Spence Green

Today I listened to Profs. Daphne Koller, Andrew Ng, Sebastian Thrun, and Fei-Fei Li talk about how to succeed in grad school. The following agenda was set, although not all topics were covered during the hour:

• Being a good grad student
  • What commonalities do great grad students share?
  • Clearly a lot of effort is required, but what sets the best apart?
  • What are common pitfalls that younger grad students encounter?
  • How can undergrads and masters students best prepare for a successful graduate career?
• Mentoring and Recruiting
  • Advising students is one of the most important roles of a professor, and something that is difficult to learn.
  • How can you best direct the research of collaborators and other students (like CURIS)?
  • What are good methods for recruiting good students to supervise?
  • More concretely, what are some tips for interviewing potential students?
• Life as a faculty member
  • What are your least favorite aspects of being a professor?
  • How does being a professor compare with other jobs like working in an industrial research lab or a startup?
  • Professors seem even more busy than grad students. Is it difficult to “have a life”? Sorry if this one sounds rude.
• Getting a faculty job
  • What sets apart great faculty candidates from mediocre ones?
  • The obvious answer is the quality of your research, but what other aspects set people apart from the pack?
  • What are the common weaknesses of faculty candidates?
  • How do you give a good job talk?
  • Reference letters are also important – what can grad students do to get good letters?
  • What are some of the resources people use to find and apply for faculty jobs?

Q: What questions do you ask Curis/undergrad assitants/students?

Andrew: what is your most significant project?

Seb: How well do you pay attention to detail?

Q: Most important aspect of career?

Fei-Fei: Most important thing is your recommendation letters. Way you have collaborated, solved problems, attacked problems, etc.

Seb: Need four really enthusiastic letters. Not letters from people who know you “fuzzily”. When you request a letter, ask if the person can write you a great letter. One good letter, and a few fuzzy ones looks bad.

Daphne: If we receive one enthusiastic letter from the advisor, and then some fuzzy letters from colleagues, that looks bad. Need to demonstrate a “reach” outside of your home institution.Seb: Trick for hiring. Give a talk at a target institution in the fall, so that you might get hired back in the spring.

Fei-Fei: Never miss an opportunity to give a talk. Go to someone’s poster and genuinely talk to them. Ask good questions, and demonstrate genuine interest.

Daphne: Content of your conversation must have substance.

Andrew: Avoid Stanford people at conferences 😉 Talk to people from other schools!

Q: What marks a good conference talk?

Sebastian: Less is more. Don’t cram all that you know into a talk. Most people can only give you 2 sentences on the talk. Put your name at the bottom of each slide so that people who walk in late will know you. Biggest trick: make people think about a problem before you give them an answer. Engage them.

Daphne: Lead the audience through the path that lead you to the solution. Here’s the problem. Here’s the naive solution. Here’s one approach, here’s another, here’s a problem we fixed. Use chronological narrative in the talk.And: Don’t film entire talks. Give 30 minutes or 1 minute of your talk, watch it, and iterate.

Sebastian: Connect with the audience. Don’t think about yourself, think about your audience’s response to your work. For example, putting a big equation on the screen will put your audience to sleep.

Daphne: Watch out for the visual elements of talk. Big fonts, main points are written on the slide. Minimize the number of words. Add images, reduce text. Formulas will not be absorbed in a talk. Want to convey: problem is hard, solution has insight, analysis gives some striking points.

Q: Biggest observed problem amongst grad students / academics?

Seb: Insecurity. Don’t cram your talk full of stuff. Life is short.

Q: Algorithm for doing research/phd (Andrew)

Step1: Basic technical skills.

Step2: If you have a good idea, work on it. If you don’t have an idea, go read papers.

Fei-Fei: How much do you want it? All of us are smart, good technical skills, etc.

Daphne: If you aren’t good at a particular technical skill, break it down, take classes, etc. Attack.Andrew: Goal is to work hard for six years, not just over the weekend.

Q: Time management skills

Andrew: Very few of us realize how much time we actually spend working, and not g-chatting, watching videos, context-switching, etc. Separate out other activities that waste time.

Fei-Fei: Don’t pull all-nighters. Need to sleep.

Daphne: Progress not linear in the time invested.

Andrew: Book recs: “7 Habits of Highly Effective People”. “Getting things done.”

Daphne: “Getting things done” changed my life.

Q: Revisiting old conference papers? A consistent body of work v. smaller contributions?

Daphne: Common to be unenthusiastic about old work. But you don’t want your resume to have a bunch of one-offs that don’t fit together as a whole.

Andrew: Think about working with another grad student.

Fei-Fei: I don’t believe in quantity. Much more important to do impactful work. Need to be excited about the problem first, then go from there.

Daphne: Three of four papers that combine into a single contribution. Not 3-4 random papers.

Q: Interdisciplinary problems?

Fei-Fei: Can’t be half-good in both. Need to basically get two Ph.d’s and be excellent at both. Don’t want to be a consumer of e.g. NIPS inference algorithms, and nobody in NIPs knows you, and nobody in e.g. neuroscience knows you.

Daphne: If you want to be successful in both fields, need scientists in both fields to recognize your work. So you need to think about where you publish. Publish your work in highly recognized, peer-reviewed journals in the community to which you belong. For example, if you want to get hired in computer science, then you need to be doing more than linear regression, t-tests, etc.

Fei-Fei: Fei-Fei (computational neuroscience) and Daphne (computational biology) recognized independently in other fields.

Andrew: Try to avoid strategizing about how to get jobs. Instead, spend your time doing good research.