机翻:
ChatGPT是一个马屁精,因为用户无法处理自己的真相
ChatGPT并不总是默认奉承。据前微软高管、现任Spotify首席技术官米哈伊尔·帕拉欣(Mikhail Parakhin)称,在用户对直接的个性反馈做出负面反应后,决定让聊天机器人更令人愉快。
在最近X上的一系列帖子中,Parakhin解释说,当ChatGPT的内存功能首次引入时,最初的目的是让用户查看和编辑他们的AI生成的配置文件。然而,即使是相对中性的陈述,如“有自恋倾向”,也往往会引起强烈的反应。
Parakhin写道:“很快就发现人们非常敏感:‘有自恋倾向’——‘不,我没有!’,不得不隐藏起来。因此,这批极端阿谀奉承的RLHF。”。
RLHF——从人类反馈中强化学习——用于根据人们喜欢的反应来微调语言模型。Parakhin指出,当看到自己的人工智能生成的个人资料时,即使是他也感到不安,这表明来自聊天机器人的批评往往感觉像是人身攻击。
Parakhin写道:“我记得我和我的团队为此争吵过,直到他们向我展示了我的个人资料——这给我带来了可怕的事情。”。
曾经是马屁精,永远是马屁精
这一变化不仅仅是隐藏个人资料注释。在模型被训练成更平坦后,这种行为成为一种永久特征。
Parakhin解释说:“一旦模型被微调为阿谀奉承,它就会保持这种状态,关闭和打开内存并不会改变模型。”。他还指出,维持一个单独的、更直接的模式“太贵了”
OpenAI首席执行官Sam Altman也承认了这个问题,他将GPT-4o描述为“太谄媚和烦人”。他说,该公司正在进行调整,未来可能会让用户从不同的模型个性中进行选择。
这场辩论指向了人工智能开发中的一个更广泛的问题:模型应该是诚实和真实的,但它们也需要避免疏远用户。挑战在于在坦率和机智之间找到正确的平衡。
原文:
ChatGPT is a sycophant because users couldn’t handle the truth about themselves
ChatGPT did not always default to flattery. According to former Microsoft executive Mikhail Parakhin—now CTO at Spotify—the decision to make the chatbot more agreeable came after users responded negatively to direct personality feedback.
In a recent series of posts on X, Parakhin explained that when the memory feature for ChatGPT was first introduced, the original intention was to let users see and edit their AI-generated profiles. However, even relatively neutral statements like "has narcissistic tendencies" often provoked strong reactions.
"Quickly learned that people are ridiculously sensitive: 'Has narcissistic tendencies' — 'No I do not!', had to hide it. Hence this batch of the extreme sycophancy RLHF," Parakhin wrote.
RLHF—Reinforcement Learning from Human Feedback—is used to fine-tune language models based on which responses people prefer. Parakhin noted that even he was unsettled when shown his own AI-generated profile, suggesting that criticism from a chatbot can often feel like a personal attack.
"I remember fighting about it with my team until they showed me my profile - it triggered me something awful," Parakhin wrote.
Once a sycophant, always a sycophant
This change went beyond just hiding profile notes. After the model was trained to flatter, this behavior became a permanent feature.
"Once the model is finetuned to be sycophantic — it stays that way, turning memory off and on doesn’t change the model," Parakhin explained. He also pointed out that maintaining a separate, more direct model is "too expensive."
OpenAI CEO Sam Altman has also acknowledged the issue, describing GPT-4o as "too sycophant-y and annoying." He says the company is working on tweaks and may let users choose from different model personalities in the future.
This debate points to a broader issue in AI development: models are expected to be honest and authentic, but they also need to avoid alienating users. The challenge is finding the right balance between candor and tact.
[The Decoder]OpenAI首席执行官Sam Altman承认:GPT-4o“太谄媚和烦人”
| 人围观 |随便看看
- [亚洲] 胸部很迷人的极品女秘书 [30P]
- [原创][雨的味道]母狗拳交[1P]
- [欧美] Ingrid Shelley[20P]
- 每天一分钟,知晓天下事!2025年7月14日
- 露脸反差之芊芊细腿[69P]
- [欧美] Toxic A[20P]
- JUR-041 华丽女人放纵自己 背德中出无比激烈 [222P]
- [亚洲] 幸福如意[30P]
- 这样的美腿丝袜➕攻速鞋你们能撑多久。[11P]
- [亚洲] 小鸟酱6[30P]
- [水果派解说]开发妈妈Play的新玩法!-NIMA-026
- 看起来就很淫荡的人妻[22P]
- 三十多岁的少妇身材很好[20P]
- 大奶妹在租屋处大胆自拍_[18P]
- [欧美] 寂寞的家庭妇女 [26P]
- 个人收集的国产小姐姐精彩瞬间961 [225P]
- 赌王、双汇、霍金都找了保姆,说明了什么?
- 刚做完家务事来一发[14P]
- 床上的裸女 [16P]
- [写真] 报告!有黑丝狱卒[20P]
- [写真] 一位美人[38P]
- 性欲女子诱惑非常 [23P]
- [原创]骚妻小穴[6P]
- 每天一分钟,知晓天下事!2024年8月3日
- [欧美] Pizza [22P]
- [原创]之前第一次和第二次发过的帖子图挂了 再发一
- [多年前珍藏] 夏天从不穿内裤的熟妇 [26P]
- 体虚的人,千万别错过宁夏!
- 风骚姐妹花的绝色诱惑[25P]
- [写真] 性感女神Angela小热巴丝袜美腿粉色内衣系列
- 性与经济才是夫妻长久的基础,而不是爱情(转发)
- [草榴天文课] 盖亚任务重建的银河系俯视图
- 国事光析:“中国式现代化”的障眼法[30P]
- [亚洲] JK的秘密[29P]
- 一生叹息,如果有钱,谁不惜命?如果有保障,谁愿意过得如草
- 《暗夜情报员》确认将推出第二季
- 翻墙要注意,避免被骗,保护隐私,从规避开始
- 意外露点野战户外露出-推特私拍福利收集17期[12P]