3/ What is LATS? 1. Select: the best scoring action thus far
2. Expand & act: Generate candidate actions and execute
3. Reflect & evaluate: Observe outcomes and score decisions
4. Backpropagate: update applicable trajectory scores So… Monte-Carlo Tree Search!
RESEARCH
-
LATS: Monte-Carlo Tree Search for AI Decision Making
By
–
-
Subobject-level Image Tokenization: Chen et al. Research
By
–
Subobject-level Image Tokenization Chen et al.: https://
arxiv.org/abs/2402.14327 #Artificialintelligence #DeepLearning #MachineLearning -
Vector Matching Engine Performs Fast Similarity Search on Embeddings
By
–
QBE (Query-By-Example) does feature vector similarity search. This is better… The @abacusai Vector Matching Engine performs searches on large stores of vector embeddings from text, images, documents & more, in milliseconds using #DeepLearning models: https://
abacus.ai/vectorstore -
Autonomous Agents vs Diffusion Transformer Models
By
–
autonomous agent will be better than diffusion + transformer model
-
SAS Innovate 2024: Leading AI Data Analytics Conference in Vegas
By
–
SAS Innovate is April 16-19 at ARIA Resort & Casino in Las Vegas. Register now to join this leading #AI / Data / #Analytics experience of 2024: https://
sas.com/gms/redirect.j
sp?detail=PLN22951_174944999
… by @SASsoftware ————
#SASInnovate #DataScience #ML #SASVisionary #ad #DecisionScience -
Interview Answer Prompt with Resume
By
–
10. Answering Interview Questions ChatGPT prompt: "Please provide an answer to the interview question: [Insert Question] using my resume [Paste Resume]. Make sure to use descriptive language and include measurable results."
-
PhD Defense on Robot Sounds: Ethnography and Pepper Experiments
By
–
Now this is a PhD defense! Congratulations to Dr. Adrian Latupeirissa from KTH on his amazing work on robot sounds – starting from ethnographic work on robots in films to live experiments on the Pepper robot in a museum. It was a pleasure participating on his examining committee.
-
AI-Generated Content’s Massive Impact on Global Worldviews
By
–
100% — AI is going to be creating so much of the world’s online content soon. At scale, I think how it acts/behaves will quite literally shape a lot of the world’s views.
-
Karpathy releases tokenization tutorial covering UTF-8 and byte pair encoding
By
–
ICYMI – Andrej Karpathy has released an excellent video tutorial on "Tokenization" couple of days back. ⦿ Basics covered: Strings, Unicode code points, and encodings like UTF-8.
⦿ Byte pair encoding algorithm explained and implemented in Python.
⦿ Delving into complexities: -
Magic AI Startup Achieves Major Breakthrough in Active Reasoning
By
–
Magic, a startup building AI coding “coworkers”, claims to have made a massive AI breakthrough. The startup claims its model now can ingest up to 3.5M words (nearly 5x Gemini 1.5) and made an ‘active reasoning’ breakthrough reminiscent of OpenAI's Q*.