cover of episode 72. 做实习的缺点?ICML x HCI 的参会体验 (嘉宾:Bryan Wang)

72. 做实习的缺点?ICML x HCI 的参会体验 (嘉宾:Bryan Wang)

2023/8/19
logo of podcast 李丁聊天室

李丁聊天室

AI Deep Dive AI Insights AI Chapters Transcript
People
B
Bryan Wang
李丁
Topics
Bryan Wang: 此次ICML会议是Bryan第一次参加机器学习领域的会议,主要目的是参与并协助组织AI and HCI workshop。他认为大型语言模型的出现使得机器学习领域开始关注人机交互,但同时也存在重复造轮子的现象,因此HCI和ML领域的交流合作至关重要。workshop的成功举办促进了两个领域之间的交流,他发现许多机器学习研究者都关注以人为中心的计算。他认为参加不同领域的会议能激发新的想法,拓展研究视野。 在实习方面,他总结了实习的三大要素:项目的多样化、人脉建立和地点选择。他分享了在Google、Adobe和Meta实习的经历,Google实习让他深入了解了HCI和AI的交叉领域,学习了机器学习的实践技能;Adobe实习让他在思想上得到启发,学习了如何看待创造力和问题;Meta实习让他认识到Meta的HCI团队规模庞大,但公司较为谨慎,审批流程较长。他认为不同公司有不同的文化和资源,会影响研究方向和项目选择。 地点选择方面,他认为实习地点的选择能帮助他更好地了解不同城市的生活方式,从而帮助他决定未来想居住的城市。他比较了旧金山和西雅图的生活,并表达了对Redmond的评价。 最后,他谈到了实习的缺点,即项目主题的选择受限于公司方向,可能难以找到与个人兴趣完全匹配的项目。 李丁: 作为一名全职研究员,李丁分享了他招聘实习生的两种方法:一种是选择最优秀的人才,然后根据他们的能力调整项目;另一种是根据公司项目需求选择合适的人才。在面试过程中,他主要考察候选人的研究能力、学习能力和项目匹配度。选择实习生时,需要平衡实习生的资历和项目需求,考虑项目的长期发展和人才培养。他认为,在不同阶段,对实习生的选择标准也会有所不同。早期阶段可以更开放,后期阶段则更注重与自身研究方向的匹配度。他同时还分享了Adobe公司鼓励学术合作的特色,以及在招聘实习生时,如何权衡senior和junior实习生的利弊。

Deep Dive

Key Insights

What were the key takeaways from Bryan Wang's experience at ICML?

Bryan Wang co-organized the AI and HCI workshop at ICML, aiming to bridge the gap between HCI and machine learning communities. The workshop was successful, with 60 papers presented and a balanced mix of AI and HCI researchers. He found it valuable to engage with ML researchers who are increasingly interested in human-centered computing, leading to new ideas and perspectives.

What are the three key elements Bryan considers important for internships?

Bryan emphasizes three key elements for internships: diversifying research portfolio, building connections, and experiencing the location. Diversifying research involves working on projects outside one's usual scope, connections help in networking and future opportunities, and location helps in deciding where to live post-PhD.

How does virtual internship compare to in-person internship in terms of building connections?

Virtual internships are less effective for building deep connections compared to in-person internships. Bryan found that in-person internships allowed for better bonding and understanding with team members, whereas virtual internships lacked this depth. He recommends choosing in-person internships if given the option.

What are the differences in internship experiences at Google, Adobe, and Meta?

Google offers abundant resources and perks, such as access to GPUs and massage benefits, focusing on AI and engineering. Adobe has a creative, art-focused vibe, influencing Bryan's research on human-AI systems for creativity. Meta, particularly in Reality Labs, provides strong support for AR/VR research but has more bureaucratic overhead in accessing tools and data.

What are the downsides of doing internships during a PhD?

The main downside is that internship projects must align with the company's interests, which may not always match the intern's personal research interests. This can limit the scope of projects and make it challenging to find a topic that satisfies both the intern and the company. It requires balancing diversification of research with personal and company alignment.

How does Bryan approach choosing a city for post-PhD life based on internship locations?

Bryan uses internships to experience different cities and decide where he might want to live post-PhD. He considers factors like food, weather, and lifestyle activities. For example, he found Redmond boring but enjoyed Seattle and San Francisco for their vibrant food scenes and outdoor activities.

What are the benefits of internships for PhD students according to Bryan?

Internships provide PhD students with financial stability, a break from academic responsibilities, and a focused time to work on projects. They also offer a chance to restart and gain new perspectives, which can be motivating. Additionally, internships help in building a diverse research portfolio and professional connections.

How does Bryan's manager at Meta approach project selection for interns?

Bryan's manager at Meta encourages branching out into projects that may not be directly related to the team's main research but still align with the company's broader goals. This approach allows interns to explore new areas and potentially expand the team's research scope.

Chapters
Bryan Wang, an HCI researcher, shares his experience attending the ICML conference, focusing on the co-organized AI and HCI workshop. The workshop aimed to bridge the gap between HCI and ML communities, highlighting the overlap in problem-solving and the potential for cross-disciplinary collaboration. His experience sparked new ideas and perspectives by interacting with researchers from diverse backgrounds.
  • Attended ICML conference to co-organize AI and HCI workshop.
  • Workshop successfully presented 60 papers.
  • The event fostered collaboration and knowledge sharing between AI and HCI researchers.
  • Many ML researchers are now more interested in human-centered computing.

Shownotes Transcript

嘉宾:Bryan Wang)

「在西雅图做研究实习」的第三期里,我和多伦多大学的 Bryan Wang 聊了他最近去夏威夷参加 ICML 的经历,他是一名 HCI researcher,在 ICML 这个 Machine Learning 会议里参会体验怎么样?

这个暑假 Bryan 在西雅图东区的 Meta Reality Labs 实习,之前他也在 Google 和 Adobe 实习过,他分享了这几件公司他的不同体验。

和之前的几期不一样的,我们两讨论到了暑假出来实习的缺点。同时,Bryan 也问了我,作为一名 Fulltime researcher,怎么选择实习生,怎么挑选项目的问题。

如果你是一名对研究,实习,业界工作有兴趣的同学,可以关注一下我的这个系列。可以在下方订阅。

  • 0:00 开场

  • 2:28 Bryan 参加 ICML 的体会

  • 7:25 实习中重要的三要素 diversify, connection, location

  • 16:30 Virtual 实习的挑战

  • 20:52 Meta/Google/Adobe 三家实习的比较

  • 25:44 Meta 实习的不足

  • 32:57 暑假出来实习的缺点

  • 35:30 作为全职研究员,李丁招 intern 时有哪些考虑?

  • 43:15 Bryan 结尾前安利实习

李丁的联系方式(喜欢程度按照这个顺序):

  • Signal (+1) 206-745-2849 强烈推荐 / 端对端加密 / 无审查 / 需翻墙

  • 邮件 [email protected]

  • Whatsapp (+1) 206-745-2849

  • Youtube 留言:https://youtu.be/XDqW41pQDfE

  • 微信 ID: lidingpublic 不喜欢微信 / 审查严重 / 用户体验差

联系李丁来聊聊天啊!

邮件订阅:https://liding.substack.com)

制作团队

  • 李丁泽宇

https://dingzeyu.li)