OpenAI, Google and Anthropic Are Struggling to Build More Advanced AI “Three of the leading artificial intelligence companies are seeing diminishing returns from their costly efforts to develop newer models.” Scaling laws are breaking down economically.
@hardmaru
-
The Real Winner of AI and Tech Market Competition
By
–
The real winner https://
japantimes.co.jp/business/2024/
11/06/companies/nvidia-ai-apple-largest-company/
… -
The Coming AI Startup Bust and Bubble Economy
By
–
The Coming AI Startup Bust “I think we are in an AI startup bubble and that is okay…” @asianometry YouTube: https://
youtu.be/6aS0Dlqarqo -
AI Companies Work: Analysis and Discussion
By
–
Related article: https://
benn.substack.com/p/do-ai-compan
ies-work
… -
AI’s $600B Question: Four Months Later, Still Searching for Answers
By
–
Four months after @DavidCahn6
’s “AI’s $600B Question”, are we close to having any answers, or even the concept of an answer? -
AMD’s missed opportunity in deep learning stack development
By
–
And imagine where would AMD be today if they got the deep learning stack done right much earlier, back in 2017.
-
High Quality 5T Token Dataset Released for LLM Training
By
–
They also released a 5T token high quality, processed dataset for training LLMs https://
huggingface.co/datasets/Zyphr
a/Zyda-2
… -
Dennett on Nature’s Need-to-Know Principle in Creature Design
By
–
Makes me think of Daniel Dennett’s book:
— hardmaru (@hardmaru) 15 octobre 2024
“Nature makes heavy use of the Need to Know principle, and designs highly successful, adept, even cunning creatures who have no idea what they are doing or why.”
(From Bacteria to Bach and Back) https://t.co/1C7zUNilJ2Makes me think of Daniel Dennett’s book: “Nature makes heavy use of the Need to Know principle, and designs highly successful, adept, even cunning creatures who have no idea what they are doing or why.” (From Bacteria to Bach and Back)
-
New SOTA 7B LLM Released Under Apache-2.0 License
By
–
Cool to see a new SOTA 7B LLM released under Apache-2.0. Hopefully this will help bring state space models to mainstream! https://
huggingface.co/Zyphra/Zamba2-
7B
… -
Meta-learning Demands More Trial-and-Error Than Standard Learning
By
–
Meta-learning requires a lot more trial-and-error than learning.