Group 1 - The CEO of Google DeepMind, Demis Hassabis, expressed surprise at OpenAI's early introduction of advertisements in AI chatbots, indicating that monetization through ads has been a cautious topic of discussion within Google [1][2] - Hassabis noted that while ads are a core revenue source for Google, DeepMind is not under pressure to make hasty decisions regarding advertising in AI chatbots [1][2] - He emphasized the complexity of integrating ads into AI assistants, as users expect to trust their assistants, raising questions about how advertising would function in this context [1] Group 2 - Users have shown significant aversion to ads in AI conversational settings, with OpenAI previously attempting to recommend applications during chats, which was perceived as intrusive [2] - Hassabis highlighted the fundamental differences between chatbots and search products, stating that user intent is clear in search scenarios, making ad integration more natural, whereas chatbots serve as digital assistants [2] - DeepMind is committed to a scientific and cautious approach in its development, stating that it will not make impulsive decisions due to short-term pressures [2]

Core Insights - Koah, a startup, recently raised $5 million in seed funding, focusing on integrating advertising into AI applications as a key monetization strategy [2][5] - The company aims to address the challenge of monetizing AI products, particularly in markets outside the U.S., where subscription models may not be viable [3][5] Group 1: Advertising Strategy - Koah is not attempting to introduce ads into ChatGPT but is focusing on "long-tail" applications built on large models, targeting user bases outside the U.S. [2][3] - The company has already implemented ads in several applications, including AI assistant Luzia and parenting app Heal, with notable brands like UpWork and Skillshare as advertisers [4] - Koah's advertising approach has shown a click-through rate of 7.5%, significantly outperforming traditional methods, with early partners earning $10,000 within 30 days of platform launch [4] Group 2: Market Insights - The CEO, Nic Baird, emphasizes that the current consumer AI landscape requires diverse revenue models, as reliance solely on subscriptions can lead to user fatigue and churn [5] - Baird's team identifies AI conversations as being positioned in the middle of the purchase funnel, highlighting the need to capture user intent effectively [7] - The company aims to understand user needs better rather than focusing on traditional banner ads, seeking to provide relevant recommendations [7]

来 源丨 新智元 （ID： AI_era ） 作者丨 KingHZ 为什么GPT-5会让老用户失望，但免费用户沸腾到要限流？ 为什么一个聊天模型，会突然懂得「花钱买成果」？ 为什么OpenAI用GPT-5在下一盘大棋，可能让广告公司和搜索引擎都睡不着？ 答案，全藏在一个看似普通的技术名词里—— 路由器 。 它不仅能决定GPT-5的「智商」，也决定了OpenAI的「钱途」。 GPT-5是真「天才」 对许多ChatGPT Pro和Plus用户等「真爱粉」来说，GPT-5让人失望，呼吁GPT-4回归。 你要真觉得GPT-5智商不行，那就大错特错了。 GPT-5 Pro智商高达148，真超越了人类天才的140！ 此外，在多个基准测试上，Epoch AI独立评估了GPT-5，表现非常亮眼： 在数学上，GPT-5表现尤其突出，刷新了FrontierMath和OTIS Mock AIME 2024-2025的最佳纪 录； 在编程测试SWE-bench Verified上，GPT-5进步明显，稍低于Claude 4系列； 在GPQA多选题测试中，逼近或略高于Grok 4和Gemini 2.5 Pro。 左右滑动查看 在推理和 ...

新智元报道 编辑：KingHZ 【新智元导读】GPT-5来了！智商148、数学纪录被刷新、英伟达点头认可，但OpenAI真正的杀手锏，不在智商，而在分配智商的「路由器」。 为什么一个聊天模型，会突然懂得「花钱买成果」？ 为什么OpenAI用GPT-5在下一盘大棋，可能让广告公司和搜索引擎都睡不着？ 答案，全藏在一个看似普通的技术名词里——路由器。 你要真觉得GPT-5智商不行，那就大错特错了。 GPT-5 Pro智商高达148，真超越了人类天才的140！ 此外，在多个基准测试上，Epoch AI独立评估了GPT-5，表现非常亮眼： 在数学上，GPT-5表现尤其突出，刷新了FrontierMath和OTIS Mock AIME 2024-2025的最佳纪录； 它不仅能决定GPT-5的「智商」，也决定了OpenAI的「钱途」。 GPT-5是真「天才」 对许多ChatGPT Pro和Plus用户等「真爱粉」来说，GPT-5让人失望，呼吁GPT-4回归。 在编程测试SWE-bench Verified上，GPT-5进步明显，稍低于Claude 4系列； 在GPQA多选题测试中，逼近或略高于Grok 4和Gemini ...