以下是内容的中文翻译：
Instagram 负责人亚当·莫塞里（Adam Mosseri）分享了他对产品开发不断变化的格局、人工智能的影响以及他的领导哲学的见解。他强调了 Meta 团队结构的重大转变，即从大型专业团队（例如，由十三名 Android、iOS、服务器工程师、产品经理、设计师、数据科学家和研究员组成的团队）转变为更小的“小组”（pods）。
这些新的小组通常由四到六名全能型工程师和一名“产品专家”（product staff）组成，后者是传统产品经理角色的演变。这位产品专家是全能型人才，能够利用新的人工智能工具，执行传统上由设计师、数据科学家和研究员处理的任务。只有在需要深厚、新颖的专业知识时，才会引入专家。莫塞里认为，这些更小的团队能够促进更快的决策，并减少“委员会式设计”（design by committee）的问题。
他指出，这一转变正在模糊传统的职能界限，并引起专家们的焦虑。虽然资深专家仍将是必需的，但人工智能正在降低个人跨职能贡献的门槛，使工程师的工作更多地关乎规划和审查，而非单纯的代码编写，也让设计师能够编写程序，或数据科学家能够提出设计方案。莫塞里强调，人类大脑在需要“品味”、判断力和战略的领域仍将是最有价值的。尽管人工智能可以提供帮助，但定义愿景、设定战略方向以及应对复杂的限制（人员、竞争、监管环境）仍然需要人类的领导。他认为，如果人工智能“懒惰地”提供战略，它将产生可预测、缺乏灵感的结果。
莫塞里强调，“品味”是一项关键技能，随着“创造事物”变得更容易，其重要性正在上升。他认为设计师因其与生俱来的品味而极具价值，他相信这种品味难以自动化。在招聘时，他看重毅力、快速学习能力和自我意识。在快速变化的世界中，好奇心和“敢于尝试、不怕犯错”的意愿（他将其比作学习一门新语言，并愿意听起来很愚蠢）正变得越来越重要。随着团队规模缩小，那些只专注于管理超大型组织的角色可能会减少。
关于人工智能成本，莫塞里指出，虽然 Meta 目前对工程师使用AI没有令牌限制，但如果AI使用的“烧钱”速度接近或超过一名工程师的薪水，这种情况可能会改变。他预计AI价格将经历“过山车”般的波动，成本可能因使用量增加而先上升，最终则因前沿模型之间的竞争而下降。
谈到 Instagram 的算法，莫塞里驳斥了一个普遍的误解，即算法对用户的兴趣有详细的语义理解。历史上，它依赖于生成难以解读的关联的嵌入模型。然而，大型语言模型（LLMs）现在正使算法能够用自然语言描述这些兴趣，从而让用户有更多自主权来理解和调整“他们的算法”。他为算法推荐信息流辩护，而非按时间顺序的信息流，解释说按时间顺序的信息流会激励发布者过度发布内容，随着时间的推移降低用户满意度。他认为这是为了实现系统可扩展性并保持用户参与度所必需的权衡。
关于人工智能生成内容的兴起，莫塞里认为，尽管存在挑战，但它将成为 Instagram 的“顺风”。他认为，在一个充满合成内容的世界里，人们将越来越多地寻求人类的创造力、真实性和个人联系。Instagram作为一个重要的创作者平台，有望从这一趋势中受益。他主张对AI内容进行标注而非过滤，这样用户可以做出明智的决定，并了解内容背后的账户信息。他提到“塑料梦境序列”（plastic dream sequence）作为他喜欢的一种AI艺术，因为它具有清晰的美学和创造性视角。
莫塞里承认受到竞争对手的影响，特别是 TikTok，因为它能够通过基于探索的排名“让小众创作者脱颖而出”，从而帮助小众创作者找到受众。Instagram 已在此领域投入巨资，以提高原创性、突破性内容和时效性。
最后，莫塞里谈到了公众的审视和争议，这在他的职位上是司空见惯的。他相信透明度，即使面对强烈异议，也会向公众解释决策和权衡。他以发展的眼光看待批评，回忆起职业生涯早期的困境，并强调任何有争议的辩论都并非像表面看起来那样简单。在家庭方面，他为自己的三个年幼的孩子设定了严格的屏幕时间限制，要求他们通过努力“赢得”设备使用时间，并批准他们使用的应用程序。他也开始和他的大儿子一起进行“情感编程”（vibe coding），即利用人工智能来开发游戏，强调数字素养和创造，而非被动消费。
Adam Mosseri, Head of Instagram, shared his insights on the evolving landscape of product development, the impact of AI, and his leadership philosophy. He highlighted a significant shift in team structures at Meta, moving from large, specialized teams (e.g., "a baker's dozen" of Android, iOS, server engineers, PM, designer, data scientist, researcher) to smaller "pods." These new pods typically comprise four to six generalist engineers and one "product staff" member, an evolution of the PM role. This product staff member is a generalist who can perform tasks traditionally handled by designers, data scientists, and researchers, leveraging new AI tools. Specialists are brought in only when deep, novel expertise is required. Mosseri believes these smaller teams foster faster decision-making and reduce "design by committee."
This shift, he notes, is blurring traditional functional lines and causing anxiety among specialists. While senior specialists will always be needed, AI is lowering barriers for individuals to contribute across functions, making engineers' jobs more about planning and reviewing than raw coding, and enabling designers to program or data scientists to propose designs. Mosseri stresses that human brains will remain most valuable in areas requiring "taste," judgment, and strategy. While AI can assist, defining a vision, setting strategic direction, and navigating complex constraints (personnel, competitive, regulatory landscapes) still require human leadership. He posits that if AI *lazily* provides strategy, it yields predictable, uninspired results.
Mosseri emphasized that "taste" is a critical skill, trending up in importance as building things becomes easier. He views designers as highly valuable because of their innate taste, which he believes is difficult to automate. For hiring, he looks for grit, quick learning, and self-awareness. Increasingly, curiosity and a willingness to "try things and make mistakes" (likening it to learning a new language and being willing to sound foolish) are paramount in a rapidly changing world. Roles focused solely on managing very large organizations may trend down as team sizes shrink.
Regarding AI costs, Mosseri noted that while Meta currently has no token limits for engineers, this may change if the burn rate of AI usage approaches or exceeds an engineer's salary. He anticipates a "roller coaster" of AI prices, with costs potentially rising due to increased usage before eventually falling due to competition among frontier models.
Discussing Instagram's algorithm, Mosseri debunked the common misconception that it has a detailed semantic understanding of users' interests. Historically, it relied on embedding models that created illegible correlations. However, LLMs are now enabling the algorithm to describe these interests in natural language, giving users more agency to understand and adjust "their algorithm." He defended algorithmic feeds over chronological ones, explaining that chronological feeds incentivize content overload from publishers, reducing user satisfaction over time. He sees it as a necessary trade-off for a system that scales and keeps users engaged.
On the rise of AI-generated content, Mosseri believes it will be a "tailwind" for Instagram, despite the challenges. He argues that in a world abundant with synthetic content, people will increasingly seek out human creativity, authenticity, and personal connection. Instagram, being a major platform for creators, is well-positioned to benefit from this trend. He advocates for labeling AI content rather than filtering it, allowing users to make informed decisions and providing transparency about the accounts behind the content. He mentioned "plastic dream sequence" as an example of AI art he enjoys for its clear aesthetic and creative point of view.
Mosseri admitted to being influenced by competitors, particularly TikTok, for its ability to "break small talent" through exploration-based ranking, which helps niche creators find audiences. Instagram has invested heavily in this area to improve originality, breakout content, and recency.
Finally, Mosseri addressed public scrutiny and controversy, a constant in his role. He believes in transparency, explaining decisions and trade-offs to the public, even when facing strong disagreement. He puts criticism in perspective, recalling early career struggles, and emphasizes that no contentious debate is as simple as it often appears. On the home front, he implements strict screen time boundaries for his three young children, requiring them to earn device time and approving their apps. He's also begun "vibe coding" (using AI to build games) with his eldest son, emphasizing digital literacy and creation over passive consumption.