#ML suggests all that relaxing #whale song might just be human-esque gossiping
"Contextual and combinatorial structure in #spermwhale vocalisations," was undertaken by MIT Computer Science and Artificial Intelligence Lab (#CSAIL) researchers and #ProjectCETI. The team used a relatively simple #machinelearning algorithm to analyze the communicative sounds #spermwhales make, thought to be a system of fixed messages – are actually much more complex than previously thought.
https://www.theregister.com/2024/05/09/ai_whale_language/
#CSAIL
»#Researchers at #MIT #CSAIL and Project #CETI believe that they have unlocked a kind of #spermwhale “#alphabet” with the aid of #machinelearning technologies.« https://techcrunch.com/2024/05/07/machine-learning-aids-in-discovery-of-sperm-whale-alphabet/?eicker.news #tech #media
MIT entwickelt bahnbrechende Lösung: AskIt reduziert die Komplexität der LLM-Integration in Softwareprojekte erheblich!
#KI #AI #askit #mit #csail #softwareentwicklung #integration #llm #dsl #promptengineering #codegenerierung #programmingbyexample #llms
https://kinews24.de/askit-geniale-innovation-fuer-llms-von-mit-forschern
Fluorescent Filament Makes Object Identification Easier https://hackaday.com/2023/08/17/fluorescent-filament-makes-object-identification-easier/ #3dprinterfilament #machinereadable #3dPrinterhacks #machinevision #infraredtags #infrared #mitcsail #qrcode #CSAIL #label #FDM #mit #ir
Trustnet/Reheadline and socially crowdsourced countermeasures against the AI-generated misinformation/disinformation deluge — @karger at #MIT #CSAIL #ImaginationInAction
Turing laureate Michael Stonebraker's exciting talk on developments in the DBOS world.
-
#DBOS #operatingSystems #technology #CSAIL #MIT #CambridgeMA #Boston
Packed early in the morning, #MIT #CSAIL + Imagination In Action
-
#computerScience #AI #machineLearning #CambridgeMA #Boston
Last was an intriguing talk by Marco Pavone on building trust in autonomous vehicles at #CSAIL. Integrating #LLMs into driving simulation is something I haven't heard before, although it strikes me as a more incremental improvement over existing methods than something likely to push #AVs to wide deployment https://www.youtube.com/live/HjOt-4k6haI?feature=share&t=760 (6/6) #AI
@blacklight @DrHyde @tyil The #JargonFile started being passed around and accumulating entries through the 1970s at #Stanford, #MIT, #CarnegieMellon, Bolt Beranek and Newman, and other pre-Internet computing centers. Some terms date back to the late 1950s MIT Tech Model Railroad Club, a very early progenitor of #hacker culture.
Far more than you ever wanted to know here: https://en.wikipedia.org/wiki/Jargon_File
Next was a nice talk by Monroe Kennedy III on collaborative #robotics at #CSAIL. There's interesting work here on robot dexterity and human prediction using a variety of #sensors https://www.youtube.com/live/ii8ZNXaZ0hg?feature=share&t=912 (5/7)
First was an interesting talk by Jeannette Bohg on scaling #robot learning for long-horizon tasks at #CSAIL. Robot manipulation over long time scales is challenging, and the novel approach to tackle this that was presented here combines YouTube videos, #LLMs, and skill libraries https://www.youtube.com/watch?v=Ca-CxLZ2mq8&t=1s (2/7) #robotics
Last was a great talk by Graham Neubig on evaluating and learning model #explainability at #CSAIL. The field of #MachineLearning explainability doesn't have good quantitative evaluation tools, and the work presented here, which aims to use ML to evaluate explainability, is quite compelling. I would like to see some evaluations of this technique with human raters to validate it more thoroughly, hopefully that's coming soon. Highly recommend https://www.youtube.com/watch?v=CtcP5bvODzY (5/5)
Next was an interesting talk by James Glass on recent progress in audio-visual language learning at #CSAIL. After a quick summary of the latest in the space, Glass presents some fascinating work using crowdsourced spoken image captions across multiple languages. The results aren't perfect, but demonstrate how multimodality might unlock training effective models with far less data than current #LLMs https://www.youtube.com/watch?v=6CfBMQ3cvJs (3/5) #MachineLearning
Last was an enthralling talk by @animesh_garg on building generally autonomous #robots at #CSAIL. The argument that making truly autonomous systems requires embodiment, pre-defined causal and world models, and data is explored from a variety of perspectives here, demonstrating some impressive results. Highly recommend https://www.youtube.com/watch?v=jHf8ysoqyCw (10/10)
Next was an engrossing talk on scene representation by Vincent Sitzmann at #CSAIL. There might be a slight technical barrier to entry, but the new ways to think about scene understanding and #visualization are brilliant and elegant. Highly recommend https://www.youtube.com/watch?v=vX6nsIvMYaU (5/8)
@SexyCyborg This was a pretty active area of AI for a while. I think Reid Simmons and Tom Mitchell, and/or their students at CMU where looking at learning-by-watching. Quite a bit of work also out of #CSAIL lab at #MIT.
The field of #AI has forgotten a lot.
Next was a more technical talk at #CSAIL by Rob Fergus on #data augmentation for image-based #ReinforcementLearning. This is an extremely hard problem, and the authors show some impressive progress https://www.youtube.com/watch?v=Ny2CpgPrtB8 (8/10)
32 years ago today Tim Berners-Lee and Robert Cailliau emailed colleagues outlining a proposal for a hypertext project that they dubbed "World Wide Web".
(v/ @w3c )
#TimBernersLee #RobertCailliau #hypertext #WorldWideWe
"WorldWideWeb: Proposal for a HyperText Project" 👇
Client Info
Server: https://mastodon.social
Version: 2025.04
Repository: https://github.com/cyevgeniy/lmst