Yunyao and I work in a research center of a major technology company. At this time of a year, we attend job talks every week. These talks are technical presentations given by candidates for our research or development positions, as part of their two day job interviews here. Not surprisingly, the majority of the talks are given by newly graduated computer science PhDs who have published a lot of technical papers in premier computer science conferences. After a while, we kind of see some common deficiencies of these new graduates. Below is an instant messaging conversation we had right after attending a talk, which I hope would be helpful for some people.

YY: How do you like the talk?

HH: good slides

YY: yes, very well made

HH: lack of big picture in question handling

YY: but he seems to have some problem explaining the details (to outsider)

HH: I think he lacks big picture. Those questions could normally be answered by a few simple sentences

YY: yeah

HH: but he tried to explain the details, which further confused people. All the questions were about the big pictures, but he couldn't even understand what people were asking

YY: that's typical when the work is all about formulas, because he puts slides in such a way to help people understand, but it is not clear what the exact output his program would generate

HH: right

YY: it would be better that he first goes through actual input/output, and then explain the meaning of the output

HH: yeah, I am afraid they have not thought about that

YY: we are not interested in blackbox. For example, it is still unclear to me how he ... (omitted)

(a few exchanges on technical details are omitted)...

YY: I didn't ask too many questions, because I didn't want to make the situation worse for him

HH: right, I think he should have stepped back from the stories of his papers. Do not use the same ones in the job talk, instead, create a coherent new story for the job talk. It is the same problem with B (another candidate who gave talk here before).

YY: yes, and don't refer back to the paper. I am not going to read it.

HH: right, the talk should be self contained and coherent. The stories of the papers are normally not coherently fit to one another.

YY: No. Anyway, I am under-impressed.

HH: he just tried to repeat the stories of his papers, which are tenuous at best.

YY: yeah, I think the fact that his slides use simplified examples made things worse because it confuses people

HH: right. Anyway, I think it is a common problem with new graduates. They didn't seem to think too much about the problems they were solving. Everything is paper driven, and they just can't tell a story straight.

YY: exactly. This is because most people are not driven by problems, but by writing papers.

HH: right. They have a hunch, an idea, and they start writing a paper about it.

YY: (two smiles)

HH: I think this conference driven publishing culture in computer science contributes to this problem

YY: I think so

HH: well, these are the things beyond individual's control, so we can't really blame the student. If they are lucky, they get good adviser or mentors, then it is less of a problem...

YY: yeah, the student start from scratch, so it's really up to the adviser

HH: well then students often look up the publications of faculty members. If one faculty member publishes a lot, students will swarm to him instead of another one who publishes less (but maybe with higher quality). So it is a self perpetuating problem. Anyway, there's nothing we (in industry) can do about it.

YY: yeah.


comments powered by Disqus