Awesome meeting Huzaifa! He did some really interesting work showing that model reprogramming can sometimes perform fine tuning for private ML. Check it out: https://
x.com/NicolasPaperno
t/status/1624118654278172692
…
RESEARCH
-
Model Reprogramming for Private Machine Learning
By
–
-
ChatGPT Passes Bar Exam According to CTO
By
–
The CTO of one company mentioned today that ChatGPT “has passed the bar exam”.
-
Top AI Research Papers Published This Week
By
–
Top papers among thousands of papers published in the last week in AI research.
-
Robust Algorithm Design Advances Machine Learning Efficiency
By
–
Robust algorithm design is important across the field of #ML. Today we discuss our innovations over the previous year that advance the state of the art by building algorithms with improved efficiency, performance and speed. Read more at https://
goo.gle/3jJFdUv -
SpeechT5 vs FastSpeech2 vs VITS: TTS Model Comparison
By
–
Ofcourse. It depends on what do you optimise for SpeechT5 TTS (AR) does sound better than vanilla FastSpeech2 (NAR) however worse than VITS/PortaSpeech. Significantly worse than TorToise TTS. However it is considerably slow.
-
ChatGPT vs Stockfish Chess Match Prediction
By
–
Some madlad decided to pit ChatGPT against Stockfish, the predominant AI chess engine
— swyx 🐣 (@swyx) 10 février 2023
see if you can tell who’s who, and predict who will win pic.twitter.com/zWGed293VjSome madlad decided to pit ChatGPT against Stockfish, the predominant AI chess engine see if you can tell who’s who, and predict who will win
-
Privacy Attacks on Split and Federated Learning Methods
By
–
Very nice to see this work by @sanjayatwork and @mointweets. It's always great to see attacks against methods that are only "intuitively" privacy preserving, like split learning and federated learning (without differential privacy on top). https://
x.com/NicolasPaperno
/NicolasPapernot/status/1624108058702905360
… -
Unified DNAS Compresses TinyML Models for Neural Processors
By
–
#TenstorrentTop10 research paper series: Unified DNAS for Compressible TinyML Models for Neural Processing Units by Fedorov, Matas, Tann, Zhou, Watmough and @tenstorrent VP of Machine Learning Matthew Mattina #neuralnetworks #udc #machinelearning https://
tenstorrent.com/research/tenst
orrent-top-10-udc-unified-dnas-for-compressible-tinyml-models-for-neural-processing-units/
… -
CICERO by Meta AI: Strategic Dialogue Generation Process
By
–
#CICERObyMetaAI generates strategic, free-form dialogue by: Predicting the moves other players are likely to make. Creating a plan based on those predictions. Generating messages. Filtering the final messages for clarity and value. Read more on our blog
-
Stability AI Open-Sources Checkpoints Thanks
By
–
Also Big Thanks to @StabilityAI and @RiversHaveWings for open-sourcing the checkpoints!