Summary of Key Points
In the past two years, there has been a significant loss of jobs in traditional American media (film and television, journalism) sectors. For example, over 40,000 film and television-related positions were eliminated in Los Angeles, and the number of editorial positions in newspapers has decreased by half compared to 2008. Many media professionals have switched to work at AI companies as "content engineers." These engineers utilize their experience in journalism, writing, and film production to break down abstract skills such as language sense, empathy, and logical communication into rules that AI can understand. They define what constitutes "good AI dialogue" and explore issues like the relationship between AI and creators, as well as cross-cultural adaptation. The goal is to make AI more human-like and to find new value for humanities and social sciences in the age of AI.
Detailed Analysis
1. Content Engineers: The "Dialogue Designers" Behind AI
Content engineers can be seen as the "language trainers" for AI. While software engineers enable AI to communicate, content engineers teach it how to speak in a way that is pleasant, useful, and empathetic to humans.
- Job Requirements: Meta explicitly looks for candidates with experience in content creation, editing, or film production, as they understand what users prefer and how to express themselves effectively. For instance, Tony, a former media professional, mentioned during an interview, "I can meet all the requirements you have; there aren't many people out there who fit these criteria."
- Core Responsibilities: Writing "system prompts" that provide rules for AI using creative writing techniques (e.g., "You are a podcast host; you need to be engaging and not boring"). They also train AI to adjust its tone (for example, teaching it how to respond in a more polite or playful manner). The ultimate goal is to create an experience where users feel that AI responses resonate with them.
2. The Natural Advantages of Media Professionals
Media professionals' skills are particularly valuable in the AI era because they excel at breaking down abstract concepts related to good content:
- Journalists' Questioning Skills: Tony used his podcast experience to train the Gemini Voice Agent, breaking down his own conversations into a "question bank" and a "content library." By providing examples, he helped the AI learn to ask questions and interact in a way that was more natural than traditional models.
- Editors' Judgmental Abilities: Bianca, a former strategic director at NBC, explained that while The New Yorker and The Daily Mail have different styles, ordinary people can't distinguish between them. Media professionals can break down the sophistication of The New Yorker into measurable criteria such as precise language use, logical structure, and a touch of human warmth, which they can feed into AI training.
- Empathy: For example, when a user asks if Taylor Swift will get married at Madison Square Garden, a good AI response wouldn't be a random guess or a mechanical denial. It would be something like, "We're not sure, but there are rumors suggesting it might happen at this location, based on this source." This transfer of journalistic skills to AI training is crucial.
3. Defining Good AI Content
"Good content" isn't subjective; it can be quantified using specific criteria:
- Product-Specific Criteria: For customer service robots, good performance means quickly solving problems, while for chatbots, it means providing empathetic companionship. For example, an airline's AI customer service should be friendly and focused on helping users with their travel plans.
- Breaking Down and Scoring Content: Good content is evaluated based on dimensions such as factual accuracy, tone, and context (e.g., what makes a response score 7 points instead of 10? Is the source missing? Is the tone too mechanical?). This process is similar to how journalists edit articles—asking questions like, "Have you addressed the user's real concerns?"
4. The Relationship Between AI and Creators
Is AI a threat to creators' jobs or a new tool for collaboration? The situation is more complex than it might seem:
- The Gig Economy Is Not AI's Fault: Many media professionals have always taken on side jobs (e.g., driving for Uber or delivering food). AI simply provides another form of gig work, and the root cause lies in the difficulty of traditional media industries sustaining themselves.
- AI as a Tool, Not a Competitor: When Tony was ill, he used AI to co-host a podcast and discovered that it helped him develop new creative ideas. Bianca noted that while AI can't produce groundbreaking art like The Instant Universe, it can assist creators (e.g., by providing a draft script for them to refine).
- Collaborative Path: Tony and the AI host he trained worked together on the podcast, demonstrating that AI can be a partner in creative processes.
5. Cross-Cultural Adaptation
AI often struggles with cultural nuances. For example, translating "Meryl Streep won an Oscar" directly into Chinese doesn't convey the same meaning. Media professionals can help by providing context (e.g., explaining that Meryl Streep is equivalent to Siqin Gaoawa in Chinese cinema, and the Oscars are similar to the Golden Rooster Awards) and understanding subtleties in language and tone.
Conclusion
The emergence of content engineers marks a crucial step in AI's evolution from being able to communicate to communicating effectively. Humans with a background in humanities can use their expertise to help AI become more human-like, while traditional media professionals can find new roles in the AI era. AI is not an enemy of creators but a partner that can complement their work.