Building with AI? This session dives into the landscape of multimodal and large language models, helping you choose the right one for your application — whether it involves text, images, audio, or video. Gain insights into model strengths, limitations, and practical integration tips with Witthawin Sripheanpol at #FOSSASIASummit2025

Click the link below to watch on the FOSSASIA YouTube channel:
https://youtu.be/zDohfhP3kW0

#30DayChartChallenge Día 10: ¡Buceando en la Distribución del VIX!

En lugar de solo ver la línea del VIX, hoy analizamos su "distribución de probabilidad" por Presidencia de EE.UU. (Clinton -> Trump 2º). ¡La forma lo es todo!

Usando #rstats y #ggplot2, estas densidades facetadas nos permiten investigar:
* Modos Dominantes: ¿Cuál era el nivel "normal" de VIX (el pico más alto)? ¿Cambió mucho?
* Multi-modalidad: ¿Hay evidencia de múltiples estados de volatilidad (picos secundarios) dentro de un mismo mandato?
* Riesgo de Cola: ¿Qué tan probable era el "pánico" (VIX > 35)? ¡Compara las colas derechas!

Estos patrones reflejan los distintos regímenes de volatilidad y la percepción del riesgo sistémico. No es solo el nivel, ¡sino la "estructura" de la incertidumbre lo que importa!

Datos: Yahoo Finance via #quantmod.
Código: https://t.ly/kikdo

Google expands Search's AI Mode to millions more Labs users in the US, now with multimodal capabilities—allowing users to ask complex questions about images. #GoogleSearch #AI #Multimodal #ArtificialIntelligence #TechNews #GoogleAI #Innovation #SearchEngine #MachineLearning

NEWS: Meta has unveiled Llama 4, its latest AI model, featuring advanced multimodal capabilities that integrate text, video, images, and audio processing. This release includes Llama 4 Scout and Llama 4 Maverick, both open-source and designed to enhance Meta’s AI assistant across platforms like WhatsApp, Messenger, and Instagram. Is this a new benchmark in AI versatility?
#Llama4 #AI #Multimodal
https://ai.meta.com/blog/llama-4-multimodal-intelligence/

Gemma 3 4B it -QAT- q4 by Google is a pretty good multimodal model given its size and level of quantization. Even better it runs on CPU and is relatively fast. The description of the picture below was generated by it.

Great to learn new vocabulary!

OpenAI’s new AI image generator is potent and bound to provoke - The arrival of OpenAI's DALL-E 2 in the spring of 2022 marked a turning po... - https://arstechnica.com/ai/2025/03/openais-new-ai-image-generator-is-potent-and-bound-to-provoke/ #autoregressiveimagegenerator #multimodalimagegeneration #4oimagegeneration #aiimagegenerator #machinelearning #autoregressive #imagesynthesis #multimodalai #multimodal #chatgpt #chatgtp #dall-e3 #biz⁢ #dall-e #gpt-4o #openai #ai

Wow! I didn't really like Gemma 2, but Gemma 3, released today, is awesome. It comes in four sizes, 1b, 4b, 12b and 27b. It's super fast and except for the 1b version it can even handle images.

The 27B version apparently outperforms both DeepSeek v3 and LLaMA3-405 on the ChatbotArena benchmark.

It's also the first small model I've tested that's good at German.

#Gemma 3 Released with #Multilingual & #Multimodal Capabilities

Blog: https://developers.googleblog.com/en/introducing-gemma3/
Video: https://youtu.be/UU13FN2Xpyw

#ai #llm #Google

Is this [holding a butterfly shaped like S Moody Ave] #multiModal?

Amazing real time AI conversation demo. very close to what you could see in the HER movie: https://www.sesame.com/research/crossing_the_uncanny_valley_of_voice#demo
#ai #multimodal #conversation #model #sesame #HER

Microsoft announces Phi-4-multimodal and Phi-4-mini: https://azure.microsoft.com/en-us/blog/empowering-innovation-the-next-generation-of-the-phi-family/
#llm #phi4 #multimodal #release #instruct #ai

Soliaktion von #teilauto für #dvb.

#Multimodal ist das, was für die Herren Böhm und Zastrow immer außerhalb ihrer Vorstellungskraft bleiben wird.

https://teilauto.net/multimodal

AI Demos from Meta: https://aidemos.meta.com/
#ai #llm #multimodal #meta #facebook #demo #tech

From @thomassmits posting on bluesky @thomassmits.bsky.social‬:

Making Victorian visual culture searchable: Introducing an open-access and AI-powered dataset of 72,000 illustrations from the Illustrated London News (1842-1890) (w. Bethany Warner,
@pfyfe.bsky.social and @bcgl.bsky.social )

https://openhumanitiesdata.metajnl.com/articles/10.5334/johd.284

A 'Flask application enables scholars to use both textual and visual queries to search the 72,000 images' https://github.com/tpsmi/ilnmultimodalsearch

Code for extraction and multimodal AI embeddings: https://github.com/tpsmi/multimodaliln

In Ländern mit viel kombiniertem Rad- und Zugverkehr werden nicht die Stellplätze für Fahrräder in den Zügen ausgebaut - das würde nicht skalieren. Sondern es werden riesige Radabstellanlagen direkt an den Bahnhöfen gebaut, inklusive Stellplätzen für Bikesharing-Systeme, so dass auch Gäste mit dem Rad weiter können.

Hier z.B. der Fläche vor dem Bahnhof Antwerpen-Berchem. Auf der Rückseite des Bahnhofs gibt es noch ein Fahrradparkhaus.

Carradice Barley for a change, on the train home after a somewhat slow ride in very early this morning. My train turned up this times. #cycling #BikeTooter #commute #multimodal

Morgähn. Das #Brompton und ich sind heute mal wieder #intermodal unterwegs mit Frau #SBB. (Ich natürlich nur mit FFP3-Maske im Zug. Es ist voll. Niemand muss stehen, aber fast alle Sitzplätze sind belegt. Hier und da ist ein Schniefen zu hören. Bin froh um die Maske.)