Their model can also be run in reverse to guess real-world sounds from human vocal imitations, similar to how some computer vision systems can retrieve high-quality images based on sketches.
RESEARCH
-
AI Model Generates Realistic World Sounds and Audio Imitations
By
–
The model can effectively take many sounds from the world & generate human-like imitations of them — including noises like leaves rustling, a snake’s hiss, & an approaching ambulance siren. pic.twitter.com/q2sOB0Tl9d
— MIT CSAIL (@MIT_CSAIL) 5 décembre 2024The model can effectively take many sounds from the world & generate human-like imitations of them — including noises like leaves rustling, a snake’s hiss, & an approaching ambulance siren.
-
AI Model Simulates Human Vocal Tract and Speech Production
By
–
The researchers engineered their system to produce & interpret sounds much like we do. They first built a model of the human vocal tract that simulates how vibrations from the voice box are shaped by the throat, tongue, & lips. Then, they used a cognitively-inspired AI algorithm
-
O1 targets early adopters: PhDs, scientists, and engineers first
By
–
Not at all surprised that the o1 demos were extremely advanced and technical (and not business-oriented). They're catering toward first adopters. And those first adopters will be PhDs, scientists, and engineers.
-
Machine Learning and CT Imaging Predict Lung Abnormalities
By
–
Machine learning used with CT imaging foretells lung abnormalities
#AI #AIio #BigData #ML #NLU #Futureofwork @ahier @guzmand
@DavidBrin @denisegarth @dez_blanchfield @diioannid @DioFavatas @gerald_bader -
PaliGemma2: Exploring Google’s Latest Multimodal AI Model
By
–
Want to know more about it, read out detailed blogpost about it: https://
huggingface.co/blog/paligemma2 -
Token-level bit efficiency metric for model evaluation
By
–
I need to think about this more, but I'm proposing token-level calculation of that concept: sum(token_bits / token_bytes for _ in all_tokens) … and not: sum(tokens_bits) / sum(tokens_bytes) Because the second one would hide bad predictions for longer tokens in the sum(),
-
AI’s Future Impact on Workforce Integration and Equity
By
–
In the second issue of the Dialogues magazine from @Google and @atlanticrethink
, Maya Kulycky, Google Research VP, discusses how the future of work will depend on how AI is introduced to the workforce, to ensure everyone benefits. Read the full article ↓ -
Tokenization Boundaries and Cross-Entropy Loss Optimization
By
–
I found the exact tokenization boundary of the text and the corresponding token index. I can do the normalization but it's not perfect… (Testing it shortly.) Computing the cross-entropy like this rewards vocabularies with more tokens and easy predictions (on average) vs.
-
Open Buildings AI Maps African Urban and Rural Areas
By
–
From urban areas to rural settings, mapping buildings in Africa & the Global South presents unique challenges. The Open Buildings dataset overcame these hurdles w/ an innovative #AI model, creating a valuable resource for researchers & aid workers. More→ https://
goo.gle/4eUoLXS