abrichr

11 ms·

121.
by abrichr 3y ago
Perhaps this qualifies: Mark Zuckerberg: First Interview in the Metaverse | Lex Fridman Podcast #398 https://www.youtube.com/watch?v=MVYrJJNdrEg
122.
Learning Unsupervised World Models for Autonomous Driving via Discrete Diffusion(arxiv.org)
2 points by abrichr 3y ago|1 comments
123.
by abrichr 3y ago
> Learning world models can teach an agent how the world works in an unsupervised manner. Even though it can be viewed as a special case of sequence modeling, progress for scaling world models on robotic applications such as autonomous d
124.
by abrichr 3y ago
Try this custom instruction: - Skip any preamble or qualifications about how a topic is subjective. - Be terse. Do not offer unprompted advice or clarifications. Speak in specific, topic relevant terminology. Do NOT hedge or quali
125.
by abrichr 3y ago
Thank you for letting me know!
126.
by abrichr 3y ago
> No one ever shows what these should and could actually be used for - incredibly boring and repetitive tasks. Agreed! That’s precisely why we created https://openadapt.ai .
127.
by abrichr 3y ago
> Rabbit means to solve that by creating a "LAM", a "Large Action model", which is a service by Rabbit that will click interfaces for you. https://openadapt.ai is an open source app that runs on your local
128.
by abrichr 3y ago
The “neuron” in a neural network is just a non linear function of the weighted sum of the inputs (plus a bias term). See the “definition” section in https://en.wikipedia.org/wiki/Perceptron .
129.
by abrichr 3y ago
The success of Github in creating software, and the success of software in advancing scientific progress. Maybe something like nbdev.fast.ai. In any case, it was just a thought, and likely not an original one. I would welcome it if someone
130.
by abrichr 3y ago
Tests and data.
131.
by abrichr 3y ago
> Peer review is basically Github anonymous PRs that has the author pinky swear that the code compiles and 95% of test cases pass. It should be possible to use something like Github to *verify* "that the code compiles and 95% of tes
132.
by abrichr 3y ago
I understand, thank you for clarifying :) My point was that academics could use Github (or something like it)
133.
by abrichr 3y ago
> Github
134.
by abrichr 3y ago
Very promising, excited to learn more! Any thoughts on State Space Models? Eg: https://github.com/havenhq/mamba-chat https://arxiv.org/abs/2311.18257
135.
by abrichr 3y ago
Thank you for your perspective! (We are looking into both for https://github.com/OpenAdaptAI/OpenAdapt )
136.
by abrichr 3y ago
Thank you for the great library and examples! Can you please comment on how simpleaichat compares to https://github.com/outlines-dev/outlines ?
137.
by abrichr 3y ago
This is exactly why in https://github.com/OpenAdaptAI/OpenAdapt we have implemented three separate PII scrubbing providers. Congrats to the op on shipping!
138.
by abrichr 3y ago
> Whoever is going to be able to let "normal" Mac users to install and run a local copy of an LLM, is going to reap tons of commercial benefits. (e.g. DMG, click-install, run. No command line). https://gpt4all.io h
139.
by abrichr 3y ago
This is a great point, and something that may be at least partially addressable with current methods (e.g. RLHF/SFT). Maybe (part of) what's missing is a tighter feedback loop between a) limitations experienced by the human users
140.
by abrichr 3y ago
Open source question answering over videos: > With the binding of unified visual representations to the language feature space, we enable an LLM to perform visual reasoning capabilities on both images and videos simultaneously.
141.
by abrichr 3y ago
From https://arxiv.org/abs/2106.09685 : > LoRA: Low-Rank Adaptation of Large Language Models > An important paradigm of natural language processing consists of large-scale pre-training on general domain data and a
142.
by abrichr 3y ago
> Note: You'll be asked to approve code before it's run.
143.
by abrichr 3y ago
My guess is Neural Radiance Fields [1], or something quite similar. https://www.matthewtancik.com/nerf
144.
by abrichr 3y ago
e.g. https://thenewamerican.com/us/economy/markets/gadhafi-s-gold...
145.
by abrichr 3y ago
According to [1], the key differentiator is parallax, i.e. the positions of objects change relative to each other, depending on how the viewer is looking at them. [1] https://www.youtube.com/watch?v=pmqr6ydIzDA
146.
by abrichr 3y ago
Thanks for the tip! Automating repetitive GUI workflows is the goal of https://github.com/OpenAdaptAI/OpenAdapt
147.
by abrichr 3y ago
Agreed! This is part of the motivation behind https://github.com/OpenAdaptAI/OpenAdapt
148.
by abrichr 3y ago
This type of use case is exactly why are building https://github.com/OpenAdaptAI/OpenAdapt
149.
by abrichr 3y ago
This is exactly why in https://github.com/OpenAdaptAI/OpenAdapt we have put a lot of emphasis on scrubbing screenshots before sharing them with the model, and built tools to visualize the results.
150.
by abrichr 3y ago
Sorry to hear that. What was the loophole?

More ›