Masthash

#Multimodality

Habr
1 week ago

OmniFusion: выходим за границы текста

Кто-то ещё сомневается, что в мире машинного обучения происходит революция? Уверен, мы являемся свидетелями преобразования привычного взаимодействия с данными, поиска информации, да и вообще работы как таковой. Ведь умные ассистенты (ChatGPT, GigaChat, Bard) готовы взять на себя даже самые сложные задачи. Но не всегда возможно сформулировать проблему в виде текстового запроса, иногда требуется информация из других “модальностей” — картинка, звук, 3D и тд. Ниже я разберу какие именно есть способы соединения больших языковых моделей (LLM) с дополнительными форматами данных, а также опишу как устроена наша новая модель OmniFusion.

https://habr.com/ru/companies/airi/articles/775108/

#чатбот #gigachat #multimodality #language_model

Habr
1 week ago

Kandinsky 3.0 — новая модель генерации изображений по тексту

Без чувства современности художник останется непризнанным. Михаил Пришвин В прошлом году на АI Journey мы представили модель Kandinsky 2.0 — первую диффузионную мультиязычную модель генерации изображений по тексту, которая может генерировать изображения на основе русскоязычного текста. За ней последовали новые версии — Kandinsky 2.1 и Kandinsky 2.2 , которые значительно отличались по качеству и своим возможностям от версии 2.0, и стали для нашей команды серьёзными вехами на пути к достижению лучшего качества генерации. Спустя год после релиза нашей первой диффузионной модели мы представляем новую версию модели генерации изображений по тексту — Kandinsky 3.0! Это результат длительной работы нашей команды, которую мы вели параллельно с разработками версий Kandinsky 2.1 и 2.2. Мы провели много экспериментов по выбору архитектуры и проделали большую работу с данными, чтобы сделать понимание текста и качество генераций лучше, а саму архитектуру — проще и лаконичнее. Также мы сделали нашу модель более «отечественной»: теперь она значительно лучше ориентируется в российском и советском культурном поле. В этой статье я кратко опишу ключевые моменты новой архитектуры, стратегию работы с данными и, конечно, продемонстрирую возможности нашей модели на примере генераций.

https://habr.com/ru/companies/sberbank/articles/775590/

#kandinsky_30 #multimodality #sberai #sberdevices #airi #generative_models #kandinsky #computer_vision #texttoimage #animation

UKP Lab
2 months ago

A warm welcome to Falko Helm, who has just started as a PhD candidate at the UKP Lab! 👋 Falko researches #Multimodality & Structure for #TransformerModels. He is also interested in #GraphTheory to handle linked documents. Find more about Falko here: https://github.com/Falko1

A photo of Falko with the headline "Welcome"
Trufi Association
2 months ago

Watch the discussion on enhancing active mobility! Experts D. Taylor Reich Carlosfelipe Pardo share insights on the power of open data and technology in this enlightening webinar recording. Data-driven decision-making and advocacy can transform cycling infrastructure.

https://tinyurl.com/yry7anl2

#opendata #opensource #sustainabletransport #ActiveTransport #ActiveMobility #Cycling #Walking #CommuteByBike #mapping #Multimodality #mobility #OpenStreetMap #VGI #multimodal #mobileapp #webinar

Trufi Association
2 months ago

🚴‍♂️ Join us this Thursday for a deep dive into the world of active mobility! 🚶‍♀️ Don't miss out on insights from experts Taylor Reich and Carlosfelipe Pardo. Discover the impact of data in shaping our cities. Register now:

https://www.linkedin.com/events/unpackingthepowerofdatainactive7109631505320071168/theater/

#opensource #sustainabletransport #transportation #ActiveTransport #Cycling #Walking #CommuteByBike #MobilityAsAService #mapping #Multimodality #mobility #OpenStreetMap #VGI #GTFS #GBFS #multimodal #webinar

Trufi Association
2 months ago

Active mobility is the future, and data is driving the way. Join our webinar with experts Taylor Reich and Carlosfelipe Pardo to explore the intersection of walking, cycling, and data.

https://tinyurl.com/2bpchc76

#opendata #opensource #sustainabletransport #transportation #ActiveTransport #ActiveMobility #Cycling #Walking #CommuteByBike #digitaldevelopment #Multimodality #mobility #populartransport #OpenStreetMap #VGI

Gaël Le Bris
2 months ago

Congratulations, MCO & Brightline! #Orlando International is now the first true U.S. #multiport: https://www.gobrightline.com/press-room/2023/the-countdown-is-on-brightline-orlando-to-officially-launch-service. More about multimodal #airport facilities in our upcoming #ACRP
#research report from Project 10-33 on the future of airport access... Stay tuned! ✈️🚄

#AirTravel #aviation #mobility #multimodality #STEM #transportation

Trufi Association
3 months ago

Are you a MaaS innovator? 📊 Trufi Association is your indispensable partner for GTFS and integration into apps, analytics, and transport tech solutions. Join us in revolutionizing transport for the global South and North!

#informaltransport #opendata #opensource #publictransport #sustainabletransport #transportation #ict4d #ict4dev #MaaS #MobilityAsAService #publictransit #mapping #Multimodality #mobility #populartransport #OpenStreetMap #GTFS

https://tinyurl.com/ync3fl4d

UKP Lab
3 months ago

We are pleased to welcome Aishik Mandal, who has just started his PhD at the Ubiquitous Knowledge Processing Lab! 👋 Aishik’s research concerns #Privacy-aware #Multimodality and #DialogueSystems. You can find out more about him here: https://jitaishik.github.io/

Trufi Association
3 months ago

Wrangling transportation data in the global South is no easy feat. It's our specialty. We harness community crowdsourcing and OpenStreetMap to provide accurate GTFS informed by the people who know the city better than anyone.

https://tinyurl.com/2gr8z78e

#bus #digital #digitaldevelopment #ict4d #ict4dev #informaltransport #MaaS #mapping #Mobility #multimodal #Multimodality #opendata #opensource #opensteetmap #publictransit #publictransport #PublicTransportation #sustainabletransport

Jonas Nölle
4 months ago

If you're interested in #multimodality and facial expression pragmatics come to our talk at #ICLC16 today in session 4 (3F), 11.45.
I'll present work from @facesyntax on how facial expressions can modulate the perceived confidence and doubt of spoken answers in English & Chinese.

Cover slide with title: Multimodal markers of confidence and doubt: Inferring Feeling of Knowing from facial movements
Result plot for Western (British English) participants
Abstract, which can be found at https://iclc16.github.io/abstracts/ICLC16_BoA.pdf on page 525
ISCA
4 months ago

Come “Share” With Us Our Enthusiasm… On Twitter!
Virginia Calabria & Sophia Fiedler

The International Society for Conversation Analysis, aka ISCA or @ISCAupdates, is an inclusive community with a global outreach.

Wherever people study #everythingILEMCA, they are part of the ISCA community. ISCA organises
https://www.conversationanalysis.org/come-share-with-us-our-enthusiasm-on-twitter/
#Uncategorized #academictwitter #EMCA #everythingILEMCA #ILEMCA #ISCA #LSI #multimodality

Jonas Nölle
6 months ago

Excited to share: after my current postdoc at the FACESYNTAX project finishes, I'll join @ozyurek_a at the Multimodal Language Department @mpi_nl next year!
We'll use #VR and mocap to probe multimodal communication and language evolution in unprecedented ways. Looking forward to pushing boundaries with wild, insightful experiments. 🕹️😎
#Multimodality #VR #Interaction #LanguageEvolution

Blurry image of a participant gesturing while wearing a VR headset.
EU AgencyForRailways
6 months ago

RT @JoDoppelbauer: Honoured to participate today at @STARS4Rail webinar.
Uptaking advanced technologies by SMEs in the #Railway sector can only boost its strategic importance within the #multimodality framework. https://t.co/Ulp97RWO4H

🐦🔗: https://n.respublicae.eu/ERA_railways/status/1664644568589934593

Inautilo
6 months ago

#Design #Outlooks
Decoding the Future: the evolution of intelligent interfaces · How we interact with technology is about to change significantly https://ilo.im/134u0j

_____
#IntelligentInterface #ProductDesign #UxDesign #IxDesign #UiDesign #DigitalDesign #WebDesign #AI #AR #VR #Ubiquity #ContextAwareness #Multimodality #Collaboration

Enjoying @LingLass's methodological talk on compiling a corpus of internet memes on health-related topics #ICAME44! #multimodality #CorpusLinguistics

Screenshot from the Zoom meeting showing a slide with six health-related memes, the speaker, and the conference room on-site.
GForester
7 months ago

Are you curious about multimodality in storytelling?
My latest blog post explores the benefits and challenges of using multiple modes of communication, such as images, text, and sound, to convey a message.
Learn how I used Instagram stories to create an immersive experience and engage a wider audience.

https://medium.com/@s5104976/multimodality-in-storytelling-905ccf9857e4

#multimodality #storytelling #instagram

Trufi Association
7 months ago

"Trufi Association collaborated with the Stadtnavi project to develop a native app for iOS and Android... the resulting platform is a testament to the benefits of collaboration and open-source development."

– Axel Thévenet

#development #project #android #ios #collaboration #opensource #sustainability #publictransport #multimodality

https://joinup.ec.europa.eu/collection/open-source-observatory-osor/news/stadtnavi-drives-cleaner-air-open-source

Roman Klinger
7 months ago

Reminder: I offer 3 positions for #postdocs/#PhD students in #emotionanalysis from text (1 w/ #argumentmining, 2 w/ #multimodality/#computervision). Apply until May 10, meet me @eaclmeeting. Details: https://www.romanklinger.de/projects/job-emcona.pdf and https://www.romanklinger.de/projects/job-item.pdf #NLProc #eacl2023

Alessio Pomaro
7 months ago

🧠 Greg Brockman talks about the #future of #ChatGPT in a TED talk.
⚠️ Will it be an everything app and the virtual assistant of our dreams? I don't know, but the new features are amazing!

#AI #GPT4 #LLM #ArtificialIntelligence #OpenAI #From #multimodality #TextToImage

https://www.linkedin.com/posts/alessiopomaro_the-inside-story-of-chatgpts-astonishing-activity-7056151317907464192-L78i

Me:
What would be the best hashtags to use for that post?

ChatGPT:
Based on the content of the post, the following hashtags may be appropriate:

#AGI #AI #ArtificialIntelligence #MachineLearning #DeepLearning #DataScience #AlphaZero #AlphaGo #MetaLearning #SelfCompetition #MultiModality #Robotics #PhysicalInteractions #Research

All right, here goes! I’m going to try my hand at a bit of #sociolinguistics #scicomm here and describe my latest publication for a lay audience. It’s my first try at this, so be gentle (but do ask questions if you have any!).

My study helps answer the question “What is identity, exactly, and how does it work?” It’s part of a long tradition of studies that have answered parts of that question, dating back several decades now. /1

#identity #multimodality #positioning #MembershipCategorization

So far I haven't really talked about my #research here, but I've seen other #academics announcing their books, so I think I'll at least try this out and see how it goes?

I'm one of the three editors of the new volume #Multimodal #Communication in #Intercultural #Interaction, and if you're interested in purchasing a copy for yourself (or more likely getting someone at your #university to do so), there's a hefty discount code in this flyer!

#Multimodality #Sociolinguistics #CognitiveLinguistics

20% Discount with Discount Code. MULTIMODAL COMMUNICATION IN Intercultural Interaction 

Edited by Ulrike Schröder, Elisabetta Adami and Jennifer Dailey-O'Cain 

Series: Routledge Studies in Language and Intercultural Communication 

This collection brings together a range of perspectives on intercultural communication in multimodal interaction, bridging cognitive, social and functional approaches toward promoting cros—disciplinary dialogues and taking research at the intersections of these fields into new directions. 

This book will be of interest to students and scholars in intercultural communication, multimodality, sociolinguistics, cognitive and interactional linguistics, and semiotics.

20% Discount Available - enter the code AFLO1 at checkout. This code expires on 30 June 2023. 

To request a copy for review, please contact: https/m.email.taylorandfrancis.com/Review_copy_request
Simone
11 months ago

Join us on zoom in January for a wonderful line-up of speakers at our workshop “Multimodal Digital Curating” Registration link and full programme👇
Please register here:
https://uni-koeln.zoom.us/meeting/register/tJwqceusrDstHdG68_57CwtvJ32NqzFRI4MH

Full program: https://agmedien.de/multimodal-digital-curating/

#multimodality #digitalcuration #mediaanthropology #collaboration #experimentation

Anu Lahtinen
11 months ago

Viime päivien Lintula-sekoilut palauttivat mieleen vuoden 2013, jolloin minulla mielestäni oli kanttia jopa sanoa jotain Twitterin merkityksestä tapauksessa "Tv-viihde sosiaalisen median pyörityksessä: Conan O'Brienin tapaus" Lähikuva-julkaisussa. Toiset ajat, toisen ajan some!

Soundtrack Charles Aznavour: “Yesterday When I Was Young”

#Lähikuva #ConanOBrien #TalkShowHistory #TwitterHistory #FanCultures #Multimodality

https://journal.fi/lahikuva/article/view/121170

Marco Magirius
1 year ago

I extended the deadline, so that everybody can enjoy their vacation.

#callforpapers #CallForPaper #multimodality #diversity #inclusiveness #literatureeducation #l1education

Details at https://www.ph-heidelberg.de/fileadmin/wp/wp-magirius/MagiriusCfP.pdf
Raffaella Bottini
1 year ago

How can we assess multimodal constructs❓Great discussion on #multimodality and #languagetesting during the Cyril Weir Lecture by @drjenrowsell@twitter.com #LTF2022 @UKALTA2@twitter.com

Francesco Ragazzi
1 year ago

Since a lot has happened lately, it could be a good time to re-do an #introduction / #introductions.

As a political sociologist I work with text, digital methods & film on matters of #security, #counterterrorism & #surveillance, from a critical perspective.

I currently run the project @securityvision on the politics of computer vision with amazing media artists/designers/visual anthropologists: Ruben van de Ven (@r), Cyan Bae, Ildikó Plájás & Elka Smith.

At Leiden University I'm trying to promote research through making (#film, #photography, #art & other methods) at @recntr, together with Mark Westmoreland and Julian Ross.

I geek out on photo & cinema #cameras & #lenses as well as #floss and #coding, although more from the outside. Working on my first "hello world" in #python

I'm likely to post on #security #computervision #biometrics #facialrecognition #masssurveillance #privacy #technology #film #cinema #multimodality #academia

Die Entwicklung geht von datenspezifischem Output zu allgemeineren KI-Modellen. Auch Meta sucht den KI-Gral und leistet mit data2vec einen multimodalen Beitrag.
Machine Learning: Auf der Suche nach einem Alleskönner-Algorithmus