Maximus November retweeted
Today we're releasing GPT4All, an assistant-style chatbot distilled from 430k GPT-3.5-Turbo outputs that you can run on your laptop.
Maximus November retweeted
Unbelievable satellite imagery of one of the most intense March cyclones on record for California with the eye of the storm making landfall near San Francisco.
Amazing.
Maximus November retweeted
A special weather statement has been issued for Oakland CA, Fremont CA and Hayward CA until 6:00 PM PDT – at San Ramon, CA

ALT This graphic displays a special weather statement plotted on a map. The statement is in effect until 6:00 PM PDT. The statement includes Oakland CA, Fremont CA and Hayward CA. Strong winds associated with the edge of a strong low pressure system will impact portions of Alameda and Contra Costa and northwestern Santa Clara Counties through 600 PM PDT. The threats associated with this storm are wind gusts up to 55 MPH and no hail is expected. Seek shelter in a sturdy structure until the storm passes.
Maximus November retweeted
There was a thread on here last week where two Turing award winners are jovially promoting the idea that mathematical statistics is better at determining causation than common sense. 1/10
Maximus November retweeted
Update: #Hwy17 is reopened from Granite Creek to the Santa Clara/Santa Cruz County line following the removal of downed trees/snowfall.
Maximus November retweeted
How can LLMs such as GPT-3 and ChatGPT achieve greater factual accuracy without relying on an external retrieval search engine?
Our #ICLR2023 paper shows that recitation can help - like humans!
Recitation-Augmented Language Models
arxiv.org/abs/2210.01296
1/N
Maximus November retweeted
Fascinating audit of social media "raciness" classifiers that don't understand context and are massively biased toward labeling images of women's bodies as sexual. Posts classified racy get shadowbanned.
By @gianlucahmd and @HilkeSchellmann. theguardian.com/technology/2…
Maximus November retweeted
Amazing new lib from @daviddobrinskiy which shows a friction-free way to create web apps with @streamlit, #nbdev, and @ProjectJupyter.
This is what all app development should feel like! Iterative, exploratory, with real time immediate feedback.
ddobrinskiy.quarto.pub/blog/…
Maximus November retweeted
Wow! February 1st snowpack numbers haven't been this good since 1995❄️
State average to date =205% of normal
3rd year of La Niña
1995= 207% of normal
Switched from 94-95 El Niño to 95-96 La Niña
@nbcbayarea #snow #California #drought
Maximus November retweeted
Once in a while, I stumble upon some nice little feature in the @PyTorch library that I haven't seen before.
Have you tried AutoAugment, yet?
Just gave it a try: takes a few epochs, but that's a 10% accuracy improvement in this simple example case
github.com/rasbt/deeplearnin…
Maximus November retweeted
Large language models (LLMs) have been criticized due to their heavy reliance on humans to create datasets for fine-tuning and RLHF, but recent research suggests that we might not even need humans for this… 🧵[1/9]
Maximus November retweeted
Our benchmark of tree-based models vs deep learning for tabular data: final version.
TL:DR: from small compute budget, @scikit_learn's HistGradientBoosting is best. With finer tuning of hyperparams, XGBoost brings a gain (here n ranges from 3,000 to 10,000)
Maximus November retweeted
Idealists and indirect realists often stress how selective and distorting our evolved perceptual systems are -- without, it seems, realizing that if you apply the same considerations to introspection you end up with illusionism
Maximus November retweeted
Training factually correct LLMs requires vast amounts of factually correct training data.
Curating & creating factually correct training data for LLMs is very inefficient (vs using this data directly in the first place).
Wrote a shot post about it here:
sebastianraschka.com/blog/20…
Maximus November retweeted
NEW: In @ScienceMagazine today, our latest peer-reviewed research shows Exxon scientists predicted global warming with shocking skill & accuracy between 1977 & 2003, contradicting the company's decades of climate denial. THREAD.
📰No pay wall for 2 weeks: bit.ly/ExxonKnew2point0
Maximus November retweeted
Exxon has known about global warming since the 1970s. They've lied and obfuscated ever since to maintain highly profitable operations. The current climate change plan is to give these same companies trillions in public subsidies to lead the green transition. That’s just insane.
NEW: In @ScienceMagazine today, our latest peer-reviewed research shows Exxon scientists predicted global warming with shocking skill & accuracy between 1977 & 2003, contradicting the company's decades of climate denial. THREAD.
📰No pay wall for 2 weeks: bit.ly/ExxonKnew2point0
Maximus November retweeted
These things sound AMAZING. And look at that RAM footprint!!! I'll work on a mockup using this stuff soon, I love them. Thanks to those who recommended #gamemusic #gameaudio #gamecomposer #filmcomposer
Maximus November retweeted
One hopes the drive for 16-18 maths doesn't follow Einstein's quote: "insanity is doing the same thing over and over again and expecting different results."
Maximus November retweeted
96-year-old Freeman Dyson in his last email to me #OTD three years ago: "Robots and software look to me like a bunch of useful tools smothered in public relations hype."





























