AI Roundup 09/29 -> 10/06/2022

Last week's most interesting AI news in <2 minutes

1. Google launches its own high-definition text-to-video model dubbed Imagen Video (TechCrunch)

2. DeepMind's AlphaTensor discovers a faster way to do matrix multiplication (DeepMind)
   

3. Google Brain researchers release DreamFusion - a novel text-to-3D image model (Futurism)

   

4. Researchers from Tsinghua University have released CodeGeeX, a 13 billion parameter programming model (CodeGeex)
   

5. Google AI has released FILM: a method to create high-quality slow-motion videos from near-duplicate photos (Google AI)

5. Meta releases AITemplate, a revolutionary new inference engine achieving performance improvements up to 12x on NVIDIA GPUs and 4x on AMD GPUs (Meta AI)
   

6. Researchers from Meta publishes AudioGen - a model that can generate audio samples from textual input, extend an existing audio stream, or simulate background noise (Paper's Github repo)
   

7. Anonymous researchers release Phenaki - a text-to-video model based on sequential text commands that can generate videos that can be as long as multiple minutes (Paper's websi

   

8. Tech giants launch Speech Accessibility Project to Improve voice AI understanding for people with disabilities (Engadget)

Last week’s AI Tidbits