Work log
This is a daily log of how I'm spending my time at work.
I only track uninterrupted, focused activities (i.e. not randomly responding to email or Slack).
These are mostly done in hour-long Pomodoros.
2021-04-08
Rushed day today, with back-to-back meetings with Microsoft and Noah. I should try to find a fixed time for both.
|
3h
|
Prototyping the semantics extraction with GPT-3 and running experiments.
|
Microsoft
|
|
1h
|
Project meeting
|
Microsoft
|
|
30m
|
Team stand-up meeting
|
Microsoft
|
|
30m
|
Meeting with Noah
|
PhD
|
|
30m
|
Planning next steps after the meeting, writing tasks for next week.
|
PhD
|
|
30m
|
Chat with Mina
|
Life
|
|
1h
|
Chat with Megha
|
Life
|
|
1h30
|
Polishing & testing user study again, and kickstarted a pilot with 3 people on MTurk!
|
PhD
|
|
30m
|
Running a job with ConDagger on the ternary-addition domain and 3 runs.
|
PhD
|
2021-04-07
Worked on research today for the most part. Mostly engineering to be able to run experiments in batch and faster.
|
1h30
|
CoCoLab meeting
|
PhD
|
|
30m
|
Fixing the bug in counting how many steps from the environment an agent took.
|
PhD
Research
|
|
4h
|
Making a better job running infrastructure.
|
PhD
Research
|
|
1h30
|
Writing: expanded & polished the introduction.
|
PhD
Research
|
2021-04-06
Worked at Microsoft, met WenXin and did some writing. The jobs are taking too long - I'll need to make the experiment running process better somehow.
|
1h
|
Meeting with Alex
|
Microsoft
|
|
3h
|
Prototyping the semantics extraction
|
Microsoft
|
|
1h
|
Meeting with WenXin. She's making good progress in the fractions domain!
|
PhD
|
|
1h30
|
Writing the paper. Finished the structure of the introduction.
|
PhD
Research
|
|
30m
|
Babysitting the running jobs and checking partial results. Q-Learning
|
PhD
Research
|
2021-04-05
Worked more on Microsoft. I finally have my work setup there working (VPN, etc). Pushed a little on the paper. I might need a way to make the experiments run faster - they take days for each model/domain pair.
|
4h
|
Working & meetings
|
Microsoft
|
|
30m
|
Meeting with Suvir/Erdem/Dorsa
|
PhD
|
|
1h
|
Watched 229 lecture
|
PhD
|
|
30m
|
Babysitting running jobs. Fixed a small bug in Q-Learning making it crash in the equations domain.
|
PhD
|
|
1.5h
|
Writing the paper
|
PhD
|
2021-04-04
|
30m
|
Chat with an admit about experience in the lab
|
PhD
|
2021-04-03
Just caught up with 229 and helped WenXin.
|
1h
|
Watched 229 lecture
|
PhD
|
|
1h
|
Answered WenXin's questions through Google Docs and Slack.
|
PhD
|
2021-04-02
|
1h
|
Worked on the paper
|
PhD
Research
|
|
1h
|
Meeting with Noah
|
PhD
|
|
1h
|
HAI-Education meeting
|
PhD
|
|
2h
|
More onboarding setup
|
Microsoft
|
2021-04-01
Finally did some work for Microsoft, although I still don't have the credentials for some reason. On research, I fixed one silly bug that made the jobs crash overnight, and implemented the original DRRN baseline. It's running, and we should be able to look at the curves tomorrow. Tomorrow I plan to get back on paper writing, now that the results on the 3 domains we have should be more solid.
|
1h
|
Reading the Web demonstrations paper. It has an EM-based variational inference that I didn't fully parse. I should go back to this, seems like an important technical trick to understand.
|
Microsoft
|
|
2.5h
|
Played with Vega-Lite, the NL-Viz dataset and GPT-3
|
Microsoft
|
|
1h
|
Team meeting
|
Microsoft
|
|
30m
|
Meeting with Alex Polozov
|
Microsoft
|
|
2h
|
Implemented Deep Q-Learning agent
|
PhD
Research
|
|
30m
|
Reading about semantic parsing
|
Research
|
|
1h
|
Fixed crash from last night and babysat the running jobs
|
PhD
Research
|
2021-03-31
Still waiting on Microsoft, so just worked some more on the solver. I fixed the bugs that made the baselines break over Spring break, so now we should have initial comparison results by Friday.
|
1.5h
|
CoCoLab meeting
|
PhD
|
|
1h
|
Re-read the FlashFill paper (thinking about the Program Synthesis environment)
|
PhD
Research
|
|
2h
|
Debugged the solver agents and started all runs again
|
PhD
Research
|
|
1h
|
Finished implementing and deploying this time tracking infrastructure
|
PhD
|
2021-03-30
I still don't have my corporate login at Microsoft, which makes me stuck in just reading papers. Mostly worked more on the tutoring projects and a class I might take. I also met with WenXin again.
|
1h
|
Prepare for meeting with WenXin (reviewing the code base, making small demos, updating the project README and thinking of tasks for her)
|
PhD
|
|
1.5h
|
Meeting with WenXin
|
PhD
|
|
1h
|
Writing an email to WenXin with more details on next steps and some questions that she had
|
PhD
|
|
1h
|
Watched the first CS229 lecture. I'm familiar with most (but not 100%) of the class, so I'll see with the first homework if I can take it and not use too much time. If so, this clears out all my breadth requirements.
|
PhD
|
|
1h
|
Implemented the dynamic curriculum
|
PhD
Research
|
|
1h
|
Migrated my Web site to my own generation scripts, instead of using Hugo. Now I can make custom pages very easily, such as the time tracking page.
|
PhD
|
|
1h
|
Read the "Text-to-viz" paper
|
Microsoft
|
|
1.5h
|
Some onboarding trainings and documents
|
Microsoft
|
2021-03-29
Getting back from Spring break. First meeting with Microsoft, which was really cool - I'll be working with quite a few people from both MSR and PROSE (the R&D team at Microsoft). I spent a lot of time catching up on email and Slack, and was still getting back to the rhythm.
|
1h
|
Welcome meeting with Microsoft
|
Microsoft
|
|
1h
|
Onboarding tasks and emails
|
Microsoft
|
|
2h
|
Wrote a script to generate results from the runs I left running before Spring break. Realized the runs didn't work, and started debugging. Spent some time just loading all things back to the head again.
|
PhD
Research
|
|
40m
|
Meeting with Suvir, Erdem and Dorsa
|
PhD
|