Congratulations to BAIR faculty Alberto L. Sangiovanni-Vincentelli @Abrto12 on his election to the American Academy of Arts and Sciences!
RESEARCH
-
Live Data Sculpture Uses Orchestra Music to Feed Generative AI
By
–
A live data sculpture!
— Pascal Bornet (@pascal_bornet) 23 mai 2024
The live music of this orchestra is used as real-time data to feed a generative AI model
A spendid work by Refik Anadol#datasculpture #artificialintelligence #machinelearning #deeplearning #music pic.twitter.com/el7CO197mdA live data sculpture! The live music of this orchestra is used as real-time data to feed a generative AI model A spendid work by Refik Anadol
#datasculpture #artificialintelligence #machinelearning #deeplearning #music -
Baidu CEO: AGI still 10+ years away despite rapid AI progress
By
–
@Baidu_Inc CEO Robin Li at @VivaTech
: “People talk about AGI, they say it’s probably two years away, it’s probably five years away. I think it’s more than 10 years away[..] Everyone is shocked at how fast [it] evolved [..], but to me it’s still not fast enough. It’s too slow.” -
Thoughts on 8x22b Release Without 8x7b Model
By
–
its honestly a bit odd they uploaded 8x22b first without 8x7b. but surely it cant be far behind now esp if its really sparse upcycled (do we know for sure?)
-
SolarLLM Integration Tutorial Sharing on Google Colab
By
–
Love this. May I use #SolarLLM at https://
colab.research.google.com/drive/1hiGt4dB
WEZ8Xsb3fUtONg3EEbguumaZ0#scrollTo=IqM-T1RTzY6C
… and add it as a tutorial at https://
github.com/hunkim/SolarLL
MZeroToAll
…? I will include the original link and give credit. -
Solar-1-Mini-Chat-Ja: Fast Japanese LLM Outperforms GPT3.5
By
–
Please try `solar-1-mini-chat-ja`. It outperforms GPT3.5 and Rakuten LLM, a famous model in Japan, but it's 3x faster. You can use it right away at https://
console.upstage.ai -
French DeepMind Alums Raise $220M for Safe AGI
By
–
Holistic AI, a french startups, comes out with a $220m round bang (yes, they are DeepMind alums). It's a safety and governance focused startup going for 'Full AGI': https://
pymnts.com/artificial-int
elligence-2/2024/ai-startup-h-raises-220-million-for-full-agi/
… their message is basically… "an AGI company that won't make you sick and kill you!" -
LeCun discusses intelligence nature and corporate AI race impact
By
–
Great to sit down with @ylecun for the first time, to discuss the nature of intelligence, the direction AI development is moving and why the corporate race is going to slow it down..
-
MediSwift Biomedical Language Models Accepted to ACL 2024
By
–
MediSwift has been accepted into ACL 2024! MediSwift is the first suite of biomedical language models that employ sparse pre-training techniques to significantly reduce computational costs while outperforming existing models up to 7B parameters on benchmark tasks such as
-
NeurIPS Experiment Revisited: Quality Scores Don’t Predict Citations
By
–
Revisiting the revisit of the Neurips Experiment. Still no correlation between quality scores and impact as measured by citations 10 years after the conference. https://
inverseprobability.com/talks/notes/th
e-neurips-experiment-iwcv.html
…