cool research idea for someone: text diffusion in embedding space solve any sequence-to-sequence task in three steps:
1. embed source sentence text
2. build diffusion model that maps input text embedding to target text embedding
3. invert to produce target text
INNOVATION
-

Text Diffusion in Embedding Space for Sequence Tasks
By
–
-
Satya Levels Up: Microsoft’s AI Strategy Beyond OpenAI’s Constraints
By
–
I was close on this call, except Satya leveled up and did it without the burden and limitations of the OpenAI non-profit structure and charter.
-
Meta’s Vision for the Future of Artificial Intelligence
By
–
And now: Meta's view on the future of AI
-
Science Foundation Before Technology and Products
By
–
I'm not talking about products.
I'm not even talking about technology.
I'm talking about science. Before you build products, you need technology.
Before you build technology, you need science. -
AI Success: Synchronized Timeline or Collective Failure
By
–
You're wrong.
Either everyone will succeed within a year of each other, or no one will. -
Microsoft Teams Adds AI Backgrounds Moving to Metaverse
By
–
#Metaverse https://
pcworld.com/article/213669
8/microsoft-teams-adds-ai-backgrounds-as-it-moves-to-the-metaverse.html
… -
BITS Pilani Awards 6600 Degrees Through Metaverse Platform
By
–
6,600 working professionals get BITS Pilani degrees in #Metaverse https://
thehindubusinessline.com/news/education
/6600-working-professionals-get-bits-pilani-degrees-in-metaverse/article67550973.ece
… -
Yi Open-Source Project on GitHub – Upcoming Releases
By
–
Star us on @github
, and stay tuned for more Yi open-source releases coming up! -

Musk’s 50B datacenter investment mobilizes AI infrastructure
By
–
Musk is correct: about 50B expected in datacenters, representing the mobilization of AI (Musk rather focusing on rockets and vehicles). I factor in the US$44B Musk invested in X/Twitter. https://
markets.businessinsider.com/news/stocks/sa
m-altman-openai-fired-joins-microsoft-stock-price-satya-nadella-2023-11
… -

SDXL Model with LCM and ControlNet Achieves 4x Faster Generation
By
–
This is a result from a model that combines LCM, SDXL, custom loras and multi-controlnet, ~4x faster. 1024×1024, 8 steps, HED controlnet: 3.86s