RESEARCH

→ View original post on X — @jxmnop

Embedding Models: Lower Barrier Entry for AI Startups

By

@jxmnop

–

31 October 2023 13h34

the barrier to entry for training good embedding models is 100x lower than for training good LLMs – requires significantly less compute, expertise, and data. seems like a good area for building a business

Self-Taught AI: Petar’s Path to Research Success

By

–

31 October 2023 13h00

As evidenced by Petar, who learned modern AI online, anyone motivated can join AI research. No matter your background, academic flair, or professional interest, the story of Petar Veličković surely inspires and offers valuable insights.

Google Maps’ Travel Prediction Algorithm: Curiosity and Perseverance

By

–

31 October 2023 13h00

Petar, one of the minds behind Google Maps' travel time prediction algorithm, emphasizes the value of curiosity, audacity, and perseverance in AI research.

Graph Neural Networks: Interview with DeepMind AI Researcher

By

–

31 October 2023 13h00

Interested in Graph Neural Networks and AI? I've recently chatted with @PetarV_93 , a scientist at @GoogleDeepMind and lecturer at @Cambridge_Uni . His journey spans maths, competitive programming to AI research.

→ View original post on X — @huggingface

Join Hugging Face Blog Community and Share ML Insights

By

@huggingface

–

31 October 2023 11h44

Check out the awesome blog posts written by the machine learning community, at http://
hf.co/blog If you want to write one, simply join the blog-explorers organization and follow the instructions in the organization card https://
huggingface.co/blog-explorers

Responsible AI Development: Power Demands Great Responsibility

By

–

31 October 2023 11h19

With Great Power Comes Great Responsibility. Be careful when you build and release AI models!

Pretty cool to cite Spider Man in a real world issue 😅

Listen to the full interview on the What’s AI podcast!#ai #responsibleai #aiethics #whatsai pic.twitter.com/Jj37InQisP
— Louis-François Bouchard 🎥🤖 (@Whats_AI) 31 octobre 2023

With Great Power Comes Great Responsibility. Be careful when you build and release AI models! Pretty cool to cite Spider Man in a real world issue Listen to the full interview on the What’s AI podcast! #ai #responsibleai #aiethics #whatsai

→ View original post on X — @actuiafr

OpenAI’s Strategy to Prevent Catastrophic AI Risks

By

@actuiafr

–

31 October 2023 10h35

OpenAI's strategy to prevent "catastrophic" risks from pioneering AI models https://actuia.com/actualite/la-strategie-dopenai-pour-prevenir-les-risques-catastrophiques-des-modeles-dia-pionniers/
… #AI #artificialintelligence

→ View original post on X — @sauain

Bubble Sort: Essential Sorting Algorithm for Developers

By

@sauain

–

31 October 2023 8h19

Sorting Algorithms that every developer should know about: Bubble Sort: A simple comparison-based algorithm where adjacent elements are swapped if they are in the wrong order. It's named for the way smaller elements "bubble" to the top of the list. Best for small datasets!

→ View original post on X — @soumithchintala

LLaMa2-70B Training Computation Corrected Calculations

By

@soumithchintala

–

31 October 2023 5h29

Corrected Calculations (the original claim doesn't change, as it goes from 5e+24 -> 1.1e+24 and is still two orders of magnitude behind):
LLaMa2-70B was trained on 2T tokens, and 1.7m hours of A100 GPU time. At a HFU of ~60%, LLama2-70B took ~1.1e+24 flops (1.7m * 312TFlops