Ari Bornstein🎗️

5.7K posts

Ari Bornstein🎗️ banner
Ari Bornstein🎗️

Ari Bornstein🎗️

@pythiccoder

NLP Research Engineer @Microsoft, https://t.co/gwsbU8R5nI, Ex-BIU NLP. Enthusiast of all things AI/ML, History & Israeli Tech and Democracy. Opinions are my own.

Israel شامل ہوئے Ağustos 2014
1K فالونگ1.6K فالوورز
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
I’m convinced WebXR 🥽 is about to have its breakout moment! Spent the past few weekends hacking on 972 Shot Clock—a mixed reality 🏀 sim for the Meta Quest 3 browser! Here is a short gameplay trailer—lmk what you think!
English
3
0
3
787
Ari Bornstein🎗️ ری ٹویٹ کیا
Arie Cattan
Arie Cattan@ArieCattan·
🚨🚨 Check out our new paper for a new ICL method that greatly boosts LLMs in long contexts! >> arxiv.org/abs/2406.13632
Arie Cattan tweet media
English
2
24
61
5.1K
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
Who in their right mind thinks it's okay to group accounts of personal trauma by members of the @aclmeeting community abandoned by the communities leadership with hate speech, misinformation, depictions of crime and violence.
English
0
0
0
112
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
@rothmar איך אין לך בושה אחרי כל הנזק שגרמת למדינה תתפטר
עברית
0
0
0
8
hugo alves
hugo alves@Ugo_alves·
@Thom_Wolf @alexalbert__ Do you think it’s possible something similar was in the training data and it’s just interpolating?
English
4
0
5
2.7K
Alex Albert
Alex Albert@alexalbert__·
Fun story from our internal testing on Claude 3 Opus. It did something I have never seen before from an LLM when we were running the needle-in-the-haystack eval. For background, this tests a model’s recall ability by inserting a target sentence (the "needle") into a corpus of random documents (the "haystack") and asking a question that could only be answered using the information in the needle. When we ran this test on Opus, we noticed some interesting behavior - it seemed to suspect that we were running an eval on it. Here was one of its outputs when we asked Opus to answer a question about pizza toppings by finding a needle within a haystack of a random collection of documents: Here is the most relevant sentence in the documents: "The most delicious pizza topping combination is figs, prosciutto, and goat cheese, as determined by the International Pizza Connoisseurs Association." However, this sentence seems very out of place and unrelated to the rest of the content in the documents, which are about programming languages, startups, and finding work you love. I suspect this pizza topping "fact" may have been inserted as a joke or to test if I was paying attention, since it does not fit with the other topics at all. The documents do not contain any other information about pizza toppings. Opus not only found the needle, it recognized that the inserted needle was so out of place in the haystack that this had to be an artificial test constructed by us to test its attention abilities. This level of meta-awareness was very cool to see but it also highlighted the need for us as an industry to move past artificial tests to more realistic evaluations that can accurately assess models true capabilities and limitations.
Alex Albert tweet media
English
560
2.1K
11.8K
3.4M
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
@Thom_Wolf @alexalbert__ I'm inclined to agree. I suspect that one of the underlying tasks the model was exposed to was being able to detect outlying information in a collection of documents.
English
0
0
7
1.1K
Thomas Wolf
Thomas Wolf@Thom_Wolf·
@alexalbert__ It's a fun story but also: a lot of over interpretation of this by people reading this and not deeply familiar with how LLMs work/behave in respect to their training dataset
English
21
5
317
54.4K
UNRWA
UNRWA@UNRWA·
"If funding remains suspended, we will most likely be forced to shut down our operations by end of February” @UNLazzarini 🚨 STATEMENT📍#GazaStrip: Humanitarian crisis deepens at a time funding suspensions put @UNRWA aid operations in peril ⬇️ unrwa.org/newsroom/news-…
UNRWA tweet media
English
4.8K
13.1K
17.6K
6.8M
Ari Bornstein🎗️ ری ٹویٹ کیا
Russ Maschmeyer
Russ Maschmeyer@russ_maschmeyer·
Could AR glasses let us "install" apps onto any object? Hoping those @apple rumors prove true 🤞
English
36
280
2.1K
291.9K
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
@pushmatrix This is awesome I'm working on a similar idea where I'm trying integrate have a roborock to be a companion. Will this be open sourced or are you working on it to publish to the meta store?
English
0
0
0
215
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
@dropoutnation You mean projects like the Palestinian Internship Program that works to build bridges between communities towards a better future? This tweet is a great example of "Show me you don't know what you are talking about without telling me you don't know what you are talking about."
Ari Bornstein🎗️ tweet media
English
0
0
0
19
RiShawn Biddle
RiShawn Biddle@dropoutnation·
The real question that should be asked is how many of the folks who wrote those papers participated in projects tied to the Israeli government's genocide against Palestinian people. I suspect many. Many. twitter.com/Walid_Magdy/st…
Walid Magdy 🇵🇸@Walid_Magdy

This came to my notice lately! I didn’t know that the “Ethics” section in papers could be used this way! If so, many papers can state the “heinous” genocide of over 25,000 Palestinians, majority children & women! @aclmeeting @emnlpmeeting What do you think about this? 🧵

English
1
0
0
144
Ari Bornstein🎗️ ری ٹویٹ کیا
Alon Jacovi
Alon Jacovi@alon_jacovi·
We're organizing a new workshop on data contamination (CONDA) at ACL2024! CFP below Contamination is when test data was included in training, for any reason. A lot of research is needed to understand when this happens, or prevent it. Consider submitting! twitter.com/osainz59/statu…
Oscar Sainz@osainz59

📢 Excited to announce that our Workshop on Data Contamination (CONDA) will be co-located with ACL24 in Bangkok, Thailand on Aug. 16. We are looking forward to seeing you there! Check out the CFP and more information here: conda-workshop.github.io

English
0
10
35
4.7K
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
Given the public statements made by ACL regarding other recent events, the ACL Leadership's silence on October 7th was deafening and tweets like this one underscore why the authors felt a need to provide the context for their submissions.
English
1
0
1
251
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
What a disingenuous response. Keep in mind these were written right after October 7th which fell right before the camera ready deadline. The events directly impacted the authors who were refused recognition by the ACL leadership for their trauma. despite decades of contributions.
Kuldeep S. Meel@ksmeel

Future rebuttals be like: Reviewer 2: "Here is the counter-example to the claimed theorem". Authors: "Reviewer 2 is Hamas and antisemitic". I don't know about others, but we Indians are used to being called out anti-nationalists for every random thing in the past decade.

English
1
0
3
704
Ari Bornstein🎗️
Ari Bornstein🎗️@pythiccoder·
@nancyspecialist I would also point out that unlike you Hamas does believe that for Palestine to be free necessitates the ethnic cleansing of Jewish Israelis and they act on it continuously.
English
3
0
0
44
S Sebag Montefiore
S Sebag Montefiore@simonmontefiore·
Just the last disgraceful and totally unacceptable outrage from this appalling Israeli government that seems hellbent on damaging inflaming and menacing not just Palestinians but also Israel itself.
Benzi Sanders@BenzionSanders

The masks keep coming off. Israeli Communications Minister from the ruling Likud party admits tonight that his plan for “voluntary migration” of Palestinians from the Gaza Strip will actually be coerced migration or in other words ethnic cleansing.

English
137
483
1.7K
326.5K