Jonne Sälevä (@jonnesaleva) profile picture

CS PhD student at @nlpforall, focusing on low-res + morphologically rich NLP. Ulko-🇫🇮-lainen.

Internet
Joined July 2010
Photo by Jonne Sälevä (@jonnesaleva)Photo by Jonne Sälevä (@jonnesaleva)Photo by Jonne Sälevä (@jonnesaleva)Photo by Jonne Sälevä (@jonnesaleva)Photo by Jonne Sälevä (@jonnesaleva)Photo by Jonne Sälevä (@jonnesaleva)Photo by Jonne Sälevä (@jonnesaleva)
Time to update our priors on the directionality of polling error
Another fairly grim night for our polling technologies. The magnitude of the polling miss is not the point--all the mistakes are made in the same (partisan) direction.
1
165
📢 Looking for annotators! 📢 Any Finnish speakers who would like to assist in a #NLProc project about LLMs? We have about 300 data points that we need a human annotator for. Please get in touch via DMs / by replying below. Thanks in advance!
1
1
139
"Nobel committee wanted to jump on the hype train." How much of this is on us as researchers though? Ultra-hyped results and sweeping claims of superintelligence based on a very narrow sets of benchmarks over years seem like a good way to get to a situation we're in...
68
Today seemed like a good day to read up on Hopfield nets, so I pulled out the #1 ML book I know: probml.github.io/pml-book Thanks @srush_nlp for using it as your course text at Harvard back in the day! found it by stalking the CS281 site after taking 181 :).
Yes! If you want to know more about Hopfield networks and Boltzmann machines (Hinton's extension of Hopfield), and how they relate to Ising models, PGMs, and EBMs, see sec 4.3 of my book (probml.github.io/pml-book/bo…)
1
172
Despite all the meme posts today this sentiment is very nice. Hear, hear!
One aspect of the ML community that I have always treasured is our openness to borrowing ideas from other fields. It's great that the physicists are happy with this, too!
1
69
Jonne Sälevä retweeted
Fixed it.
I was having an identity crisis when I learned ACL isn’t AI. If ACL isn’t AI but NLP is, should I still submit my NLP paper to ACL? Or worse... have I not been doing NLP at all?? Turns out I’m actually a physicist! BRB, off to claim my Nobel Prize for all my physics research!🤓
8
1
104
14,774
Jonne Sälevä retweeted
I just feel bad they made Hinton miss his MRI appointment. He's Canadian. His next chance is going to be in like 2 years.
4
10
193
13,086
guess we're all physicists now? :D wonder when will snap up the international prize in statistics as well...
1
1
62
when will ML folks snap up* lol
24
Jonne Sälevä retweeted
BREAKING: The Nobel Prize in Physics has been awarded to ptrblock for “fundamental contributions to physics”
33
98
14
2,302
152,973
The First Workshop on Natural Language Processing for Turkic Languages (SIGTURK 2024) happening now at #ACL2024 ! Swing by if you get a chance!
1
7
537
Next token prediction is just a convenient factorization of the joint distribution: logP(x1, ..., xT) = \sum_{t=1}^T logP(x_t | x_{<t}) IMO, what we really need is a way to regularize #LLM's at the sequence level. #LLM #NLProc
1
116
Suomen valtion virallinen kanta voidaan pukea sanoiksi: free Palestine. Eli kahden valtion politikka. Siinä ei ole antisemitismiä hitustakaan. Palestiinalaiset kuuluvat seemiläisiin kansoihin.
2
1
8
678
Is there a Zoom link for the plenaries? Can't seem find a clickable link on the schedule website, unlike the oral presentations which had links. #lreccoling24 #LREC
1
126
Jonne Sälevä retweeted
We are pleased to announce that CS faculty member Duygu Ataman (@_dataman_ ) has received a research award from Microsoft's Accelerate Foundation Models Research (@AFMResearch) program. Congratulations! microsoft.com/en-us/research…
2
4
12
1,198