Happy to see LLM security taken seriously by OWASP https://t.co/WQqb0coc0L Here's my (offhand) thoughts on the list: 1/

Still use ⛓️Chain-of-Thought (CoT) for all your prompting? May be underutilizing LLM capabilities🤠 Introducing 🌲Tree-of-Thought (ToT), a framework to unleash complex & general problem solving with LLMs, through a deliberate ‘System 2’ tree search. https://t.co/V6hjbUNjbt

JUST. LOOK. AT. ALL. THESE. *FREE*. ARTICLES!! Highly readable, written by experts, looking at the last 50 years & ahead to what comes next for research on #consciousness, #memory, #hearing, #neuroprosthetics, #antipsychotics, emotions, #stemcells https://t.co/LWL6KHVo6w

@ylecun @hardmaru I believe there are strong reasons to not only use gradient descent even when it is applicable. Namely that gradient descent performs inductive learning, stochastic methods can do something like hypothetico-deductivism, which is fundamentally more capable. https://t.co/krQACUK5IO

This is how little code it takes to implement a siamese net using @fastdotai and @pytorch. I share this because I continue to be amazed.

@hardmaru While one should strive to use gradient-based optim whenever possible, there are situations where the only option is gradient-free. Gradient-based is way more efficient, but not always applicable. PS: my 1st paper on perturbative learning was at NIPS 1988

This library from @ylecun’s lab implemented benchmarked versions of evolutionary computation algorithms such as: Differential Evolution, Fast Genetic Algorithm, CMA-ES, Particle Swarm Optimization. 🦎🐞🐜

For those who missed the "Neural Ordinary Differential Equations" ( https://t.co/upUhWNtpvC) presentation by David Duvenaud on NIPS, here is the video session: https://t.co/swdiFPr2Wu, it starts at 24:00.

Cards Against Machine Learning

✨🧠Friendly periodic reminder that @TensorFlow supports traditional models like random forest, linear and logistic regression, k-means clustering, and gradient-boosted trees! The syntax is a bit more verbose than scikit-learn; but you get the added bonus of GPU acceleration. 👍

10 Exciting Ideas of 2018 in NLP: A collection of 10 ideas that I found exciting and impactful this year—and that we'll likely see more of in the future. https://t.co/iv29bxYbq4

The code of the final project I worked on within the @OpenAI scholarship program is on Github. In the GridWorld environment an agent navigates according to commands to hit/avoid target cells. https://t.co/jj1m29bfio #MachineLearning #ArtificialIntelligence #AI #technology

A lot of what's considered "elective" in CS education should actually be required. HCI is a lot more relevant to most software engineers these days than something like compiler design.

When a statisticians hear, "Successful people start their day at 4 a.m.", they think: 1. Waking early makes you successful? 2. Something about success makes it hard to sleep at night? 3. Success is lethal; Only early risers survive? 4. You did your survey at 4am. #epitwitter

Still the best thing that's ever been made for me. :) ♥, @letstechno! http://t.co/BCV2JfdVP5

#Data #generation for #DataScience and #MachineLearning a must-have #skill for #DataScientist https://t.co/NaqO7RnyAI

#PyMC3 3.6 RC1 is released: https://t.co/9UPBM0o86v Please help us test it and report issues on our tracker, this is not production-ready yet. 3.6 will be the last release to support Python 2.

"Continual Match Based Training in Pommerman: Technical Report," Peng and Pang et al.: https://t.co/rxBcN4rx1Z Winner of Pommerman competition

"Malthusian Reinforcement Learning," Leibo et al.:

Really cool design of a spiking NN that appears to actually outperform LSTM and GRU! Deep Networks Incorporating Spiking Neural Dynamics https://t.co/VeJBnkgPcB