Ph.D. Program in Computer Science, The Graduate Center, CUNY

04/01/2024

This is the best video about evolution of key ideas leading to current AI paradigm. Over the years, I have watched hundreds of videos and read numerous articles, including many original papers in AI field. In my view, this video is just on another level from everything else about these topics I have seen - so I strongly recommend it to you and your students:

https://www.youtube.com/watch?v=OFS90-FX6pg

There is a popular simplistic idea that systems such as ChatGPT only learned statistics of human language - i.e. how to predict the next world. But this is not correct. As this video carefully explains, AI systems "know" (i.e.,. can use) abstract concepts and all kinds of abilities that go well beyond correlating words.

From the video description:

"We delve into the early experiments with tiny language models in the 1980s, highlighting significant contributions by researchers like Jordan, who introduced Recurrent Neural Networks, and Elman, whose work on learning word boundaries revolutionized our understanding of language processing. It leaves us with a question: what is thought? Is simulated thought, thought? Featuring Douglas Hofstadter, Michael I. Jordan, Jeffrey Elman, Geoffrey Hinton, Ilya Sutskever, Andrej Karpathy, Yann LeCun and more.

This video explores the journey of AI language models, from their modest beginnings through the development of OpenAI's GPT models. Our journey takes us thro...

05/13/2019

Communicating Your Research through Comedy: A Talk by Kyle Marian
Thursday, May 16, 2019
3:00 pm - 5:00 pm
Graduate Center, Room 9204

About the Talk
How is laughter a symptom of connection, and how can academics use it strategically for impactful communication? When communicating your research, there are plenty of methods and media that allow you to connect your message to your target audience (whether they are the general public, policymakers, patients, or stakeholders). Each form has its strengths and its weaknesses and in this introductory talk, science communicator Kyle Marian will share why the art of comedy is a powerful tool for connecting with diverse target audiences, and the lessons academics and researchers can take from a comedian’s work to apply to what they do.

We also encourage you to attend the follow-up workshop to this talk, Translate Your Research into Stand-Up, on May 23.

This event is sponsored by the CUNY Central Office Career Success – Workforce Development Initiative.

About Kyle Marian
Kyle Marian is a science communicator & former physical anthropologist, now focusing her work on multimedia and performance science communication/public outreach. She has performed in and produced public lectures, general science podcasting, science blogging, talk radio, and provided workshops training speakers for public events such as TEDx and comedy storytelling. She produces a monthly stand-up show called The Symposium: Academic Stand-Up featuring academics & researchers she’s trained to translate their obscure research & work life into comedy for wider audiences. She is also the social media manager for Guerilla Science, an international organization bringing science to new audiences in unexpected ways. She has a passion for using comedy in science communication and has recently been training with the Upright Citizens’ Brigade to hone her improv and writing skills. Internationally, she’s performed academic stand-up comedy through the UK’s Bright Club community, even taking the BBC stage during the Edinburgh Festival Fringe in 2015.

Please RSVP
Please fill out our event registration form to let us know you’re coming.

05/01/2019

Computer Science Colloquium

THURSDAY, May 9TH, 2019

4:15pm – 6:15pm

Room 9205
Larry Moss, Indiana University
Bridging the Gap Between Logic and Machine Learning in Natural Language Inference
Abstract:

The field of natural language inference (NLI) has seen strong progress in recent years, especially after the advent of deep learning. The basic goal is to see whether one natural language (NL) sentence “follows from” another, and to do this on a computer, for sentences “in nature”. Reflecting my own background, I wondered if there was anything whatsoever which 2000+years of work in logic could contribute to NLI.

This talk details work on making a connection between logic and computational linguistics. It will touch on topics such as: combinatory categorical grammar and its syntax-semantics interference; work on monotonicity pioneered by Johan van Benthem; the typed lambda calculus; natural logic, and algorithms from it; datasets like SICK, FraCaS, and SNLI; and the bidirectional language model BERT.

The goal is to see whether theory scales up to practice. I won’t give away the end of the story here, but suffice it to say the we are finding this work to be both challenging and interesting.

05/01/2019

04/18/2019

MACHINE TRANSLATION TALK
FRIDAY APRIL 19, 2019
THE GRADUATE CENTER
365 FIFTH AVENUE
ROOM 7395 - 11:00 a.m.
Speaker: Marine Capuat

Lost in Translation: Neural Models for Low-Resource Machine Translation & Detecting Semantic Divergence Across Languages

Abstract:
Despite impressive progress in neural machine translation, natural language processing systems that truly break language barriers remain elusive.

First, neural machine translation performs poorly for many of the world’s languages. I will present recent work on designing differentiable optimization objectives that better exploit limited training data and improve translation quality in low resource settings.

Second, cross-lingual transfer in NLP often relies on the assumption that translation preserves meaning. I will argue that this assumption often does not hold in practice. At the sentence level, semantic divergences between source and target language are surprisingly common in parallel corpora, and these divergences negatively impact neural machine translation. At the lexical level, words in two languages rarely cover the exact same semantic space. I will present our work on characterizing differences in meaning by predicting semantic relations between words in different languages.

Joint work with PhD students Xing Niu, Yogarshi Vyas and Weijia Xu.

Bio:

Marine Carpuat is an Assistant Professor in Computer Science at the University of Maryland, College Park. Her research focuses on multilingual natural language processing and machine translation. Marine is the recipient of an NSF CAREER award, research awards from Google and Amazon, best paper awards at *SEM and TALN, and an Outstanding Teaching Award. She received a PhD in Computer Science and a MPhil in Electrical Engineering from the Hong Kong University of Science & Technology, and was a postdoctoral researcher at the Columbia Center for Computational Learning Systems.

03/05/2019

Computer Science Seminar

Universal Information Extraction

Heng Ji, Rensselaer Polytechnic Institute

Monday, March 25TH, 2019

3:30pm – 5:00pm

Room C198

Abstract:

The big data boom in recent years covers a wide spectrum of heterogeneous data types, from text to image, video, speech, and multimedia. Most of the valuable information in such "big data" is encoded in natural language, which makes it accessible to some people — for example, those who can read that particular language — but much less amenable to computer processing beyond a simple keyword search. Information Extraction (IE) and Information Retrieval (IR) on a massive scale share the same goal of creating the next generation of information access in which humans can communicate with computers in any natural language beyond keyword search, by extracting and presenting the important and relevant information embedded in big data. IE aims extract structured facts from a wide spectrum of heterogeneous unstructured data types. Traditional IE techniques are limited to a certain source X (X = a particular language, domain, limited number of pre-defined fact types, single data modality, ...). When moving from X to a new source Y, we need to start from scratch again by annotating a substantial amount of training data and developing Y-specific extraction capabilities. In this talk, I will present a new Universal IE paradigm to combine the merits of traditional IE (high quality and fine granularity) and Open IE (high scalability). This framework is able to discover schemas and extract facts from any input data in any domain, without any annotated training data, by integrating distributional semantics and symbolic semantics. It can also be extended to hundreds of languages, thousands of fact types and multiple data modalities by constructing a multi-lingual multi-media multi-task common semantic space and then performing zero-shot transfer learning across sources. I will also discuss possible research directions toward a symbiosis between universal IE and IR, using open-domain knowledge graphs constructed from this common space as an intermediate representation.

Bio: Heng Ji is the Edward P. Hamilton Chair Professor in Computer Science at Rensselaer Polytechnic Institute. She received her Ph.D. in Computer Science from New York University. Her research interests focus on Natural Language Processing, especially on Information Extraction and Knowledge Base Population. She was selected as "Young Scientist" and a member of the Global Future Council on the Future of Computing by the World Economic Forum in 2016, 2017 and 2018. She received "AI's 10 to Watch" Award by IEEE Intelligent Systems in 2013, NSF CAREER award in 2009, Google Research Awards in 2009 and 2014, IBM Watson Faculty Award in 2012 and 2014, and Bosch Research Awards in 2015, 2016 and 2017. She coordinated the NIST TAC Knowledge Base Population task since 2010, and served as the Program Committee Co-Chair of several conferences including NAACL-HLT2018.

03/05/2019

Computer Science Colloquium

Evidence and the Language of Betting
Glenn Shafer, Rutgers

THURSDAY, March 7TH, 2019
4:15pm – 6:15pm
Room 9205

Abstract:

In his Art of Conjecturing (1713), Jacob Bernoulli turned the calculus of betting into a calculus for evaluating evidence. This produced both the Bernoullian statistic and the non-additive evidential probabilities I studied in A Mathematical Theory of Evidence (1976).

We now teach Bernoullian statistics (confidence intervals, significance tests, p-values, etc.) to a million students a year in the United States. But recent studies have shown that no one understands it - not the students, not the teachers, and not the researchers who use it. This talk, which draws on my forthcoming book with Vladimir Vovk (Game-Theoretic Foundations for Probability and Finance, Wiley, May 2019), shows how we can improve this situation by bringing betting back into the picture.

Picture: https://zh.m.wikipedia.org/wiki/赌博

02/26/2019

Lecture：Amazon Far-Field Speech Recognition for Amazon Alexa: Challenges and Solution

Björn Hoffmeister
Skylight Room 9100
TUESDAY, FEBRUARY 26TH, 2019
3:00PM – 4:30PM
TUESDAY, FEBRUARY 26TH, 2019

02/06/2019

Wishing good beginning of the Spring semester to all our students and faculty!

02/06/2019

Computer Science Colloquium
Cooperation in Humans and Machines
Patrick Shafto, Rutgers University

THURSDAY, FEBRUARY 7TH, 2019

4:15pm – 6:15pm

Room 9206

Abstract:

Cooperation, specifically cooperative information sharing, is a bedrock principle of human intelligence. Machine learning, in contrast focuses on learning from randomly sampled data, which neither leverages others cooperation nor prioritizes a means for communicating what has been learned. I will discuss ways in which our understanding of human learning may be leveraged to develop new machine learning, and form the foundation of improved integration of machine learning into human society.

The Colloquium is supported by generous contributions from the Bloomberg, Information Builders, Inc., and Netlogic, Inc.

10/24/2018

New York Combinatorics Seminar - Friday October 26, 2018
http://userhome.brooklyn.cuny.edu/skingan/CombinatoricsSemina

Location: Room 4419, CUNY Grad Center (365 Fifth Avenue)
Time: 11:45 am

Speaker: Shadisadat Ghaderi (Guttman Community College, CUNY)

Title: The Matroid Intersection Conjecture

Abstract: The theory of finite matroids was introduced by Whitney to capture and generalize the concept of linear independence in vector spaces. This theory was later generalized to infinite sets in a series of papers and recently it attracts a substantial amount of attention. The most important open problem in the field is the Infinite Matroid Intersection conjecture proposed in 1990 by Nash-Williams. It says that each pair of matroids on the same ground set has the Intersection Property and is a suitable restatement in the infinite case of the well-known finite matroid intersection theorem of Edmonds. Some progress was made towards proving this conjecture, but it is still open even for finitary matroids on a countable ground set. We introduce two different methodologies to approach the conjecture. Our results enable us to prove that the Matroid Intersection Conjecture is true some families of matroids.

Sponsored by the Graduate Center's Math Department and Computer Science Department Fridays 11:45 am - 12:45 am in Room 4419 This seminar covers a wide range of topics in combinatorics and its applications. The CUNY Graduate Center is located at 365 Fifth Avenue (at the corner of 34th Street), New Yo...

09/14/2018

Wishing good beginning of the Fall semester to all our students and faculty!