We build enterprise software to power the Industrial Generative AI revolution. 

Zapata AI

Industrial problems share seven common challenges that make them difficult to solve with traditional computational solutions: data disarray, large solution spaces, unpredictability, time constraints, compute constraints, demands for accuracy and precision, and security requirements. 

Importantly, the challenges that define industrial problems are not limited to settings that would traditionally be considered “industrial,” such as manufacturing plants, warehouses, construction sites, or resource extraction sites. They can be found in enterprises in any industry that operate on a global, industrial scale.

Learn more about the challenges of industrial problems and how #IndustrialGenerativeAI addresses each challenge:

https://lnkd.in/e6swza2W

At the end of last week, a group led by Alán Aspuru-Guzik at the University of Toronto with Zapata AI and Insilico Medicine quietly dropped this pre-print on ArXiv to get some feedback before submitting it to a peer-reviewed journal (so please do take part - link in the comments). We are not making any claims of superiority to GPUs or any other form of computing or claims of compound superiority. Just a very simple claim that a molecule that was generated on a real quantum computer was synthesized and tested. We can call it a hybrid approach, since we had to use a GPU-driven generative system. And it is a real possibility and a real milestone. 
This approach will get a fair dose of criticism and skepticism. But you probably remember how our joint work with Alán Aspuru-Guzik in 2018 (published in 2019) called Generative Tensorial Reinforcement Learning (GENTRL) was criticized. But now, our generative software is deployed by 10 out of top 20 pharma companies with a substantial community and validation. 
As quantum computing scales up, more generative chemistry and generative biology tasks will be possible. I hope that we will be able to simulate entire worlds with populations from first principles (see paper titled "Complexity of life sciences in quantum and AI era"). We will likely see significant hype around quantum as generative software advances - that is why we wanted to set some bar (low bar, but still a bar) for experimental validation. And I don't want to sound like Derek Lowe (who I deeply respect), but one thing that I learned over the past 10 years is that what matters are clinically-approved drugs helping patients. And there you need to move with a speed of traffic even if you can accelerate from 0 to PCC in fraction of the time and cost. 

I am excited to show work done in my  lab and Igor Stagljar’s  at University of Toronto  Vector Institute  collaboration with Zapata AI ,  Insilico Medicine and Christoph Gorgulla ‘s lab at St Jude’s. We develop a workflow that uses classical and quantum computers for drug discovery. We apply to find a novel inhibitor for the “undruggable” KRAS family of cancer-related proteins. 

We incorporate a small (16-qubit IBM  generative QCBM model in a complex AI pipeline that employs Zapata AI ‘s Orquestra, Insilico Medicine ‘s powerful Chemistry42 suite and our own group’s SELFIES, Stoned-Selfies and Christoph Gorgulla ‘s VirtualFlow. Igor Stagljar ‘s lab contributed his lab’s impressive assays to characterize the compounds. 

This would not have been possible if our teams were not so interdisciplinary and connected many of the dots. The visions of Yudong Cao Christopher Savoie, Alex Zhavoronkov combined for this moment. 

Being co-founder and scientific advisor for Zapata AI and a very early scientific advisor of Insilico Medicine this maxes me very happy and proud for all the teams coming together.

Many important players include Mohammad Ghazi Vakili Akshat Nigam Alex Aliper

Department of Computer Science, University of Toronto Department of Chemistry at the University of Toronto

For the first time, quantum-enhanced generative AI has generated viable cancer drug candidates, targeting the historically "undruggable" KRAS protein.

Working in collaboration with scientists at Insilico Medicine, the University of Toronto, and St. Jude Children's Research Hospital, Zapata AI scientists demonstrated the first instance of a generative model running partially on quantum hardware outperforming state-of-the-art classical models in generating viable cancer drug candidates. 

The research points to a promising future of hybrid quantum generative AI for drug discovery using today’s quantum devices. Two molecules generated by the quantum-enhanced generative model were distinct from existing KRAS inhibitors and showed a superior binding affinity over the molecules generated by purely classical models.

"We’re looking forward to taking this research further to move the discovered molecules through the drug discovery pipeline, apply our methodology to other disease targets, and extend our quantum-enhanced generative AI to other industrial use cases with complex design challenges," said Yudong Cao, CTO and Co-founder of Zapata AI. 

The research is currently published on ArXiv as it awaits peer review. To access the research and learn more, follow the link below: 

https://lnkd.in/d625eMx7


“Our technology is derived from math inspired by quantum physics… The hard part is turning that discipline of physics into useful technology. Fortunately, our work in this area has many transferable and positive implications for generative AI.” -- Christopher Savoie, Zapata AI’s CEO and co-founder, via the recent investor and analyst webinar as noted in new Inside Quantum Technology article here: https://lnkd.in/gGWPfeN6 

ZPTA

AI:ZAPATA

Zapata Computing Holdings Inc. Warrant

Last Close

0.00

25/10 20:00

Market Cap

744.94K

Beta: -

Volume Today

66.52K

Avg: -

Filings Not Available