San Francisco, CA
Joined June 2009
- Tweets343
- Following1.1K
- Followers867
- Likes3.2K
My current thoughts and action items on the theoretical scientific research disruption due to AI, based on reading all the action on X and chatting with research, market intelligence and startup leaders the last two weeks. Opinions subject to change overnight !
* Credit as usually intended in academia will devalue (intellectual origination of the idea/question); credit as usually intended in entrepreneurship (idea worth little, execution is what matters) will emerge as a strong currency.
Action item: embrace the tinkering and hybridization with AI tools in all research and innovation endeavors.
* Reputation, track record, trustworthiness, ambition, opinions, ecosystem capability (including access to quantum compute) will matter more to get collaborations and funding than being a technically specialized powerhouse. Mutual trust and alignment matters more than anything when product quality and quantity is commoditized.
Action item: foster project inclusivity in my community - we are all together as humans and colleagues navigating the singularity.
* Peer review human verification will likely become an “emergent phenomenon”: when some result or preprint becomes sufficiently popular it will be picked up and built upon and automatically de-facto reviewed, possibly by journals or their equivalent in this new world. I think there is a play in education/training here.
Action item: use social media and web posting generously; be much more selective in formal journal/conference submissions: overall quality of the artifacts need to be 5x than before or it is not worth it.
* Scientific results that interface the physical world, leverage unique nonpublic data, or present phenomenological / heuristic / non-rigorous evidence and insights, are now more precious and difficult than analytical or numerical advances. I still believe in the possibility of huge genius leaps of creativity by humans making the difference, but I suspect we will see those feats from AI multiagent reasoning very soon. Our job will be to steer the investigations with taste.
Action Item: revisit our collective technical wisdom about “intuition-based” conjectures/hypothesis and work collaboratively with AI to crack / advance those ideas!
More after the exciting all-hands internal workshop on how we use AI in the USRA RIACS Quantum Team next week! Https://riacs.usra.edu @USRAedu @NASA
Rare opportunity to be considered for USRA quantum team. Very flexible towards a variety of profiles (junior to senior). Work to be done in USA (but UK possible). Topics: AI Reasoning, QEC/FT Architectures, Quantum Algos, Quantum Thermodynamics. Likely NASA affiliation.
workforcenow.adp.com/mascsr/…
I am hosting the Quantum + Space Workshop at Quantum World Congress tomorrow! Super lineup. Join! linkedin.com/posts/davideven…
Davide Venturelli retweeted
Boeing, Axiom Space, Vescent, Qrypt, USRA join Infleqtion, Accenture, and Nebula Space Enterprises in DOE’s Quantum in Space MOU! Learn more: bit.ly/3PExDGP
Looks like I am the top expert in "AI infrastructure optimization" according to AI itself !!!! 🤓 LOL
HPE, qolab, 1QBit, UCSB, Synopsis, AMAT, USRA/NASA, FNAL punching hard in this latest position paper on quantum supercomputing! community.hpe.com/t5/blogs/b…
The eventual integration of quantum technologies into the IT/HPC infrastructure of the world is inevitable. Whatever the future holds, today there is a new force bringing stakeholders together to develop and deploy solutions that are
inherently modular and transparently engineered. Good luck to TreQ, the quantum engineering company!
TreQ has secured $5 million in funding to develop open-architecture quantum computers, with support from @LavrockVC , @CreatorFund , @firstminutecap and @GreenSandsE 👏
thequantuminsider.com/2024/1…
#Quantum #Ecosystem #CreatorFund #QuantumComputing #Funding
Happy to advertise the great work led by our @phase_filip in collaboration with @UDelaware (Safro group) and @rigetti - to be presented in a few weeks at the IEEE High Performance Extreme Computing Conference: A Multilevel Approach For Solving Large-Scale QUBO Problems With Noisy Hybrid Quantum Approximate Optimization arxiv.org/pdf/2408.07793 - truly one of the largest scale experimental demo of hybrid quantum computing with remarkable performance and outlook for improvement.
The paper creates a new framework that can be profitably executed leveraging Ising machines of arbitrary size to solve problems of arbitrary size. We have some numerical evidence that for soc-epinions graphs (26k+ nodes MaxCut) if subproblems can work on thousands of variables then the decomposition method beats the state-of-art. We also have initial experimental evidence that quantum solvers as implemented by us are qualitatively different and can potentially allow faster convergence in some "oracle" terms.
Lots to built upon. It is painful to work in the era of utility scale "augmented NISQ" we live in - but we are up to the challenge because lessons from theory cannot be the only guidance. Lots of QPU hours to burn. Lot of noise to tame. No pain no gain. Even if reliable advantage is not behind the corner, with every sweat drop we learn something valuable. We are far past running toy problems.
More discussions on these general considerations at the IEEE QCE24 "The Next NISQ Phase" Panel: qce.quantum.ieee.org/2024/pr…
If you think Ising solvers do not intersect this next wave of gen-AI think again! In our recent preprint arxiv.org/abs/2407.00071 with Icosa and HPE we look at the statistics from LLM trying to think, exploiting it to define a combinatorial cost function that if optimized could lead to a powerful chain-of-thought. Intriguing!
Finally out the review paper summarizing the scientific output of the NASA Quantum AI Lab for the last 4 years… sciencedirect.com/science/ar… - the easiest way to answer the question “what is NASA doing in quantum computing” :) we should do this every 2-3 years going forward (see also papers in 2017
sciencedirect.com/science/ar… and 2019 books.google.com/books?hl=en…)
Davide Venturelli retweeted
Improving Quantum Approximate Optimization by Noise-Directed Adaptive Remapping
scirate.com/arxiv/2404.01412
Say you want to ≈solve a diagonal Hamiltonian but only have access to a quantum device that is fairly noisy (like, well, everyone who has access to quantum devices) 1/n
What you need is to encode your Hamiltonian evolution in dual-rail subspaces (doubling the number of qubits), average over N orderings of the encoding and then correct the measured estimators with a prefactor. The end result beats symmetry-verification discarding and probabilistic error cancellation. 2/3
But it is more general than that! You don’t need dual rail (any k-hot work), you don’t need qubits (qudits work) and the theory is formulated very generally for Lindblads in analogy to decoherent free subspaces. Work sponsored by @sqmscenter 3/3
In our recent paper “Uniformly Decaying Subspaces for Error Mitigated Quantum Computation” led by @NishchaySuri we present a nice way to correct noisy expectation values of observables quantum or digital simulation in the presence of inhomogeneous dissipative noise arxiv.org/abs/2403.00163 1/3
