谷歌发布开源编译器HEIR，通过同态加密技术实现加密数据上的直接计算，平衡AI应用中的隐私与安全，解决医疗、金融等领域数据保护与功能需求间的矛盾。

谷歌正通过同态加密技术推动隐私AI的实用化。该公司发布了开源编译器HEIR，可对加密数据直接进行计算，在保护用户隐私的同时实现AI推理。同态加密允许服务器处理加密数据并返回加密结果，而无需暴露原始信息，解决了传统加密在数据保护与服务功能之间的权衡问题。HEIR编译器能将预训练AI模型转换为处理加密输入，旨在让非专家也能轻松部署加密推理。谷歌已与多家硬件加速器公司合作，并联合多所高校开展研究，目前已有四篇基于HEIR的同行评审论文发表。为展示同态加密的进展，谷歌分享了四个应用案例：深度学习推荐模型、信用卡欺诈检测、网络流量异常检测和热词检测，所有示例代码均已开源。谷歌正致力于让同态加密更易开发、运行更快，并实现行业普及。

根据评论内容，主要观点和论据总结如下：

1. 对谷歌信任度低（评分：无） - 用户普遍不信任谷歌，认为数据在他人服务器上就不属于自己。 - 关键引用："Encryption or not, if it's on somebody else's server, it isn't yours. I don't believe Google has my best interest." (FloatArtifact) - "This is the same Google that doesn't have e2ee on their password manager by default." (Cider9986)

2. 技术可行性与商业可行性存疑（评分：无） - 全同态加密（FHE）计算开销巨大（约1000倍），商业可行性低。 - 关键引用："HE and other techniques have very high overheads(~10^3) on inference tasks and thus aren't very commercially viable." (sabretooth1405) - "Has FHE really progressed so far that it's now so efficient... I thought even basic operations like FHE addition were still thousands of times more complex." (bmenrigh)

3. 本地运行模型更优（评分：无） - 多数评论认为，本地运行AI模型比FHE更实际、更隐私。 - 关键引用："Private AI is practical by running the model locally, every much more so than any homomorphic encryption scheme." (nullc) - "The most private AI is the one running on my own hardware, not in some giant data center." (meindnoch)

4. 对FHE概念和安全性提出质疑（评分：无） - 部分评论质疑FHE是否真正安全，以及能否验证计算过程。 - 关键引用："Does this rely on the Trust Me Bro model, or is there some way for the client to verify that the provider actually isn't able to see your inputs?" (noident) - "One flaw with FHE is that it guarantees only that you need the key to see the inputs or outputs... but not necessarily that the computation is the one you want." (LoganDark)

5. 对谷歌动机的怀疑（评分：无） - 评论认为谷歌推广FHE是为了保持对AI模型的控制，而非真正保护用户隐私。 - 关键引用："I think they forgot 'or advertising' at the end." (traceroute66) - "Anything to keep people from running local models." (mohamedkoubaa)

6. 技术进展与历史对比（评分：无） - 有评论指出20年前FHE空间开销巨大（如1MB输出），质疑是否已有突破。 - 关键引用："I could do an encrypted (int)x + (int)y = (int)z computation... but was like 1MB in size." (drblast)

7. 能源消耗问题（评分：无） - 部分评论担忧FHE的高资源消耗会加剧环境问题。 - 关键引用："Great, private AI, at the cost of >1000x the resource usage... AI companies weren't already using quite enough energy to cook the planet." (meindnoch)

8. 对术语的批评（评分：无） - 有评论认为"全同态加密"是矛盾修辞，本质是降低安全性以换取可计算性。 - 关键引用："Proper encryption means the ciphertext is indistinguishable from noise. So...in order to be able to process on it, you have to make it not indistinguishable from noise." (filup)