Chris Price

On Cloud has the opportunity to do the funniest thing

Update on this paper from last year on AI in scientific discovery. MIT put out a press release: "no confidence in the provenance, reliability or validity of the data...[or] veracity of the research." "The author is no longer at MIT."

AI agents are poised to transform the commercialization of advanced materials. At Atomscale, we’re driving this transformation with our next-generation AI agents, purpose-built for atomic-scale engineering. Read on: The development of new materials and their integration into

Exciting news — Atomic Data Sciences is now Atomscale! Our new name reflects our evolution from automating data analysis in materials science to building a comprehensive intelligence layer for atomic-scale engineering. As we grow, our mission remains clear: to drive

@xie_tian @Sergei_Imaging Simulation is nice to have but at the end of the day, physical process design, optimization, scaling, and yield ramp consumes 99% of the lifecycle for a new materials platform.

@xie_tian @Sergei_Imaging Lab-style automation / SDLs in the current framework will never scale to production, but characterization will. Better continuous feedback for existing equipment already unlocks massive efficiency gains.

Real progress is being able to make and characterize 10,000 materials from sufficiently complex composition space in a day. Really doesn't matter that much how far computation is scaled beyond SoTA

New post on the impact of real-time feedback with end-to-end AI for atomically engineered materials - making a lot of progress @AtomicDataSci ! Check out our substack at atomicdatasciences dot substack dot com or the link in the replies.

@RehaMathur @draparente General RW ability for models>tools is critical - but the most limiting step is extracting/passing context-rich and physics-rich info between tools. Step 1 is automating ETL pipelines for process and characterization data; building this with closed-loop agents @AtomicDataSci

@draparente yes precisely, we need to allow models to both "read & write" tools and super cool, something like that would be awesome, esp i imagine companies like thermo or bruker might be v slow to adopt such standards

Self-driving labs & autonomous science are super exciting, but feels like we will hit a bottleneck in automation as cognition improves. There's an inherent asymmetry - the most valuable experiments to automate (like in vivo work) are often the least automatable

We’re excited to share this perspective, co-authored by co-founder @ccprice19, on the current state of AI/ML application to accelerate discovery and synthesis of advanced electronic materials. We continue to believe there is substantial value to purpose-built AI/ML tools for

@XirtamEsrevni @ChurchillMic @Robert_Palgrave Elevating real-world materials characterization data to be trained on at scale is needed (working on this @AtomicDataSci). Even with better theory+more compute, synthesis is tied to the individual equipment and simulation-only digital twins will be extremely difficult.

@ChurchillMic @Robert_Palgrave I'm not sure there is an "instead" it's just being careful and knowing that there are limitations to the data we use.

Nice tool, glad they open-sourced it, but: 1. I believe @Robert_Palgrave has opined, critically, on this statement in previous comments: "... MatterGen, with the caveat of compositional disorder between Ta and Cr." I think this is problematic, rather than a minor point. It suggests that ground-state structure prediction can be misleading when described as "discovery." In reality, finite-temperature effects and kinetic factors play significant roles, which means such predictions do not fully capture the necessary temperature & processing complexity (obviously). 2. In my opinion, just predicting a "novel" composition and structure that is GS stable with ideal properties is insufficient for practical use by materials synthesis chemists. We must provide guidance on a "recipe" for synthesis. Or, more future focused, integrate this approach into a self-driving lab with both a prediction head (structure/property prediction) and a process head (synthesis/process optimization) would make it more actionable and improve prediction head via RLHF. Others are doing this, but my hope is they are doing so in a vertical integration schema (@josephfkrause ?) . 3. It is essential to recognize that, to my knowledge, ML/AI models used for material predictions are strictly trained on DFT data, meaning their predictions are inherently constrained by the accuracy and limitations of DFT. It is an approximation to the many-body e-e interactions by reformulating the problem into an effective single-electron system, corrected by our best informed guess for an electron density functional. While DFT works well for many material predictions, its approximations falter when dealing with systems where strong e-e correlations dominate (e.g., superconductors, Mott insulators). Therefore, caution and skepticism are necessary when applying DFT-based ML/AI models to such materials. Feel free to criticize my comments, this is meant to engage.

An early demo of major efficiency gains unlocked by making it quick and easy to create datasets and proxy models over real samples of cutting edge materials. Much more to come!

RT @AtomicDataSci: Great newsletter @techreview on the recent Nobel Prize win for computational protein discovery and great summary @VanRij…

Breaking the fourth wall of Founder Mode™ to share our vision to solve decades-old challenges in advanced materials, ultimately shaping the future of computing and energy. Follow along with us: atomicdatasciences.substack.com/p/overcoming-m…

A great day to discuss AI for advanced materials synthesis at the NEMC Fall Meeting @Mass_Tech - check out our platform to quantify your analytical and microscopy data for fab R&D, scale-up, and SPC faster than an @TomBrady 2 minute drive

Ending the week with a big update from @AtomicDataSci - real automation and AI accelerating advanced materials synthesis. Chat with us about AI for materials and stay tuned for more to come!

Great to have Prof. Swastik Kar highlighting @AtomicDataSci work with Northeastern University’s experimental quantum lab at Quantum Massachusetts 2023 - thanks @Mass_Tech @QuantumDaily for putting on a great event!

Big weekend for us at NAMBE 2023 - looking forward to showing our product!