Greg Warzecha (@greg_warzecha) 's Twitter Profile
Greg Warzecha

@greg_warzecha

Entropy is the ultimate boss battle. Credit assignment is your secret power. Solving AGI.

ID: 42932929

linkhttps://genotic.com calendar_today27-05-2009 17:39:12

389 Tweet

245 Followers

2,2K Following

Sauers (@sauers_) 's Twitter Profile Photo

After a network memorizes the training data, cross-entropy still pushes the logits to grow bigger and bigger, so argmax of the logits is the same but cross-entropy decreases. If this goes on long enough, your softmax hits the limit of floating-point precision. [1/2]

After a network memorizes the training data, cross-entropy still pushes the logits to grow bigger and bigger, so argmax of the logits is the same but cross-entropy decreases. If this goes on long enough, your softmax hits the limit of floating-point precision. [1/2]
Derya Unutmaz, MD (@deryatr_) 's Twitter Profile Photo

I’ve been saying how awesome o1-Pro was from the moment I started using it, and the more I used it, the more I realized it was even more amazing than I initially thought! I’m now trying to contemplate just how much more incredibly awesome o3 will be!

Andrej Karpathy (@karpathy) 's Twitter Profile Photo

This weekend falling deeper into the rabbit hole of contaminants exposure in daily life... I am a bit surprised how weak the U.S. regulations are compared to other countries around industrial chemical use. E.g. the lab PlasticList recommended for testing had this infographic

This weekend falling deeper into the rabbit hole of contaminants exposure in daily life...

I am a bit surprised how weak the U.S. regulations are compared to other countries around industrial chemical use. E.g. the lab <a href="/plasticlistorg/">PlasticList</a> recommended for testing had this infographic
Elon Musk (@elonmusk) 's Twitter Profile Photo

So many people in Europe lack hope for the future or think Europe is “bad” in some way. Pervasive pessimism. This will lead to the end of Europe. Therefore, it must change.

Samuel Hume (@drsamuelbhume) 's Twitter Profile Photo

Top 5 advances in medicine this week (🧵) 1. A tool to image inflammation Inflammation happens in a range of diseases - autoimmunity, aging, cancer - but we lack a tool to visualise it This study used CD45, which is expressed by all - and only - immune cells, as a marker of

Andrej Karpathy (@karpathy) 's Twitter Profile Photo

"Move 37" is the word-of-day - it's when an AI, trained via the trial-and-error process of reinforcement learning, discovers actions that are new, surprising, and secretly brilliant even to expert humans. It is a magical, just slightly unnerving, emergent phenomenon only

Derya Unutmaz, MD (@deryatr_) 's Twitter Profile Photo

Scientific journal paywalls must be abolished! I call upon all major journal publishers to tear down these walls! Let knowledge be free-remove one of the greatest obstacles to human progress! I also urge all scientists not to publish in any journal that is not open access!

samsja (@samsja19) 's Twitter Profile Photo

Synthetic-1 is only the tip of the iceberg of what is coming. We are making progress on the three essential pillars of open-source decentralized AGI: * Pretraining (intellect-1), * synthetic data gen (synthetic-1), * RL (soon) Beyond the research, we are also building

Ben Norton (@benjaminnorton) 's Twitter Profile Photo

There's a common US talking point falsely claiming China's manufactured goods are so affordable because workers supposedly aren't paid well. That's not true. Wages in China are much higher than in neighboring countries in SE Asia. China produces goods so cheaply because: 1. A

There's a common US talking point falsely claiming China's manufactured goods are so affordable because workers supposedly aren't paid well.

That's not true. Wages in China are much higher than in neighboring countries in SE Asia.

China produces goods so cheaply because:

1. A
Andrej Karpathy (@karpathy) 's Twitter Profile Photo

Agency > Intelligence I had this intuitively wrong for decades, I think due to a pervasive cultural veneration of intelligence, various entertainment/media, obsession with IQ etc. Agency is significantly more powerful and significantly more scarce. Are you hiring for agency? Are

TuringPost (@theturingpost) 's Twitter Profile Photo

How does Multi-Head Latent Attention (MLA) reduce memory use? MLA is like zipping and unzipping stored data to save memory. It compresses the key-value (KV) cache into a much smaller form using low-rank key-value joint compression. Here's how MLA works: ▪️ KV pairs are

How does Multi-Head Latent Attention (MLA) reduce memory use?

MLA is like zipping and unzipping stored data to save memory.

It compresses the key-value (KV) cache into a much smaller form using low-rank key-value joint compression.

Here's how MLA works:

▪️ KV pairs are
no.mind (@the_no_mind) 's Twitter Profile Photo

This is Thomas Seyfried. He’s a professor of biology, who’s studied cancer for 30+ years. His message? Cancer isn’t bad genes or bad luck—it’s damaged mitochondria. This flips everything you’ve been told about how to treat & prevent cancer: 🧵

This is Thomas Seyfried.

He’s a professor of biology, who’s studied cancer for 30+ years.

His message? Cancer isn’t bad genes or bad luck—it’s damaged mitochondria.

This flips everything you’ve been told about how to treat &amp; prevent cancer: 🧵
Baptiste (@baptistevicini) 's Twitter Profile Photo

Google DeepMind's CEO just stunned 60 Minutes viewers. The Nobel prize winner revealed: • An AI that can see and understand in real-time • A plan to end ALL diseases in 10 years • Exactly when AI will surpass human intelligence Here are his 4 most jaw-dropping insights:🧵

Google DeepMind's CEO just stunned 60 Minutes viewers.

The Nobel prize winner revealed:

• An AI that can see and understand in real-time
• A plan to end ALL diseases in 10 years
• Exactly when AI will surpass human intelligence

Here are his 4 most jaw-dropping insights:🧵
Jousef Murad (@jousefm2) 's Twitter Profile Photo

Physics-Informed Neural Networks (PINNs) - Conor Daly | Podcast #120 Physics-Informed Neural Networks (PINNs) integrate known physical laws into neural network learning, particularly for solving differential equations. 🎬 Watch now: youtube.com/watch?v=eKzHKG…

Physics-Informed Neural Networks (PINNs) - Conor Daly | Podcast #120

Physics-Informed Neural Networks (PINNs) integrate known physical laws into neural network learning, particularly for solving differential equations.

🎬 Watch now: youtube.com/watch?v=eKzHKG…
Patrick Hsu (@pdhsu) 's Twitter Profile Photo

Genomes encode biological complexity, which is determined by combinations of DNA mutations across millions of bases In new Arc Institute work, we report the discovery and engineering of the first programmable DNA recombinases capable of megabase-scale human genome rearrangement