INNOVATION - AI Dynamics

DeepMatcher: Transformer Network Achieves State-of-the-Art Feature Matching

By

–

15 January 2023 18h31

DeepMatcher is a transformer-based network showing robust local feature matching, outperforming the state-of-the-art methods on several benchmarks. 10 of 11

→ View original post on X — @dair_ai

15 January 2023

Transformers in Reinforcement Learning: A Comprehensive Survey

By

@dair_ai

–

15 January 2023 18h31

A Survey on Transformers in Reinforcement Learning 8 of 11

→ View original post on X — @dair_ai

15 January 2023

Scaling Laws for Generative Mixed-Modal Language Models

By

@dair_ai

–

15 January 2023 18h31

Introduces scaling laws for generative mixed-modal language models. 9 of 11

→ View original post on X — @dair_ai

15 January 2023

Large Language Models Simulate Universal Turing Machine with Memory

By

@dair_ai

–

15 January 2023 18h31

Investigates the use of existing LMs (e.g, Flan-U-PaLM 540B) combined with associative read-write memory to simulate the execution of a universal Turing machine. 7 of 11

→ View original post on X — @dair_ai

15 January 2023

DeepMind Tracr: Compiler Converting RASP to Transformer Weights

By

@dair_ai

–

15 January 2023 18h31

DeepMind proposes Tracr, a compiler for converting RASP programs into transformer weights. This way of constructing NNs weights enables the development and evaluation of new interpretability tools. 3 of 11

→ View original post on X — @dair_ai

15 January 2023

DreamerV3 Achieves Minecraft Diamond Collection Without Human Data

By

@dair_ai

–

15 January 2023 18h31

DreamerV3 is a general algorithm to collect diamonds in Minecraft from scratch without human data or curricula, a long-standing challenge in AI.

2 of 11https://t.co/pvsmnX3TPO
— DAIR.AI (@dair_ai) 15 janvier 2023

DreamerV3 is a general algorithm to collect diamonds in Minecraft from scratch without human data or curricula, a long-standing challenge in AI. 2 of 11

→ View original post on X — @dair_ai

15 January 2023

Computer CoPilot Vision Now Tractable Through Text Translation

By

@karpathy

–

15 January 2023 18h00

Computer CoPilot. Was very much the vision with OpenAI Universe https://
openai.com/blog/universe/ , though it was too early. Now feels tractable if you translate everything to/from text (e.g. like in WebGPT). Could be built e.g. as an extension of natbot https://
github.com/nat/natbot

→ View original post on X — @karpathy

15 January 2023

ChatGPT Productivity Tips for Software Developers

By

@bamitav

–

15 January 2023 17h39

How to use ChatGPT to boost your #developers productivity https://
dev.to/developeratul/
how-to-use-chatgpt-to-boost-your-developer-productivity-3e5a
… #ChatGPT #OpenSource #RLHF #100DaysOfCode #100Devs #javascript #Python #tech #architect #AI #ML #DL #AIEthics #OpenAI #chatgpt3 #code #Coding #GPT3 #gpt4 #gptchat #gpt3chat #chatbot #ChatbotAI

→ View original post on X — @bamitav

15 January 2023

LeCun’s Evolving Stance on AI Research Paper Concerns

By

@garymarcus

–

15 January 2023 17h12

What are you claiming exactly? I stand by all the concerns there and would go further and note that Yann LeCun ridiculed the paper at the time but now basically makes all the same points in his own talks.

→ View original post on X — @garymarcus

15 January 2023

McKinsey AI Strategy: Building Blocks and Data Intelligence

By

@deeplearn007

–

15 January 2023 17h05

McKinsey: Using #AI Strategically – Focus on Building Blocks, Making Sense of Data https://
mckinsey.com/capabilities/s
trategy-and-corporate-finance/our-insights/artificial-intelligence-in-strategy
… @pierrepinna @sallyeaves @PawlowskiMario @Xbond49 @gvalan @psb_dc @HaroldSinnott @Shi4Tech @mikeflache @Nicochan33 #MachineLearning #DeepLearning #Fintech #Datascience

→ View original post on X — @deeplearn007

15 January 2023