language is irreducibly contextual and multimodal. phd @uwcse (nlp) & former eng @google, now indie dev. posting: academia ∪ "AI" ∪ building software ∪ travel
Unlike any sane person who gets a PhD in NLP right now, afterwards I made a game. I just released it in early access https://t.co/OSfNQZjiaE Talk to NPCs who talk back at you, try to persuade your way out of sticky situations
Talk to Me Human (my game) is done, out, released on Steam. I went unhinged with the bonus challenges in the second half to keep myself entertained. Spent so much time laughing hysterically in a room alone playtesting it
you CANNOT convince Vincent he's in actual danger
Your PhD advisor has at least five jobs. They’re probably really good at some and maybe bad at one or two.
It surprised me how close (though abstract) the analogies are to traditional roles from industry. Except teacher, of course.
An emergent skill of mastering any craft—using software, cooking, even playing a game—is knowing what's actually important. An expert builds a precise workflow using a tiny fraction of the available control plane. At a deeper level, they know on which aspects to focus, and which to ignore.
Here’s the link. Try the free demo if you’re still undecided!
"I've never played a game that had me dying laughing like that!" — real unsolicited feedback (!)
https://t.co/eON7FXKz7T
Hello, last chance for 50% off Talk to Me Human! Early access ($4.99) — which includes the complete game — is ending. I’m releasing the full version soon on Steam and web, and the price will bump to $9.99.
The most cursed word in talking about research is “interesting.”
If someone says “X would be interesting,” they probably mean “X hasn’t been done before” and that X is one of:
- just something they cooked up, it may not even be worth pursuing research-wise
- a fun application of an existing idea
- a fun application with a fundamentally new idea
- an incremental tweak
- a clever idea (i.e., an incremental tweak people will like)
- a potential component of a research paper
- a potential entire research paper
- a potential research redirection
- a total moonshot project
- research-adjacent, but not paper-driven; a more fundamental issue that they like to think about and discuss
You might reply, “Interesting!” and mean one of:
- I’m not paying attention and maybe want to leave the conversation.
- I don’t actually understand what you’re saying.
- I get the gist, but you didn’t use any examples / talked too fast / I haven’t thought about the problem enough, so I need time to process.
- I think you’re proposing category Y (above), but I think it’s actually a lower category Z (above), or it’s been done before.
> (Most common scenario here: I think that you are claiming this is a novel idea worthy of a research paper, but it’s actually almost the same thing as Foo et al. (2019), so I don’t think it’s worth pursuing.)
- I think you’re proposing category Y (above) and I agree.
- I think you’re proposing category Y (above), but I actually think it could be upgraded to a higher category (Z) above.
> This one is really fun and one of the best parts of research conversations.
I’m pleased to announce the release of Weatherspread, a web app for visualizing recent weather and tourism data!
I built this app with friends Tristan and Shelby to aid in planning travel and events. Usually you plan things far enough out (say, 3-6 months) that there’s no reliable forecast. So historical data is your best bet.
We’ve observed that the last few years is a much better indicator of what the weather will be like than most data sources which aggregate from 1900 or even 1950.
We’ve also added tourism stats, and synced it to weather data. This has been particularly eye-opening. It lets us see when a country experiences over-tourism (e.g., Croatia and Austria) relative to its population.
There’s lots more we’d like to build, but we’re releasing it now to get feedback. It’s free and there’s no ads, so if you’re interested and would like to support us, there’s a box to give your email. In the future, we may consider a premium app version or an patreon-like support option.
Thanks for checking it out!
complete blog post is now live! more tips, and reframed into 2-step process: (1) get accepted by page 1, (2) avoid rejection with the rest. download both PDFs too. linked in next.
working on a post that's basically "how to get a paper accepted," using a case study one of my own that went from reject (2.5, 3, 3) to accept (4, 4.5, 4.5) with just one week of revisions