimental research cases, how the way we work is changing, and how organizations and individuals can harness these changes. "In the AI era, how do we adapt and get ahead?" The core principles and specific case studies of change are presented in a friendly yet engaging manner.
1. Opening: Anticipation for the Age of Human-Computer Interaction
The talk begins in a casual tone. The speaker engages the audience by asking how fascinating AI-based conversational services (voice interaction, conversational CX, etc.) are in reality, lightening the mood with everyday jokes about managing family and children, and worrying about college tuition.
"Who wants to get voice interaction support while driving? ... But please never do it while actually driving."
"My daughter is currently studying psychology in New York, paying $100,000 a year. And my other daughter has become interested in ornithology. What's that? It's the study of OJ Simpson. Just kidding."
After humorously connecting with the audience's everyday life, the main topic of Co-Intelligence -- collaborative intelligence between humans and AI takes center stage.
2. Introduction of Ethan Mollick and AI in the 'Era of Distrust'
The moderator introduces Professor Ethan Mollick as "the world's most influential person in AI according to TIME magazine" and through his various AI research and publications (especially "Co-Intelligence"), noting that while he can't be on stage in person, he's participating via video.
Professor Mollick's opening words are also lighthearted.
"I can't actually hear the applause, so I'll just assume it's there!"
He shows a fake avatar video of himself, pointing out the fact that current AI can create 'convincing' fake videos from just a few seconds of video and audio samples.
"This is a completely fake video. In just 30 seconds, it created my avatar and can make it say anything. ... Everyone, you can no longer trust even your own eyes."
By showing scenes of 'himself' speaking in various languages, he vividly conveys how the boundaries of real-world surveillance and trust are blurring.
3. AI's Mass Adoption and the Reality of 'AI Contamination'
Professor Mollick emphasizes how widely and undetectably AI has now spread.
- Mass adoption:
"Currently, 70% of college and high school students use AI at school. And the detectors that claim to identify AI usage are virtually useless."
- Content contamination:
"Since 2022, we absolutely cannot reliably distinguish between AI-written text and human-written text. Our world has already been 'contaminated' by AI, and this trend is irreversible."
He points out that this is no longer an era where AI technology is given first only to wealthy corporations or institutions (Goldman Sachs, NSA, etc.) -- anyone in the world can now access sophisticated AI.
-
"Right now, a child in Mozambique can use more advanced AI than Goldman Sachs. In fact, some major US corporations and institutions have even banned AI in the workplace, but in reality, everyone is using it on their smartphones."
4. The Era of Uncertain Exploration: Practical AI Application and Experimentation
He emphasizes that even within the AI 'industry,' nobody has the answers in this era.
"I talk with Anthropic, Google, OpenAI, and various consulting firms every week, but even they don't know exactly what works best or what's most suitable for what. ... Experimentation and discovery is now homework that everyone must do together."
Professor Mollick uses actual research, including experiments conducted with BCG (Boston Consulting Group), to demonstrate the following:
- Groups using ChatGPT-4 saw an 'immediate' 40%+ improvement in work performance, were 26% faster, and increased total output.
- Below-average participants saw even greater gains from AI assistance (43%!), and in fact:
"Results submitted directly from AI scored the best. ... When people added their own thoughts or additional ideas, the results actually got worse."
- This isn't limited to specific fields: In law school, medical school, and elsewhere, the phenomenon of 'the lowest performers disappearing' occurred when AI was integrated -- a remarkable equalization effect.
5. AI's Capabilities: Creativity, Persuasion, and Change at the Cliff's Edge
He clearly demonstrates with examples that AI is surpassing humans in an increasing number of areas.
- Creativity experiment:
"In a Wharton startup idea contest, 35 out of the top 40 ideas were generated by GPT-4."
- Transcendent persuasion:
"In a direct debate experiment with AI, people were 81.7% more persuaded by AI's arguments than by human arguments. And after just three months of debating with AI, conspiracy theory beliefs significantly decreased. ... There's a positive side, but it's also frightening that conversations alone can change people's deep beliefs."
- Major impact on high-paying, highly-educated, creative professions:
"The jobs most affected by AI are those that are more educated, more creative, and better paid. My own job (business school professor) ranked 22nd in impact in a 2016 survey."
He then lists 'AI-least-affected' jobs like professional dancers, athletes, roofers, and laborers, and vividly conveys the speed of technology adoption with the example that even an Amish roofer uses AI for email proposals.
6. The Power of Frontier Models and Live Demos: What AI Can Do
Emphasizing that AI's performance improves dramatically with model scale:
- Bloomberg spent $10 million building a 'specialized' AI, but the far more general-purpose and larger-scale GPT-4 outperformed it.
- He notes that currently GPT-4, Gemini, Claude, Meta Llama 3.1, and Grok-2 are the five representative frontier models.
He then demonstrates through live chat AI demos how "virtually every stage of actual work" -- from trivial tasks to high-level business ideation, product development, marketing campaigns, and real-time feedback and improvement -- can be handled instantly by AI.
"From the content I normally teach students, to generating actual MBA-level business ideas, drafting emails, and preparing customer interviews -- it's all possible right away with ChatGPT."
"In fact, 18 months ago, our company could do real-time new product planning, marketing, website creation, social media, and video promotion all with AI in just 30 minutes. Now it's getting even faster and more commonplace."
He also demonstrates generating market research, product launches, case studies, and new course curriculum design -- all at a Solid A- level -- within 59 seconds.
"In this era where AI can do all of this, I believe the nature of our work fundamentally has to change."
7. The Essential Transformation of Work: 3 Types of Tasks and the Arrival of the 'Agent Era'
Professor Mollick analyzes that organizations are now differentiating into human-only work, hybrid (centaur/cyborg-type) work, and agent tasks.
- Centaur type: A model where humans and AI divide roles and collaborate
- Cyborg type: A model where AI and humans merge more deeply and work organically
- AI 'agents' (autonomous systems) are currently in early development stages, but he shows with concrete examples that autonomously judging and executing tasks is a clear goal for the AI industry in 2025.
"I had an AI agent called Devon build a 10-K analysis website, and it independently planned, built, requested feedback, set pricing at $50-100, and even conducted actual client negotiations!"
8. Lessons from the 'Jagged Frontier' and Real Innovation in How We Work
Professor Mollick presents through firsthand experience that even tasks AI seemed unable to do (e.g., crossword puzzles) can be accomplished when new prompts or approaches emerge.
"There are things AI does well and things it doesn't, but the problem is we don't know which is which until we try. For example, it might struggle with summarizing in exactly 25 words, but it creates poetry brilliantly."
He also emphasizes that traditional organizational systems (the first org chart from 1864, Agile, etc.) were designed solely for 'humans,' and therefore don't fit an AI era where multiple intelligences coexist.
- In practical applications, he adds examples where product reviews, internal meetings, and feedback processes that used to take one to two weeks are now completed in a single day by AI, and AI even provides better feedback than actual people.
9. New Principles Needed Now and Organizational Responses
Professor Mollick proposes four fundamental principles for wisely utilizing AI.
- Actively try AI everywhere:
"Wherever AI can play a role, try it as long as it's ethically and legally permitted. That's how you learn what AI does well and what it doesn't."
- What you do best won't be easily replaced:
"The work you love most, are best at, and get paid the most for won't be easily replaced by AI. Leave the rest to AI and focus on growing your own unique value."
- Prompting isn't hard -- just have a conversation:
"Don't treat AI rigidly like software. Approach it naturally, as if you're having a real conversation. That's the most effective way to use it."
- Right now is the weakest AI will ever be:
"Everything I demonstrated today is already 'outdated.' Remember that this is the weakest AI you'll ever use!"
At the organizational level, the key points are:
- Cut out work that AI has made obsolete,
- Actively explore new work that AI has made possible,
- If there are no clear answers, activate internal experiments (labs, skunkworks, etc.) to find your own answers.
"Nobody is secretly hiding something while leaving only your company behind. Everyone is experimenting and searching for answers. ... Innovation incentives and experimentation spaces must stay alive within your organization."
10. Closing: The Path of Co-Evolution Leading to Human Flourishing
Professor Mollick closes by emphasizing that AI must be used for human flourishing and empowering individuals, customers, and society as a whole.
"Change doesn't come to you on its own -- it depends on how we receive and utilize it. Wisely using AI for the 'growth and flourishing' of employees, customers, and our world is the real answer."
He concludes the talk with a heartfelt farewell.
Conclusion
Professor Ethan Mollick's Co-Intelligence talk emphasizes that the disruptive changes AI will bring are already a reality beside us, and now is the time for individuals, organizations, and society to actively experiment and become 'partners' in innovation. AI is the ultimate assistant that fills in our weaknesses and helps us focus on what we truly do well, reminding us that this is an era of experimentation where we must discover new possibilities amidst uncertainty. Don't hesitate -- invite AI into your life and work like a friend!