@zuxfoucault.bsky.social
Reposted
Brought to you by Austin Cheng (@auhcheng.bsky.social) — meet the newest member of our team: Quetzal!
Named after Quetzalcoatl, the Aztec god of creation, Quetzal is a simple yet scalable model for building 3D molecules atom by atom.
📜 arxiv.org/abs/2505.13791
[1/4]
Named after Quetzalcoatl, the Aztec god of creation, Quetzal is a simple yet scalable model for building 3D molecules atom by atom.
📜 arxiv.org/abs/2505.13791
[1/4]
May 23, 2025 at 4:13 PM
Brought to you by Austin Cheng (@auhcheng.bsky.social) — meet the newest member of our team: Quetzal!
Named after Quetzalcoatl, the Aztec god of creation, Quetzal is a simple yet scalable model for building 3D molecules atom by atom.
📜 arxiv.org/abs/2505.13791
[1/4]
Named after Quetzalcoatl, the Aztec god of creation, Quetzal is a simple yet scalable model for building 3D molecules atom by atom.
📜 arxiv.org/abs/2505.13791
[1/4]
Reposted
🤖🧫Thrilled to share our latest work: "RoboCulture: A Robotics Platform for Automated Biological Experimentation", now available on arXiv!
📃 arxiv.org/abs/2505.14941
[1/3]
📃 arxiv.org/abs/2505.14941
[1/3]
RoboCulture: A Robotics Platform for Automated Biological Experimentation
Automating biological experimentation remains challenging due to the need for millimeter-scale precision, long and multi-step experiments, and the dynamic nature of living systems. Current liquid hand...
arxiv.org
May 23, 2025 at 1:14 AM
🤖🧫Thrilled to share our latest work: "RoboCulture: A Robotics Platform for Automated Biological Experimentation", now available on arXiv!
📃 arxiv.org/abs/2505.14941
[1/3]
📃 arxiv.org/abs/2505.14941
[1/3]
Reposted
Check out our work on corrected product formulas (CPF) to lower the cost of #hamiltonian #simulation #quantumalgorithms @uoftcompsci.bsky.social @cqiqc-uoft.bsky.social @vectorinstitute.ai @chemuoft.bsky.social
⚛️🖥️Excited to share our work on simulating Hamiltonian evolution on quantum computers. Led by our group member Mohsen Bagherimehrab, we introduce Corrected Product Formulas (CPFs), which significantly improve the performance of standard product formulas. [1/4]
March 25, 2025 at 2:52 AM
Check out our work on corrected product formulas (CPF) to lower the cost of #hamiltonian #simulation #quantumalgorithms @uoftcompsci.bsky.social @cqiqc-uoft.bsky.social @vectorinstitute.ai @chemuoft.bsky.social
Reposted
Thanks to #BNNBloomberg for having AC director @aspuru.bsky.social on to chat about his latest research w/Insilico Medicine + #UofT researchers that uses AI + quantum computing to speed up the discovery of anti-cancer drugs. Watch now 📺 www.bnnbloomberg.ca/video/shows/...
February 6, 2025 at 3:41 PM
Thanks to #BNNBloomberg for having AC director @aspuru.bsky.social on to chat about his latest research w/Insilico Medicine + #UofT researchers that uses AI + quantum computing to speed up the discovery of anti-cancer drugs. Watch now 📺 www.bnnbloomberg.ca/video/shows/...
Reposted
Sharing great work led by Raul Ortega Ochoa and in collaboration with Tonio Buonassisi @tvegge.bsky.social and myself.
A tomographic representation of materials or molecular descriptors clarifies the ideas of “forward” and “inverse” design, treating them equally.
arxiv.org/abs/2501.18163
A tomographic representation of materials or molecular descriptors clarifies the ideas of “forward” and “inverse” design, treating them equally.
arxiv.org/abs/2501.18163
January 31, 2025 at 9:48 PM
Sharing great work led by Raul Ortega Ochoa and in collaboration with Tonio Buonassisi @tvegge.bsky.social and myself.
A tomographic representation of materials or molecular descriptors clarifies the ideas of “forward” and “inverse” design, treating them equally.
arxiv.org/abs/2501.18163
A tomographic representation of materials or molecular descriptors clarifies the ideas of “forward” and “inverse” design, treating them equally.
arxiv.org/abs/2501.18163
Reposted
Part 1: Exciting news! Our research on quantum drug discovery is now in Nature Biotechnology. A Quantum Computer helped design KRAS inhibitors, an elusive cancer target. These inhibitors were successfully synthesized and tested, marking a milestone in healthcare innovation. 🚀
#QuantumComputing #KRAS
#QuantumComputing #KRAS
January 22, 2025 at 7:59 PM
Part 1: Exciting news! Our research on quantum drug discovery is now in Nature Biotechnology. A Quantum Computer helped design KRAS inhibitors, an elusive cancer target. These inhibitors were successfully synthesized and tested, marking a milestone in healthcare innovation. 🚀
#QuantumComputing #KRAS
#QuantumComputing #KRAS
Reposted
Glad to see our research on thr first #quantum + #ai application with experimental confirmation for #drugdiscovery out in Nature Biotechnology!
@ghazivakili.bsky.social @elonverse.bsky.social Igor Stagljar @thematterlab.bsky.social
@ghazivakili.bsky.social @elonverse.bsky.social Igor Stagljar @thematterlab.bsky.social
While we often hear of potential long-range synergy between #AI (which requires enormous computing power) + #quantumcomputing, today we have first-of-its-kind research from #UofT + Insilico Medicine that demonstrates it firsthand. Follow the link to learn more or read on below.
U of T researchers develop new approach using quantum computers to accelerate drug discovery
A study co-led by University of Toronto researchers and Insilico Medicine has demonstrated for the first time the potential of quantum computing and artificial intelligence to transform the drug disco...
temertymedicine.utoronto.ca
January 22, 2025 at 9:12 PM
Glad to see our research on thr first #quantum + #ai application with experimental confirmation for #drugdiscovery out in Nature Biotechnology!
@ghazivakili.bsky.social @elonverse.bsky.social Igor Stagljar @thematterlab.bsky.social
@ghazivakili.bsky.social @elonverse.bsky.social Igor Stagljar @thematterlab.bsky.social
Reposted
I really like @yoshuabengio.bsky.social's humility in the recent Machine Learning Street Talk episode on AI safety, whether it's on probability of doom, different risk scenarios, AGI timeline, or other topics:
Yoshua Bengio - Designing out Agency for Safe AI
Machine Learning Street Talk (MLST) · Episode
open.spotify.com
January 17, 2025 at 1:19 PM
I really like @yoshuabengio.bsky.social's humility in the recent Machine Learning Street Talk episode on AI safety, whether it's on probability of doom, different risk scenarios, AGI timeline, or other topics:
Reposted
Article: "Active Inference Tree Search in Large POMDPs"
Coauthors: Domenico Maisto, Francesco Gregoretti, Giovanni Pezzulo
www.sciencedirect.com/science/arti...
Coauthors: Domenico Maisto, Francesco Gregoretti, Giovanni Pezzulo
www.sciencedirect.com/science/arti...
Active Inference Tree Search in Large POMDPs
The ability to plan ahead efficiently is key for both living organisms and artificial systems. Model-based planning and prospection are widely studied…
www.sciencedirect.com
January 10, 2025 at 3:21 PM
Article: "Active Inference Tree Search in Large POMDPs"
Coauthors: Domenico Maisto, Francesco Gregoretti, Giovanni Pezzulo
www.sciencedirect.com/science/arti...
Coauthors: Domenico Maisto, Francesco Gregoretti, Giovanni Pezzulo
www.sciencedirect.com/science/arti...
Reposted
Goodfire.ai is open-sourcing Sparse Autoencoders (SAEs) for Llama 3.3 70B and Llama 3.1 8B!
SAEs are interpreter models that help us understand how language models process information internally by decomposing neural activations into interpretable features.
SAEs are interpreter models that help us understand how language models process information internally by decomposing neural activations into interpretable features.
January 10, 2025 at 7:09 PM
Goodfire.ai is open-sourcing Sparse Autoencoders (SAEs) for Llama 3.3 70B and Llama 3.1 8B!
SAEs are interpreter models that help us understand how language models process information internally by decomposing neural activations into interpretable features.
SAEs are interpreter models that help us understand how language models process information internally by decomposing neural activations into interpretable features.
Reposted
I stand with Geoff Hinton on the severity of the risks of building superintelligent AIs, and I disagree with Yann's response to this statement by Geoff visible below. Geoff was talking about "things" that are like a new (non-biological) species.
Yann's facebook post (and following comments) is here:
Yann's facebook post (and following comments) is here:
Yann LeCun
That's just false.
Most of us are intimately familiar with interacting with people who are smarter than us.
Most scientists actively seek to work with people who are smarter than them in at least a...
www.facebook.com
January 11, 2025 at 12:18 PM
I stand with Geoff Hinton on the severity of the risks of building superintelligent AIs, and I disagree with Yann's response to this statement by Geoff visible below. Geoff was talking about "things" that are like a new (non-biological) species.
Yann's facebook post (and following comments) is here:
Yann's facebook post (and following comments) is here:
Reposted
In a recent article, my group and I laid out some of the recommendations I have learned over the year about research. In 2025, our lab will double down on following our own advice. pubs.rsc.org/en/content/a...
Spiers Memorial Lecture: How to do impactful research in artificial intelligence for chemistry and materials science
Machine learning has been pervasively touching many fields of science. Chemistry and materials science are no exception. While machine learning has been making a great impact, it is still not reaching...
pubs.rsc.org
December 31, 2024 at 4:56 PM
In a recent article, my group and I laid out some of the recommendations I have learned over the year about research. In 2025, our lab will double down on following our own advice. pubs.rsc.org/en/content/a...
Reposted
🧵(1/7) Have you ever wanted to combine different pre-trained diffusion models but don't have time or data to retrain a new, bigger model?
🚀 Introducing SuperDiff 🦹♀️ – a principled method for efficiently combining multiple pre-trained diffusion models solely during inference!
🚀 Introducing SuperDiff 🦹♀️ – a principled method for efficiently combining multiple pre-trained diffusion models solely during inference!
December 28, 2024 at 2:32 PM
🧵(1/7) Have you ever wanted to combine different pre-trained diffusion models but don't have time or data to retrain a new, bigger model?
🚀 Introducing SuperDiff 🦹♀️ – a principled method for efficiently combining multiple pre-trained diffusion models solely during inference!
🚀 Introducing SuperDiff 🦹♀️ – a principled method for efficiently combining multiple pre-trained diffusion models solely during inference!
Reposted
Come and join a growing and exciting effort at the @uoft.bsky.social on #ai for #materialsdiscovery #chemsky as #assistantprofessor Please reshare massively
Faculty position doing AI for materials discovery! Come join us at the @uoftmse @uoft.bsky.social jobs.utoronto.ca/job/Toronto-... Plenty of exciting opportunities in the GTA and with @acceleration_c and @vectorinst.bsky.social in particular! Please share and retweet!
Assistant Professor - AI Materials Discovery
Assistant Professor - AI Materials Discovery
jobs.utoronto.ca
December 23, 2024 at 10:40 AM
Come and join a growing and exciting effort at the @uoft.bsky.social on #ai for #materialsdiscovery #chemsky as #assistantprofessor Please reshare massively
Reposted
Understanding Transformer reasoning capabilities via graph algorithms
They "Google" provide a comprehensive evaluation of transformer models’ graph reasoning capabilities and demonstrate that they often outperform more domain-specific graph neural networks.
research.google/blog/underst...
They "Google" provide a comprehensive evaluation of transformer models’ graph reasoning capabilities and demonstrate that they often outperform more domain-specific graph neural networks.
research.google/blog/underst...
December 22, 2024 at 7:36 PM
Understanding Transformer reasoning capabilities via graph algorithms
They "Google" provide a comprehensive evaluation of transformer models’ graph reasoning capabilities and demonstrate that they often outperform more domain-specific graph neural networks.
research.google/blog/underst...
They "Google" provide a comprehensive evaluation of transformer models’ graph reasoning capabilities and demonstrate that they often outperform more domain-specific graph neural networks.
research.google/blog/underst...
Reposted
exciting new workshop announcement!! join us in Singapore for Frontiers in Probabilistic Inference: Learning Meets Sampling 🌏⚡️😃 details below 👇 #ICLR2025
🔊 Super excited to announce the first ever Frontiers of Probabilistic Inference: Learning meets Sampling workshop at #ICLR2025 @iclr-conf.bsky.social!
🔗 website: sites.google.com/view/fpiwork...
🔥 Call for papers: sites.google.com/view/fpiwork...
more details in thread below👇 🧵
🔗 website: sites.google.com/view/fpiwork...
🔥 Call for papers: sites.google.com/view/fpiwork...
more details in thread below👇 🧵
December 18, 2024 at 8:38 PM
exciting new workshop announcement!! join us in Singapore for Frontiers in Probabilistic Inference: Learning Meets Sampling 🌏⚡️😃 details below 👇 #ICLR2025
Reposted
Check out our work on stereochemistry-aware molecular generative models! We add stereochemistry to benchmark oracles: rediscovery, docking, CD spectra (new!) & compare stereo/non-stereo models. More 👇
w/ Edwin Yu, Naruki Yoshikawa, @valencekjell.com @aspuru.bsky.social
➡️ doi.org/10.26434/che...
w/ Edwin Yu, Naruki Yoshikawa, @valencekjell.com @aspuru.bsky.social
➡️ doi.org/10.26434/che...
Stereochemistry-aware string-based molecular generation
This study investigates the impact of incorporating stereochemical information, a crucial aspect of computational drug discovery and materials design, in molecular generative modelling. We present a c...
chemrxiv.org
December 17, 2024 at 6:39 PM
Check out our work on stereochemistry-aware molecular generative models! We add stereochemistry to benchmark oracles: rediscovery, docking, CD spectra (new!) & compare stereo/non-stereo models. More 👇
w/ Edwin Yu, Naruki Yoshikawa, @valencekjell.com @aspuru.bsky.social
➡️ doi.org/10.26434/che...
w/ Edwin Yu, Naruki Yoshikawa, @valencekjell.com @aspuru.bsky.social
➡️ doi.org/10.26434/che...
Reposted
In collaboration with Oxford colleagues and led by @zijian-zhang.bsky.social @scao.me we made strides in Agent-driven #selfdrivinglabs ! Furthermore we applied them to #quantumcomputing merging two of my favorite fields! Check our preprint, and code.
We are happy to introduce k-agents: our knowledge-centred laboratory automation framework 🤖.
⚛️ Quantum computers are becoming more powerful—but also more complex.
🤔 How do we build AI systems that know how to operate them?
📄 Read the paper: arxiv.org/abs/2412.07978
#AI #LLM #QuantumComputing
⚛️ Quantum computers are becoming more powerful—but also more complex.
🤔 How do we build AI systems that know how to operate them?
📄 Read the paper: arxiv.org/abs/2412.07978
#AI #LLM #QuantumComputing
Agents for self-driving laboratories applied to quantum computing
Fully automated self-driving laboratories are promising to enable high-throughput and large-scale scientific discovery by reducing repetitive labour. However, effective automation requires deep integr...
arxiv.org
December 12, 2024 at 11:23 PM
In collaboration with Oxford colleagues and led by @zijian-zhang.bsky.social @scao.me we made strides in Agent-driven #selfdrivinglabs ! Furthermore we applied them to #quantumcomputing merging two of my favorite fields! Check our preprint, and code.
Reposted
1/3 🧵Quantum conundrum: we want expressive circuits, but current hardware only allows short coherence times, and so more parameters = more problems. Check out our #NeurIPS2024 paper “Quantum Deep Equilibrium Models”.
December 9, 2024 at 8:39 AM
1/3 🧵Quantum conundrum: we want expressive circuits, but current hardware only allows short coherence times, and so more parameters = more problems. Check out our #NeurIPS2024 paper “Quantum Deep Equilibrium Models”.
Reposted
The inspiration from @ncoop57.bsky.social
to get tmux to do the heavy lifting to make ShellSage work is genius! Thanks to this, ShellSage is <100 lines of code (plus comments and prompts).
But the result is amazing. Try it out and tell us how you go.
Details here: www.answer.ai/posts/2024-1...
to get tmux to do the heavy lifting to make ShellSage work is genius! Thanks to this, ShellSage is <100 lines of code (plus comments and prompts).
But the result is amazing. Try it out and tell us how you go.
Details here: www.answer.ai/posts/2024-1...
ShellSage - Your AI Bash Buddy – Answer.AI
ShellSage saves sysadmins’ sanity by solving shell script snafus super swiftly
www.answer.ai
December 5, 2024 at 8:30 PM
The inspiration from @ncoop57.bsky.social
to get tmux to do the heavy lifting to make ShellSage work is genius! Thanks to this, ShellSage is <100 lines of code (plus comments and prompts).
But the result is amazing. Try it out and tell us how you go.
Details here: www.answer.ai/posts/2024-1...
to get tmux to do the heavy lifting to make ShellSage work is genius! Thanks to this, ShellSage is <100 lines of code (plus comments and prompts).
But the result is amazing. Try it out and tell us how you go.
Details here: www.answer.ai/posts/2024-1...
Reposted
Will simply scaling up LLMs get us to AGI? My feature for Nature, w/ inputs from @fchollet.bsky.social @rao2z.bsky.social @yoshuabengio.bsky.social @melaniemitchell.bsky.social @dileeplearning.bsky.social @andrewgwils.bsky.social, Raia Hadsell, Keyon Vafa, Karl Friston www.nature.com/articles/d41...
How close is AI to human-level intelligence?
Large language models such as OpenAI’s o1 have electrified the debate over achieving artificial general intelligence, or AGI. But they are unlikely to reach this milestone on their own.
www.nature.com
December 3, 2024 at 9:55 PM
Will simply scaling up LLMs get us to AGI? My feature for Nature, w/ inputs from @fchollet.bsky.social @rao2z.bsky.social @yoshuabengio.bsky.social @melaniemitchell.bsky.social @dileeplearning.bsky.social @andrewgwils.bsky.social, Raia Hadsell, Keyon Vafa, Karl Friston www.nature.com/articles/d41...
Reposted
📣 New: AI depends on immense resources, mostly controlled by large companies.
This means "open" AI… isn't very open.
📄 From me, @meredithmeredith.bsky.social and @smw.bsky.social in Nature:
www.nature.com/articles/s41...
This means "open" AI… isn't very open.
📄 From me, @meredithmeredith.bsky.social and @smw.bsky.social in Nature:
www.nature.com/articles/s41...
December 4, 2024 at 5:31 PM
📣 New: AI depends on immense resources, mostly controlled by large companies.
This means "open" AI… isn't very open.
📄 From me, @meredithmeredith.bsky.social and @smw.bsky.social in Nature:
www.nature.com/articles/s41...
This means "open" AI… isn't very open.
📄 From me, @meredithmeredith.bsky.social and @smw.bsky.social in Nature:
www.nature.com/articles/s41...