Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

Excited to release METAGENE-1, a 7B parameter metagenomic foundation model, built to aid in pathogen detection & pandemic monitoring. Pretrained on 1.5 trillion base pairs of DNA/RNA sequenced from wastewater. A collab w/ USC, Prime Intellect, & the Nucleic Acid Observatory. 🧵

Willie Neiswanger

1,506 subscribers

13,390 görüntüleme • 1 yıl önce •via X (Twitter)

Bilim & Teknoloji Eğitim Sağlık & İyilik

Anya Rossi• Live Now

Private livecam show

12 Yorum

Willie Neiswanger profil fotoğrafı

Willie Neiswanger1 yıl önce

Metagenomic sequencing of wastewater produces vast amounts of data that can capture public health trends at a societal scale. Our goal is to train a model on this data to help in large-scale wastewater monitoring & detection of novel bio threats. 🌐

Willie Neiswanger profil fotoğrafı

Willie Neiswanger1 yıl önce

Our data pipeline is: human microbiome > wastewater > metagenomic sequences > tokens > training data. Wastewater provides a rich source of data from tens of thousands of species across the human-adjacent microbiome. In total we pretrain on over 1.5T base pairs of DNA/RNA.

Willie Neiswanger profil fotoğrafı

Willie Neiswanger1 yıl önce

METAGENE-1 shows state-of-the-art results on pathogen detection, metagenomic embedding, and other genomic tasks. We also release new benchmarks for genomic detection and embedding (eg, Gene-MTEB, based on MTEB for LLMs). See our paper for details:

Willie Neiswanger profil fotoğrafı

Willie Neiswanger1 yıl önce

We leverage the ecosystem of modern LLM tooling—in tokenization, model architecture, training, infra, etc—for performance and extensibility. METAGENE-1 is standardized & easy to use. Hugging Face: Github:

Willie Neiswanger profil fotoğrafı

Willie Neiswanger1 yıl önce

Our paper also contains an in-depth discussion on safety when releasing metagenomic models. s/o to great coauthors @olliezliu @samsja19 @johannes_hage @UpupWang @_jason136, & jefftk Looking for collaborators to build on this with us — please reach out!

Λbstract profil fotoğrafı

Λbstract1 yıl önce

RT @InstaMAT_io: 👀 Are you looking for an alternative to Adobe Substance tools? Do not miss out on InstaMAT from killer node-based material…

Anthony Gitter profil fotoğrafı

Anthony Gitter1 yıl önce

@PrimeIntellect Exciting work! For next steps, are you thinking about how this would compare to classic bioinformatics workflows (e.g. for pathogen detection? It could also be relevant to compare to the @tatta_bio gLM model and DGEB benchmark.

Willie Neiswanger profil fotoğrafı

Willie Neiswanger1 yıl önce

@PrimeIntellect @tatta_bio Thanks for the links! Yes, these are great ideas for next steps.

Alon Albalak profil fotoğrafı

Alon Albalak1 yıl önce

@USC @PrimeIntellect Wow, awesome work @willieneis and team! Congrats on the release 🎉

Willie Neiswanger profil fotoğrafı

Willie Neiswanger1 yıl önce

@USC @PrimeIntellect Thanks Alon!

Ben Lengerich profil fotoğrafı

Ben Lengerich1 yıl önce

@USC @PrimeIntellect Congrats Willie! Looks very cool.

Willie Neiswanger profil fotoğrafı

Willie Neiswanger1 yıl önce

@USC @PrimeIntellect Thanks Ben!

Benzer Videolar

Releasing METAGENE-1: In collaboration with researchers from USC, we're open-sourcing a state-of-the-art 7B parameter Metagenomic Foundation Model. Enabling planetary-scale pathogen detection and reducing the risk of pandemics in the age of exponential biology.

Releasing METAGENE-1: In collaboration with researchers from USC, we're open-sourcing a state-of-the-art 7B parameter Metagenomic Foundation Model. Enabling planetary-scale pathogen detection and reducing the risk of pandemics in the age of exponential biology.

Prime Intellect

620,098 görüntüleme • 1 yıl önce

We release Cosmos Policy 💫: a state-of-the-art robot policy built on a video diffusion model backbone. - policy + world model + value function — in 1 model - no architectural changes to the base video model - SOTA in LIBERO (98.5%), RoboCasa (67.1%), & ALOHA tasks (93.6%) 🧵👇

We release Cosmos Policy 💫: a state-of-the-art robot policy built on a video diffusion model backbone. - policy + world model + value function — in 1 model - no architectural changes to the base video model - SOTA in LIBERO (98.5%), RoboCasa (67.1%), & ALOHA tasks (93.6%) 🧵👇

Moo Jin Kim

149,333 görüntüleme • 5 ay önce

🚀 Excited to announce the first release of a novel open source programming language and platform for language model interaction! Combining prompts, constraints & scripting, LMQL elevates the capabilities of large language models. 🧵1/6 A quick tour.

🚀 Excited to announce the first release of a novel open source programming language and platform for language model interaction! Combining prompts, constraints & scripting, LMQL elevates the capabilities of large language models. 🧵1/6 A quick tour.

LMQL (Language Model Query Language)

198,966 görüntüleme • 3 yıl önce

Introducing INTELLECT-3: Scaling RL to a 100B+ MoE model on our end-to-end stack Achieving state-of-the-art performance for its size across math, code and reasoning Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more

Introducing INTELLECT-3: Scaling RL to a 100B+ MoE model on our end-to-end stack Achieving state-of-the-art performance for its size across math, code and reasoning Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more

Prime Intellect

1,142,730 görüntüleme • 7 ay önce

ODesign: A World Model for Biomolecular Interaction Design | Odin Zhang et al. - "ODesign allows scientists to specify epitopes on arbitrary targets and generate diverse classes of binding partners with fine-grained control. Across entity-, token-, and atom-level benchmarks in the protein modality, ODesign demonstrates superior controllability and performance to modality-specific baselines. Extending beyond proteins, it generalizes to nucleic acid and small-molecule design, enabling interaction types such as protein-binding RNA/DNA and RNA/DNA-binding ligands that were previously inaccessible." Technical Report: Project Page:

ODesign: A World Model for Biomolecular Interaction Design | Odin Zhang et al. - "ODesign allows scientists to specify epitopes on arbitrary targets and generate diverse classes of binding partners with fine-grained control. Across entity-, token-, and atom-level benchmarks in the protein modality, ODesign demonstrates superior controllability and performance to modality-specific baselines. Extending beyond proteins, it generalizes to nucleic acid and small-molecule design, enabling interaction types such as protein-binding RNA/DNA and RNA/DNA-binding ligands that were previously inaccessible." Technical Report: Project Page:

Leo Zang

17,064 görüntüleme • 8 ay önce

Jakob Pöltl on why he wants to stay in Toronto: -Says the team built a foundation the last 1.5 years, and he bought into the process from the beginning. -He’s ‘crazy excited’ about the team, they could be a fun team, and could challenge some of the top teams.

Jakob Pöltl on why he wants to stay in Toronto: -Says the team built a foundation the last 1.5 years, and he bought into the process from the beginning. -He’s ‘crazy excited’ about the team, they could be a fun team, and could challenge some of the top teams.

Omer Osman

12,634 görüntüleme • 1 yıl önce

Miguel speaks on Collab album w/ J.Cole: “I want to do a collab Album w/ J. Cole. That’s one they’ve been asking for & I was just texting him. A collaboration w/ him would be really special because we’re aligned in a lot of the human ways.” Miguel receives a text from Cole 🎥

Miguel speaks on Collab album w/ J.Cole: “I want to do a collab Album w/ J. Cole. That’s one they’ve been asking for & I was just texting him. A collaboration w/ him would be really special because we’re aligned in a lot of the human ways.” Miguel receives a text from Cole 🎥

Caleb

104,845 görüntüleme • 8 ay önce

Doors are open on Elysium Perdita IRL exhibition at Valuart in Switzerland! 🥀🔥 Brilliant opportunity to exhibit a body of work of both collaborative & solo series. Cinema, art & poetry intersect w evocative narratives. The story begins… 1/14🧵

Doors are open on Elysium Perdita IRL exhibition at Valuart in Switzerland! 🥀🔥 Brilliant opportunity to exhibit a body of work of both collaborative & solo series. Cinema, art & poetry intersect w evocative narratives. The story begins… 1/14🧵

Laurence Fuller

34,759 görüntüleme • 3 yıl önce

Welcome to Pirate Nation, an on-chain RPG by Proof of Play, built at the intersection of casual gaming, web3 technology, and shared ownership & incentives. A thread 🧵 covering what you need to know about our game, ecosystem, & team. Let's jump in. 👇 1/many

Welcome to Pirate Nation, an on-chain RPG by Proof of Play, built at the intersection of casual gaming, web3 technology, and shared ownership & incentives. A thread 🧵 covering what you need to know about our game, ecosystem, & team. Let's jump in. 👇 1/many

Pirate Nation

984,663 görüntüleme • 3 yıl önce

DARPA—who developed the Operation Warp Speed Death Jabs (see tweet 2)—now wants to inject people with "protein complexes" that, when hit with "different wavelengths of light" will "synthesize specific DNA or RNA sequences on demand." DARPA = literal Nazis. Nuremberg for DARPA! ----------------Partial transcription of clip--------------- "Synthetic DNA and RNA molecules could enable a host of next-generation technologies crucial for national security and global well-being. However, current methods for creating these molecules from scratch face significant limitations in terms of their scale, complexity, and use of hazardous chemicals. DARPA's Generative Optogenetics Program. "GO seeks to harness the power of light to direct the synthesis of DNA and RNA directly within living cells. If successful, this high-risk, high-reward research program has the potential to revolutionize medicine, agriculture, and manufacturing. "GO aims to unlock an era of programmable biology where we can communicate with and direct cell behavior with unprecedented speed, precision, and reach via optical signals. Performers on GO will design protein complexes that, when expressed in a living cell, use light-triggered signaling to deliver genetic information, instructing cells to incorporate specific building blocks based on different wavelengths of light to synthesize specific DNA or RNA sequences on demand. "The GO program focuses on two main research objectives. The first is achieving the core capability of synthesizing DNA or RNA directly in living cells using optical signals. The second focuses on developing error-mitigation methods that ensure high-fidelity nucleic acid synthesis."

DARPA—who developed the Operation Warp Speed Death Jabs (see tweet 2)—now wants to inject people with "protein complexes" that, when hit with "different wavelengths of light" will "synthesize specific DNA or RNA sequences on demand." DARPA = literal Nazis. Nuremberg for DARPA! ----------------Partial transcription of clip--------------- "Synthetic DNA and RNA molecules could enable a host of next-generation technologies crucial for national security and global well-being. However, current methods for creating these molecules from scratch face significant limitations in terms of their scale, complexity, and use of hazardous chemicals. DARPA's Generative Optogenetics Program. "GO seeks to harness the power of light to direct the synthesis of DNA and RNA directly within living cells. If successful, this high-risk, high-reward research program has the potential to revolutionize medicine, agriculture, and manufacturing. "GO aims to unlock an era of programmable biology where we can communicate with and direct cell behavior with unprecedented speed, precision, and reach via optical signals. Performers on GO will design protein complexes that, when expressed in a living cell, use light-triggered signaling to deliver genetic information, instructing cells to incorporate specific building blocks based on different wavelengths of light to synthesize specific DNA or RNA sequences on demand. "The GO program focuses on two main research objectives. The first is achieving the core capability of synthesizing DNA or RNA directly in living cells using optical signals. The second focuses on developing error-mitigation methods that ensure high-fidelity nucleic acid synthesis."

Sense Receptor

50,875 görüntüleme • 7 ay önce

Super excited to share the last paper of my PhD: "Hallucination in World Models is Predictable and Preventable"✨ We train a 350M-param generative world model on a large dataset w/ 210 tasks and show that we can predict *when* hallucination happens and use that to fix it! 🧵1/n

Super excited to share the last paper of my PhD: "Hallucination in World Models is Predictable and Preventable"✨ We train a 350M-param generative world model on a large dataset w/ 210 tasks and show that we can predict when hallucination happens and use that to fix it! 🧵1/n

Nicklas Hansen

53,575 görüntüleme • 21 gün önce

Karpathy told Dwarkesh that a 1 billion parameter model, trained on clean data, could hit the intelligence of today's 1.8 trillion parameter frontier. That is a 1,800x compression claim. The math behind it is more defensible than it sounds. When researchers at frontier labs look at random samples from their training corpus, they see stock ticker symbols, broken HTML, forum spam, autogenerated gibberish. Not Wikipedia. Not the Wall Street Journal. The actual pretraining dataset is mostly noise, and the model is burning parameters to vaguely remember all of it. One estimate pegs Llama 3's information compression at 0.07 bits per token. Well-structured English carries around 1.5 bits per token of real information. The trillion-parameter model is holding a roughly 5% resolution image of the internet it trained on. So when a lab ships a 1.8 trillion parameter model, the overwhelming majority of those weights are handling rough memorization. They are compression overhead for a noisy training set, taking up capacity that could be doing reasoning instead. Karpathy's proposal is to separate the two. Build a cognitive core: a small model that contains only the algorithms for reasoning and problem-solving, stripped of encyclopedic memorization. Pair it with external memory the model queries when it needs a fact. A 1 billion parameter reasoner plus retrieval beats a 1.8 trillion parameter model trying to do both. The data already supports this direction. GPT-4o runs at roughly 200 billion parameters and outperforms the original 1.8 trillion GPT-4. Inference costs for GPT-3.5 level performance fell 280x between 2022 and 2024, driven almost entirely by smaller, cleaner, better-architected models. The trend line is pointing where Karpathy says it should. The real implication for anyone tracking the AI trade: data quality is the actual constraint. The companies winning the next phase will be the ones who figured out what to train on, and what to throw away.

Karpathy told Dwarkesh that a 1 billion parameter model, trained on clean data, could hit the intelligence of today's 1.8 trillion parameter frontier. That is a 1,800x compression claim. The math behind it is more defensible than it sounds. When researchers at frontier labs look at random samples from their training corpus, they see stock ticker symbols, broken HTML, forum spam, autogenerated gibberish. Not Wikipedia. Not the Wall Street Journal. The actual pretraining dataset is mostly noise, and the model is burning parameters to vaguely remember all of it. One estimate pegs Llama 3's information compression at 0.07 bits per token. Well-structured English carries around 1.5 bits per token of real information. The trillion-parameter model is holding a roughly 5% resolution image of the internet it trained on. So when a lab ships a 1.8 trillion parameter model, the overwhelming majority of those weights are handling rough memorization. They are compression overhead for a noisy training set, taking up capacity that could be doing reasoning instead. Karpathy's proposal is to separate the two. Build a cognitive core: a small model that contains only the algorithms for reasoning and problem-solving, stripped of encyclopedic memorization. Pair it with external memory the model queries when it needs a fact. A 1 billion parameter reasoner plus retrieval beats a 1.8 trillion parameter model trying to do both. The data already supports this direction. GPT-4o runs at roughly 200 billion parameters and outperforms the original 1.8 trillion GPT-4. Inference costs for GPT-3.5 level performance fell 280x between 2022 and 2024, driven almost entirely by smaller, cleaner, better-architected models. The trend line is pointing where Karpathy says it should. The real implication for anyone tracking the AI trade: data quality is the actual constraint. The companies winning the next phase will be the ones who figured out what to train on, and what to throw away.

Aakash Gupta

507,952 görüntüleme • 2 ay önce

From strangers to PBB’s heartwarming mother & son duo of the nation. Will & Klarisse’s bond wasn’t about romance, it was care in its purest form. A dynamic built on growth, trust & an unbreakable collab. Big winners in our hearts. NANALO KAMI NG ANAK KO! HAMWAH HAMWAH HAMWAH

From strangers to PBB’s heartwarming mother & son duo of the nation. Will & Klarisse’s bond wasn’t about romance, it was care in its purest form. A dynamic built on growth, trust & an unbreakable collab. Big winners in our hearts. NANALO KAMI NG ANAK KO! HAMWAH HAMWAH HAMWAH

Iñigo 🐺

111,607 görüntüleme • 1 yıl önce

This is how DNA turns coded information into functional proteins - the building blocks of the nanomachines that keep the cells in your body alive. This complex process highlights the sophisticated interconnected systems of Life which must all exist together from the beginning, or Life doesn't happen. First, an RNA molecule is copied from a short segment of DNA. Without the specifically ordered DNA information, RNA cannot form, proteins cannot be built, cells stop working, and life ceases to exist. Life is information first. Once the RNA Molecule is created, it gets ejected from the Polymerase where it was built, and it travels through a complex molecular machine called a Nuclear Pore Complex (NPC), which is an information recognition device that controls the flow of information in and out of a cell's nucleus. The NPC is highly complex - composed of about 500-1,000 protein subunits, derived from a set of about 35 distinct proteins. Without this molecular machine, there is no regulation for what goes in and out of the cell's nucleus, which would lead to catastrophic death for the cell. It must exist for cells to exist. Once the RNA Molecule passes through the NPC, it travels to the Ribosome, a 2-part chemical factory which reads the information on RNA and uses it to construct functional proteins using a specifically sequenced chain of amino acids. Once complete, this protein will then be sent to the section of the cell it belongs to integrate into another molecular machine and do its job. The Ribosome is another highly complex molecular machine - consisting of between 56-80 proteins. Without this molecular machines, proteins cannot be built. Proteins are the building blocks of every cell in every organism on Earth. Without Ribosomes, Life doesn't exist. If you're paying attention, you'll start to realize that Life relies on a highly sophisticated interdependent network of complex machines, which all rely on each other for the function of the system. DNA requires the cell for stability, but the cell requires the proteins for its structure and function, but those proteins require DNA and RNA to be built - it's a circle of necessary interdependence. Systems like this cannot be built by evolutionary processes, which requires that each piece of the process is built by gradual incremental means over lots of time. Without all the pieces there, from the beginning, none of it works. There is only one known source of complex & interdependent informational systems like those we find in life: and that is from Intelligence. Molecular Biology is the best and most obvious evidence of the Intelligent Design in Life.

This is how DNA turns coded information into functional proteins - the building blocks of the nanomachines that keep the cells in your body alive. This complex process highlights the sophisticated interconnected systems of Life which must all exist together from the beginning, or Life doesn't happen. First, an RNA molecule is copied from a short segment of DNA. Without the specifically ordered DNA information, RNA cannot form, proteins cannot be built, cells stop working, and life ceases to exist. Life is information first. Once the RNA Molecule is created, it gets ejected from the Polymerase where it was built, and it travels through a complex molecular machine called a Nuclear Pore Complex (NPC), which is an information recognition device that controls the flow of information in and out of a cell's nucleus. The NPC is highly complex - composed of about 500-1,000 protein subunits, derived from a set of about 35 distinct proteins. Without this molecular machine, there is no regulation for what goes in and out of the cell's nucleus, which would lead to catastrophic death for the cell. It must exist for cells to exist. Once the RNA Molecule passes through the NPC, it travels to the Ribosome, a 2-part chemical factory which reads the information on RNA and uses it to construct functional proteins using a specifically sequenced chain of amino acids. Once complete, this protein will then be sent to the section of the cell it belongs to integrate into another molecular machine and do its job. The Ribosome is another highly complex molecular machine - consisting of between 56-80 proteins. Without this molecular machines, proteins cannot be built. Proteins are the building blocks of every cell in every organism on Earth. Without Ribosomes, Life doesn't exist. If you're paying attention, you'll start to realize that Life relies on a highly sophisticated interdependent network of complex machines, which all rely on each other for the function of the system. DNA requires the cell for stability, but the cell requires the proteins for its structure and function, but those proteins require DNA and RNA to be built - it's a circle of necessary interdependence. Systems like this cannot be built by evolutionary processes, which requires that each piece of the process is built by gradual incremental means over lots of time. Without all the pieces there, from the beginning, none of it works. There is only one known source of complex & interdependent informational systems like those we find in life: and that is from Intelligence. Molecular Biology is the best and most obvious evidence of the Intelligent Design in Life.

Divinely Designed

62,517 görüntüleme • 6 ay önce

From Palantir and Two Sigma to turning mechanistic interpretability into a production-grade platform, Mark Bissell (Member of Technical Staff) and Myra Deng (Head of Product) are building Goodfire on the belief that “understanding what’s inside the model” is the path to safer, more controllable AI and Goodfire just raised a $150M Series B at a $1.25B valuation to scale that bet. We sat down with Mark and Myra to unpack what actionable interpretability looks like beyond demos: why SAE features can lose to raw-activation probes on real detection tasks, how they deployed token-level PII monitoring with Rakuten across English + Japanese without training on real customer data, and what it takes to do real-time steering on a trillion-parameter model (including the infamous Gen-Z feature). We also dig into hallucination detection, surgical post-training edits for bias/unlearning, the emerging equivalence between activation steering and in-context learning, and their push to bring interpretability into training so model development isn’t just “data in, weights out.” Goodfire mark bissell Myra Deng

From Palantir and Two Sigma to turning mechanistic interpretability into a production-grade platform, Mark Bissell (Member of Technical Staff) and Myra Deng (Head of Product) are building Goodfire on the belief that “understanding what’s inside the model” is the path to safer, more controllable AI and Goodfire just raised a $150M Series B at a $1.25B valuation to scale that bet. We sat down with Mark and Myra to unpack what actionable interpretability looks like beyond demos: why SAE features can lose to raw-activation probes on real detection tasks, how they deployed token-level PII monitoring with Rakuten across English + Japanese without training on real customer data, and what it takes to do real-time steering on a trillion-parameter model (including the infamous Gen-Z feature). We also dig into hallucination detection, surgical post-training edits for bias/unlearning, the emerging equivalence between activation steering and in-context learning, and their push to bring interpretability into training so model development isn’t just “data in, weights out.” Goodfire mark bissell Myra Deng

Latent.Space

19,389 görüntüleme • 5 ay önce

1/ In 1979, Washington, Riyadh & Islamabad lit a fire which birthed the world’s first "global jihad". From CIA cash to Saudi sermons to Pakistani training camps, the machine was built. One man’s journey through it all: Ahmad Zidan, the new advisor of Abu Mohammad al-Julani🧵

1/ In 1979, Washington, Riyadh & Islamabad lit a fire which birthed the world’s first "global jihad". From CIA cash to Saudi sermons to Pakistani training camps, the machine was built. One man’s journey through it all: Ahmad Zidan, the new advisor of Abu Mohammad al-Julani🧵

Kevork Almassian

37,336 görüntüleme • 11 ay önce

Apple built a large foundation model and fine-tuned it on multiple tasks. But they are doing something very clever: They load a single model in memory and use different adapters to specialize the model on the fly. I recorded a video to show you how to write the code to do the same thing Apple is doing. I explain everything step by step. Here is what I'll show you in the video: 1. We'll load two datasets 2. Then load a large model 3. Then, we'll fine-tune the model on both datasets I'll use LoRA to fine-tune the model. This process creates two small adapters, each specializing in solving one of the datasets. The base model's original parameters will remain unchanged. From here: 4. We'll generate a list of tasks 5. We'll load the correct adapter to solve each task The large model I'm using needs 346 MB of memory, but I only need to load it once. Each adapter is only 2.7 MB. I only need to load the base model once and pair it with any of the fine-tuned adapters. Minimum memory footprint and I can solve multiple tasks. Hope this helps!

Apple built a large foundation model and fine-tuned it on multiple tasks. But they are doing something very clever: They load a single model in memory and use different adapters to specialize the model on the fly. I recorded a video to show you how to write the code to do the same thing Apple is doing. I explain everything step by step. Here is what I'll show you in the video: 1. We'll load two datasets 2. Then load a large model 3. Then, we'll fine-tune the model on both datasets I'll use LoRA to fine-tune the model. This process creates two small adapters, each specializing in solving one of the datasets. The base model's original parameters will remain unchanged. From here: 4. We'll generate a list of tasks 5. We'll load the correct adapter to solve each task The large model I'm using needs 346 MB of memory, but I only need to load it once. Each adapter is only 2.7 MB. I only need to load the base model once and pair it with any of the fine-tuned adapters. Minimum memory footprint and I can solve multiple tasks. Hope this helps!

Santiago

84,747 görüntüleme • 1 yıl önce

Dr. Kevin McKernan, a pioneering genomicist, pulls no punches on the RNA tech at the heart of mRNA vaccines. In a candid reflection, he spotlights how swapping every uracil for pseudouridine—a tweak to evade immune detection—fundamentally rewires the molecule's behavior. "You have no idea what you're doing," he says, likening it to flipping one tiny methyl group in a DNA sequencer and watching the whole system crash. Enzymes are that finicky. Scale it up: 800+ such mods on a single RNA strand? You're venturing into uncharted waters—unpredictable folding, cellular processing, even triggering RNA interference cascades. McKernan calls it "reckless insanity," especially galling in biotech's epicenter like Massachusetts, where Pfizer, Moderna, and PCR giants cluster like vines in a pandemic-fueled vineyard. Why the silence from peers? Is innovation outpacing caution, or are incentives clouding the view? As we unpack these breakthroughs, McKernan's plea echoes: Science thrives on scrutiny. What's your take on balancing speed with safety in biotech's high-stakes race?

Dr. Kevin McKernan, a pioneering genomicist, pulls no punches on the RNA tech at the heart of mRNA vaccines. In a candid reflection, he spotlights how swapping every uracil for pseudouridine—a tweak to evade immune detection—fundamentally rewires the molecule's behavior. "You have no idea what you're doing," he says, likening it to flipping one tiny methyl group in a DNA sequencer and watching the whole system crash. Enzymes are that finicky. Scale it up: 800+ such mods on a single RNA strand? You're venturing into uncharted waters—unpredictable folding, cellular processing, even triggering RNA interference cascades. McKernan calls it "reckless insanity," especially galling in biotech's epicenter like Massachusetts, where Pfizer, Moderna, and PCR giants cluster like vines in a pandemic-fueled vineyard. Why the silence from peers? Is innovation outpacing caution, or are incentives clouding the view? As we unpack these breakthroughs, McKernan's plea echoes: Science thrives on scrutiny. What's your take on balancing speed with safety in biotech's high-stakes race?

Camus

51,428 görüntüleme • 8 ay önce

🔥Really excited to see the release of PAN world model, a project I had been working over the past years. PAN is a general world model capable of simulating physical, agentic, and nested worlds, synthesizing infinite interactive experiences for training AI agents. Building on top of pretrained LLMs and video diffusion models, PAN connects language, perception, action, and latent thoughts, for long-horizon simulation and reasoning. PAN shows overwhelming performance gains over JEPA-2, Cosmos-2, and other prior models. More in the thread👇 ... 1/

🔥Really excited to see the release of PAN world model, a project I had been working over the past years. PAN is a general world model capable of simulating physical, agentic, and nested worlds, synthesizing infinite interactive experiences for training AI agents. Building on top of pretrained LLMs and video diffusion models, PAN connects language, perception, action, and latent thoughts, for long-horizon simulation and reasoning. PAN shows overwhelming performance gains over JEPA-2, Cosmos-2, and other prior models. More in the thread👇 ... 1/

Zhiting Hu

31,195 görüntüleme • 8 ay önce