Jason

Check out my latest portfolio I built! jasonsrobinson.com
#DataScience #DeepLearning #AI

The evolution of AI: From AlphaGo to AI agents, physical AI, and beyond The trajectory of AI's evolution underscores its potential to transform industries and society, and elevate our daily lives.

www.technologyreview.com/2025/02/28/1...

AI-driven blood tests, leveraging nanotube technology (50,000x thinner than a hair) and machine learning, are showing promise in detecting ovarian cancer at its earliest stages, offering a potential game-changer in cancer diagnostics.💡

Source: tinyurl.com/3ts3rhry

#AI #AIinHealth #TechForGood

the rl for deepseek-r1 - Entropic Musings - Obsidian Publish An introduction to the RL behind DeepSeek-R1 DeepSeek-R1 took us all by surprise. This would have been considered AGI 10 years ago and not only did DeepSeek release the weights, but they also blessed…

An introduction to the RL behind DeepSeek-R1 by hallerite

Since GRPO is the main RL algorithm that we want to understand in the end and GRPO is based on PPO, we should start with PPO.

publish.obsidian.md/hallerite/rl...

Blood tests are currently one-size-fits-all − machine learning can pinpoint what’s truly ‘normal’ for each patient A narrower, more personalized ‘normal range’ could help doctors better diagnose and treat disease in individual patients.

Your "normal" blood test results may not actually be normal for you. A mathematician explains how #MachineLearning can find healthy blood count ranges for individual patients and improve predictions of their risk of future disease. https://buff.ly/3P1vAfM 🩺🧪#AI #PersonalizedMedicine #CBC

They created custom instructions within the LLM specifically according to username, for a similar scenario, from that user. Positive those actions aren’t isolated to one user, but likely a distribution list of Superusers of similar celebrity.

Happy Birthday🎂

Scaling Synthetic Data Creation with 1,000,000,000 Personas We propose a novel persona-driven data synthesis methodology that leverages various perspectives within a large language model (LLM) to create diverse synthetic data. To fully exploit this methodology...

AI2 used this method to help create sufficiently diverse prompts for supervised fine-tuning, which is a strong recommendation.

But I kind of can't believe it works. What gives us confidence that an LLM can infer an author description from text, or use one to inflect text? They never see authors!

Bluesky has an impersonator problem Cryptoscammers tried to trick me using fake profiles of tech personalities. I am not alone.

The unfortunate effect of mass growth is that evil-doer opportunists will seize on these instances. Clearly @Bluesky will need to have some form of user verification coming down the pipeline. www.technologyreview.com/2024/12/11/1...

Excited about BioEmu?

Opening this position for just a few days over #NeurIPS2024. Looking especially for Bioinformatics + Structural Biology skills, MD/Stat Mech skills and/or #deeplearning architecture design + engineering skills

aka.ms/ai4science-r...

New publication (4+ years in the making): “The Diffusion and Reach of (Mis)Information on Facebook”. shorturl.at/VE2fU
We analyze the propagation of 1B+ posts across content moderation regimes, with @davidlazer.bsky.social @jatucker.bsky.social @taliastroud.bsky.social @annenbergpenn.bsky.social

Neurons in adults are more "rigid" and require more stimulus to change. This is represented by a cartoon of neurons within an adult silhouette having straight and firm connections, opposed to the neurons within a child that have more flexible and adaptable connections.

Some of my prior #research shows how certain adult neurons tied to reward resist change!

It is free access!

doi.org/10.3389/fnin...

On a macro level, this could provide a new hypothesis for why adults exhibit resistance to #learning, #anxiety, and #addiction.

#neuroscience #science 🧠🧪

a drawing of a woman with the words if now isn 't a good time for the truth i don 't see ALT: a drawing of a woman with the words if now isn 't a good time for the truth i don 't see

She will be missed! #NikkiGiovanni

Recent advances and applications of artificial intelligence in 3D bioprinting 3D bioprinting techniques enable the precise deposition of living cells, biomaterials, and biomolecules, emerging as a promising approach for engineering functi

[new review] Zhang Lab. Recent advances and applications of artificial intelligence in 3D bioprinting. Biophys Rev (Melville)
http://dlvr.it/TGbPnR

Demystifying LLMOps: A Practical Database of Real-World Generative AI Implementations - ZenML Blog The LLMOps Database offers a curated collection of 300+ real-world generative AI implementations, providing technical teams with practical insights into successful LLM deployments. This searchable res...

The ZenML team has compiled over 300 real-world examples of deploying LLMs in production. It might help technical teams tackle practical challenges with a focus on concrete implementation details while avoiding promotional fluff www.zenml.io/blog/demysti... #dataBS

Have you ever wondered how to train an autoregressive generative transformer on text and raw pixels, without a pretrained visual tokenizer (e.g. VQ-VAE)?

We have been pondering this during summer and developed a new model: JetFormer 🌊🤖

arxiv.org/abs/2411.19722

A thread 👇

1/

GitHub - NirDiamant/GenAI_Agents: This repository provides tutorials and implementations for various Generative AI Agent techniques, from basic to advanced. It serves as a comprehensive guide for buil... This repository provides tutorials and implementations for various Generative AI Agent techniques, from basic to advanced. It serves as a comprehensive guide for building intelligent, interactive A...

One of the best repos I’ve come across for #GenAI Agents.
github.com/NirDiamant/G...

Building AI trust: The key role of explainability AI systems are powerful but often operate like “black boxes,” shrouded in mystery. Here’s how companies can shed some light and drive adoption of AI solutions that users trust and understand.

Shareholder’s of companies want #AI, due to FOMO. Similar to a math assignment, they don’t care that it works, they want to know how. They want to see the work to have confidence in the solution. #XAI

www.mckinsey.com/capabilities...

Congratulations!

Competitive Scrabble 👀

Scaling just means making the app bigger. Functionality, security, etc.

Wow, impressive work! 🤩
+ It comes with #opensource code and model - can't wait to try it out! 😀
github.com/microsoft/Bi...

Thanks for sharing! Currently working on a project of different modalities for segmentation and classification tasks. This surely will help!

Hilarious 🤣

I personally agree that LLMs do not “know anything”, but extremely useful “tools”, particularly in research. They will guide you to an area of possible solutions, not absolute solutions.

Great point! The simplified dataset will allow the LLM to find patterns easier than a trained physician, but at the cost of loss context.

Congratulations 🎉

Yea there can be moments of fun but the reality when the camera is off is far from fun it. Nothing related to violence. But they can only go out at certain times, never at night, told when to go to sleep when to wake up. Work 30hrs a week for $10-$15 a month. Nah, them tiktoks a facade. Stay free💯

Great questions. Yes, technically the bias introduced by medical technicians would affect the model’s capability to better generalize the data. However, there are methods that are used to account for the bias in models, in conjunction with the technicians and human raters to mitigate bias.

Bro, they’re not, trust me!