Here is the translation of the contents into English:
In this interview, Jeff Dean and Noam Shazeer shared their 25-year experience working at Google, witnessing the company's growth from its early days to its current success, and seeing tremendous changes in the field of artificial intelligence. They discussed their work content, challenges, and achievements, as well as their vision for the future.
Jeff Dean mentioned the current challenges faced by Google, such as how to handle models with hundreds or thousands of billions of parameters, which requires loading the entire model into memory. He also discussed the possibility of using hybrid expert models in the future, which could be allocated more flexible memory resources based on the frequency and cost of usage.
Noam Shazeer talked about the issue of open research results, saying that Google has published many groundbreaking papers in the past, such as the Transformer paper, which have had a huge impact on the entire industry. However, he pointed out the challenge of balancing the publication of research results with maintaining competitiveness. He also emphasized the importance of humility and being able to let go of bad ideas in favor of better ones.
Finally, the host asked how ordinary workers could achieve a career like theirs. Jeff Dean suggested that following industry trends and communicating with colleagues, as well as paying attention to research papers, is the best way to learn about new and interesting fields. Noam Shazeer emphasized the importance of humility again.
This interview provides some valuable lessons for workers, such as being able to let go of bad ideas and accept better ones, balancing top-down and bottom-up approaches, having flexibility in resource allocation, and explaining exciting research directions to guide teams towards valuable goals.
Translation
这个访谈中，杰夫·迪恩和诺姆·沙泽尔分享了他们在谷歌工作25年的经历，他们参与了谷歌从早期到如今的辉煌发展，见证了人工智能领域的巨大变革。他们谈到了他们的工作内容、挑战和成就，以及对未来的展望。
杰夫·迪恩提到了谷歌当前面临的挑战，如如何处理拥有数百亿或数千亿参数的模型，这需要将整个模型加载到内存中。他还讨论了关于混合专家模型，未来可能会根据专家的使用频率和计算成本更加灵活地分配内存资源。
诺姆·沙泽尔则提到了开放研究成果的问题，他认为谷歌在过去发布了许多的研究成果，如Transformer论文，对整个行业的发展起到了巨大的推动作用。然而，面临着如何平衡发布研究成果与保持竞争优势的问题。他还强调了谦逊的重要性，要能够放弃不好的想法接受更好的想法。
最后，主持人问道，对于普通的打工人来说，如何获得像他们这样成功的职业生涯？杰夫·迪恩认为，了解新的有趣领域的最好方法就是关注行业动态与同事交流，以及关注研究论文。诺姆·沙泽尔则强调了谦逊的重要性。
这个访谈提供了一些对打工人来说很重要的经验教训，如要能够放弃不好的想法、接受更好的想法；要平衡自上而下和自下而上的方式；在资源分配方面需要灵活性；向大家阐明有趣的研究方向也能引导团队朝着有价值的方向前进。
Reference:
https://www.youtube.com/watch?v=kNVnr1BMpKU