We’re excited to announce that our MedSyn project is finally out after 2 years of dedicated work. Congrats to @yanwuxu6 and Li Sun! MedSyn is a fully controllable Generative AI that synthesizes anatomically realistic CT images using text prompts with extensive evaluation.

Six years ago, @yannakakis and I published a textbook on AI and Games. It has been used in universities across the world and become a standard reference for the field. Much has happened since, and it is about time for a second edition. Good news: we just put a draft online!

Google’s Gemini recently made waves as a major competitor to OpenAI’s GPT. Exciting! But we wondered: How good is Gemini really? At CMU, we performed an impartial, in-depth, and reproducible study comparing Gemini, GPT, and Mixtral. Paper: https://t.co/S3T7ediQLa 🧵

AI researcher's dilemma with preliminary results: Relinquish control of your precious p4d AWS instance and potentially not meet your deadline or burn through your budget while preparing the next round of experiments. #research #deeplearning #Transformers #GPU

💡Ongoing research considers the integration into GPT-like models by leveraging an AMI-like objective in context for language generation purposes for reducing hallucinations and toxicity. Future work will deal with the handling of discrete augmentation. 3/N #LLM #ACL2023 #GPT

miCSE is a mutual information-based contrastive learning framework that advances SOTA in few-shot sentence embedding. Learning embeddings with miCSE entails enforcing structural consistency across augmented views making self-supervised learning more sample-efficient. 2/N

Check out our paper "miCSE: mutual information Contrastive Sentence Embedding for Low-shot Sentence Embeddings" - at #ACL2023NLP - Poster Session 1, July 10th - 11:00am-12:30 pm. Joint work done with @moin_nabi. #LLMs #NLP ArXiv 📚: https://t.co/Sdw33QQT7a 🧵👇 1/N

Data visualization inspiration thanks to DALL-E: how Rothko, Basquiat, Picasso, and Monet would create an academic chart.

It’s deplorable. The Green Party supporters have succumbed to the fallacy that they can trick the law of physics, while sticking to their naive mantra of evil nuclear energy - ignoring technical advances - just to cater their Friday for a future electorate.

Today I launched a @UN crowdfunding campaign because we urgently need funds to start the emergency operation before it is too late. Learn more: https://t.co/H6q5MJWBDg

Great article about the current state of AI, and the different perspectives on how to move forward. #deeplearning #neurosymbolic #reasoning

„For the past decade youngish, urbanish, professionalish got a sweetheart deal from that whole mosaic of urban amenities that vaguely pretended to be tech companies. [..] The golden age of bougie on-demand urban-tech discounting has come to a close.“ https://t.co/QYcRqy7ANB?

ACL 2022 Virtual Poster Sessions ① Tuesday, May 24, 7 pm (GMT+1) VPS2: Machine Learning for NLP ② Wednesday May 25, 7 pm (GMT+1) VPS4: Machine Learning for NLP

"SCD: Self-Contrastive Decorrelation for Sentence Embeddings." Our ACL'22 paper proposes a non-contrastive self-supervised learning approach leveraging feature decorrelation and multi-dropout. https://t.co/Spu1578tSU #Transformers #NLP #acl2022nlp @moin_nabi

Regarding ACL, some graphs in case they’re helpful: cumulative score distributions for meta reviews (left) and paper avg reviews (right) for the Nov. deadline. Caveat: as with all conferences, decisions are not determined by scores alone, but rather a holistic view of the paper.

The good, the bad, and the ugly truth about AI not meeting expectations yearned for in medicine | #deeplearning #MachineLearning #AI #RealityCheck |

Information age snake-oil at its best: AI-powered emotion recognition. A nice synopsis on bias, self-fulfilling prophecy - culminating in an estimated multi-billion $ market - without scientific backing | #DeepLearning #MachineLearning #AI #AIsnakeoil https://t.co/5RPJ179oEx

Beautiful write up of the mRNA vaccine tech story and the arduous path of research | #COVID19 #mRNA #research #Science |

COVID-19 meets the never ending story of fail starts in the medical imaging AI community | #DeepLearning #medicalimaging #COVID19 #ArtificialIntelligence |

As much as I admire @elonmusk and his determination to change the 🌎. His reality distortion field becomes visible with @Tesla and their hyperbole of autonomous 🚗. #AI needs a paradigm shift rather than more data/engineers to create ever larger models.