jake vasilakes (@JakeVasilakes) header
jake vasilakes (@JakeVasilakes) profile picture

jake vasilakes

@JakeVasilakes

PhD student in NLP for trustworthy science communication @nactem_unimcr / @idsai_uom #BLM #GreenNewDeal

 Manchester, UK via St Paul, MN
jvasilakes.github.io
Joined August 2020
  • Tweets366
  • Following255
  • Followers59
My first ever tweet is a gentle reminder that language models like #GPT3, #BERT, etc. should never be used as knowledge bases. They may "know" that "Birds have feathers" but they also "know" all the racist bs the redditors they were trained on believe.
1
2
The amount most cities spend on annual fireworks displays could probably house a large portion of their homeless population for a year.
This Post is from an account that no longer exists. Learn more
2
Does anyone know of a decent survey on #NLProc for story/narrative understanding?
Soldering USB-A: this is actually relaxing. Maybe if the PhD doesn't work out I'll go into something with electronics. Soldering USB-C, unwittingly shorting the pins, plugging it in, and frying my comps front panel USB ports: well never mind then.
1
BREAKING my tomato plant has started growing it's first ever tomato. Updates in due course.
A small budding tomato on the vine.

ALT A small budding tomato on the vine.

2
Getting v tired of seeing NLP/ML researchers rehashing (with or without citing) previous work.
1
Vacancy for Research Associate in Natural Language Processing at @nactem_unimcr #NLP. For further details and to apply, visit: jobs.manchester.ac.uk/displa…
3
2
Most of the time I actually like doing my PhD remotely. Trying to navigate progression requirements is not one of those times.
Semantics is not distinct from pragmatics people
Gregg Caruso (@GreggDCaruso) profile pictureGregg Caruso
@GreggDCaruso18 Jun 2021
One for philosophers of language…
Image from Gregg Caruso (@GreggDCaruso)
2
Planted milkweed. Made a friend.
A monarch caterpillar on a milkweed stem.

ALT A monarch caterpillar on a milkweed stem.

2
My dad was disappointed when I started studying philosophy instead of STEM cuz he thought it was easy hippie shit. I had him read a passage from Husserl and he promptly changed his view.
This Post is from an account that no longer exists. Learn more
2
1
Honestly the only thing I've done in CS that matches the conceptual difficulty of understanding Husserl, Kant, Hegel, etc. is actual CS/NLP *research*.
Reading a bit of linguistics/phil of lang lately and I'm reminded of just how shallow and impoverished NLP currently is.
3
There are many negatives to living in the US American suburbs, but the constant lawn mowing sounds really grind my gears.
jake vasilakes retweeted
Sophia Ananiadou (@SAnaniadou) profile picture
Sophia Ananiadou@SAnaniadou4 Jun 2021
Congratulations to @amschoene. Pooling efforts @nactem_unimcr to work on mental health. I was awarded a grant for automating mental health interventions for long COVID. @idsai_uom @turinginst
1
6
Maximum efficiency grading with vim macros @+A = all tests pass! @+w = well done!
anarchism functional programming 🤝 state is inherently dangerous and should be eliminated
22
626
36
3,625
jake vasilakes retweeted
Anya Belz (@anyabelz) profile picture
Anya Belz@anyabelz12 May 2021
I'm looking for #PhD students to start this year or next who want to work on #NLG and/or #NLProc evaluation at @adaptcentre, @dcucomputing. Particularly neural network priming for language learning, or reproducibility in evaluation. Pls rt, pass on, or get in touch via email/dm.
11
165
3
250
jake vasilakes retweeted
Computer Facts (@computerfact) profile picture
Computer Facts@computerfact8 May 2021
ai is not "the future" ai is literally the past filtered through sketchy software regurgitated into an unwilling present
8
217
19
1,109
Always brilliant @SashaMTL and I looked into the darkest recesses of the @CommonCrawl to find what unsavory content large language models are trained on, and didn't like what we found! Our work was accepted into @aclmeeting and you can read about it here: arxiv.org/abs/2105.02732
What's in the Box? A Preliminary Analysis of Undesirable...

What's in the Box? A Preliminary Analysis of Undesirable...

Whereas much of the success of the current generation of neural language models has been driven by increasingly large training corpora, relatively little research has been dedicated to analyzing...

arxiv.org
1
9
1
42
Load more