What a pleasure to see the open-source and academic community at full speed on pushing smart ways to do long context with pretrained LLM Check this thread and amazing work. Pushing LLaMA up to 8k context and more with negligible degradation in quality
RESEARCH
-
AI Data Programming Machine Learning Innovation India
By
–
@jeevprabnivash #ai #data #programming #MachineLearning #INDIAai
-
Qualification Problem in AI: Testing Machine Intelligence
By
–
'Qualification problem' in AI refers to the challenge of testing a machine's intelligence. With varying ideas of what defines intelligence, specifically in cases of autonomous systems, can result in this problem. Explore its history & implications here: https://
rb.gy/w9g2o -
Reserved Instances: Training Infrastructure and Utilization Challenges
By
–
Reserved instances are trainers secret sauce! Not only because of the « guaranteed » part but also for other reasons for example it incentives researchers to train/experiment rather than brainstorm haha. Comes with other less publicized challenges though (like under-utilization).
-
Technochauvinism and AI Bias: Beyond Neutral Algorithms
By
–
In this article, we discuss the concept of #technochauvinism, shedding light on the often-overlooked societal #implications and #biases in #AI, and emphasize the importance of discerning between technological and human #solutions for various #challenges. https://
forbes.com/sites/bernardm
arr/2023/06/30/when-bytes-bias-unraveling-the-myth-of-neutral-ai/
… -
Deep Blue and Chess: Not the Way I Told It
By
–
Not any way I told it. You can point Deep Blue at chess.
-
Google’s proprietary AI significantly outperforms open source alternatives
By
–
Yea. I'm still 200% convinced that everything in the open is still garbage compared to what's inside Google. It was quite depressing honestly to see this level difference.
-
Matthew Berman Interviews H2O.ai Principal Data Scientist on LLM Fine-tuning
By
–
Interview Alert! @MatthewBerman sits down with @kagglingpascal, Principal Data Scientist, at @h2oai
. Together they delve into the fascinating world of #LLM fine-tuning, #Kaggle, and cutting-edge innovations at http://
H2O.ai and much more. https://
youtube.com/watch?v=Ur-1PI
9SMfw
… -
The Unsolved Problem of Defining Low Impact for Superintelligence
By
–
I don't think either of these have been falsified. We still don't know how to say "low impact" in a way that holds up to a superintelligence, and if you think ChatGPT demonstrates otherwise then you didn't understand the original problem.
-
Three realizations about AI alignment difficulty and lack of progress
By
–
Realized that alignment was necessary, then realized that alignment was hard, then realized we were not on track to make it or even come close.