OpenAI is not investing into Quantum Computing! #1 Bigger Picture by Quantum Leonardo
Sketch-style representation of a quantum AI processor, reminiscent of Leonardo da Vinci's drawing technique.

OpenAI is not investing into Quantum Computing! #1 Bigger Picture by Quantum Leonardo

Get excited about quantum technology, but don't go overboard!

With a 30% to 50% decline in investments in quantum technology—varying based on the source and how quantum is categorized—I've noticed an intensified effort to maintain enthusiasm for quantum technology. This optimism needs to last long enough for quantum tech companies, especially those focusing on quantum computing hardware, to fulfill their promises. According to various roadmaps from leading companies like Pasqal , QuEra Computing Inc. , IBM , Quantinuum , and PsiQuantum , early but commercially viable versions of this hardware could emerge between 2027 (optimistically) and 2030 (realistically), with a more pessimistic projection extending to 2033. I strongly agree with promoting quantum tech. I believe we should persuade others that quantum technology, built upon a century of research, uniquely merges advancements in engineering, mathematics, physics, and computer science and when it finally arrives, it will be a rock solid source of value. However, I also think that excessively trying to divert public attention to quantum might not be beneficial and hurt its image.

Last week, I heard rumours of OpenAI going into quantum. These are mainly propagated by these two articles: Is OpenAI Opening Up To Quantum?, From quantum AI to photonics, what OpenAI’s latest hire tells us about its future.

"With a new hire, subtle announcements and an obvious need, experts are wondering if OpenAI is investigating the use of quantum to power its AI empire." The Quantum Insider

Quantum, it seems, has turned into a bit of a diluted currency, used more to grab attention or lure in investors. Yet, I've encountered a totally different viewpoint too: some say if a company's main focus is quantum, they'd rather not invest in it. It's quite the conundrum, which is why stepping back and adopting a broader perspective on quantum is essential. In this piece, I'm zeroing in solely on the computing aspect, leaving the wide and wild world of Quantum AI for another day. Btw, Sam Altman, CEO of OpenAI in the recent Lex Fridman interview says that "Compute is going to be the currency of the future".

What's quantum and what's definitely NOT?! - bigger picture

Bear with me as I will not make it easy for you - let me confuse first a little bit more. The building block of modern computing, the transistor, necessitates understanding the laws of quantum mechanics (semiconductor physics). The understanding, modeling, and construction of transistors are inextricably linked to quantum mechanical calculations. Yet, we refer to today's computing power as "classical." This terminology, arguably, does not serve its purpose effectively. It likely originates from the lexicon of physics, where a distinction is made between classical and quantum physics. The term has gained traction and is widely used, suggesting we must accept it.

Now when you have learned that classical computing is built on quantum as well, let's try to disentangle this mess. Examples speak more than thousands words, so let me give my perspective. I would propose 3 categories that we have to keep in mind when we talk about quantum AND/FOR/IN computing.

  • Processors powered by (quantum) physics. Extropic led by Guillaume Verdon and Trevor McCourt assembles a new kind of processor (neither CPU, GPU nor QPU). Extropic is addressing the limitations of traditional digital computing, particularly as Moore's law slows, by developing hardware that operates efficiently in intrinsically noisy environments, mimicking the efficient, probabilistic computation seen in biological systems. They are leveraging Energy-Based Models (EBMs) to design stochastic analog circuits that can perform sampling from complex landscapes far more efficiently in terms of runtime and energy than current digital computers. Extropic's innovations include superconducting chips for high-efficiency, low-temperature operations and semiconductor devices for room temperature applications, aiming to drastically improve the performance of AI algorithms by utilizing hardware that simulates probabilistic, thermodynamic processes. Lightmatter led by Nicholas Harris is advancing photonic technologies for chips, offering hardware and software solutions that aim to reduce power consumption while boosting performance, crucial for compute-intensive AI workloads. Lightmatter plans to develop technologies that enhance AI model scalability and efficiency, addressing the demand for innovative compute and chip communication solutions.The continued scaling of computational power, aligned with current trajectories, necessitates the development of such advanced processors. If one of the proposed approaches (the two mentioned do not exhaust the list) succeeds, we may see the rise of a new tech hub called Thermodynamic Valley or Photonic Valley.
  • Solvers powered by (quantum) physics. LightSolver aims to solve NP-hard business optimization problems with the use of their product LPUs (Light Processing Units). For example, LightSolver optimized field service technician scheduling for a US telecommunications provider using a QUBO formulation, demonstrating better performance over the Gurobi solver. LightSolver's platform performed really well in generating optimal solutions rapidly, significantly reducing operational costs by optimizing routes more effectively than traditional solvers, showcasing its potential in real-time operational optimization across various industries. It is quite rare for a company to compare their optimization solver to Gurobi Optimization which is a very solid and hard to beat industry standard, hence, this demo looks impressive. Dirac 3 offered by Quantum Computing Inc. serves the same goal - to solve complex optimization problems. They have many interesting case studies as well and very flexible cloud/hybrid/on-premise offerings. Dirac 3 as stated in the technical specification is Hybrid Analog Machine with Quantum Optics and Digital Electronics. Term Entropy Quantum Computing is used as well.
  • Quantum computers (QPUs - Quantum Processing Units). These are the processors being built by tech giants 谷歌 , Amazon Web Services (AWS) , 微软 , IBM and specialized start-ups distributed across the globe (IonQ, PsiQuantum, Quantinuum, IQM, Xanadu, Pasqal, QuEra, Diraq and so on and so on). I will make a separate blog post about features of quantum computers that make them quantum but, for now, we can say that these are the ones that can execute quantum algorithms such as Shor's algorithm (breaking RSA), Grover's algorithm (search), Quantum Fourier Transform, etc. that cannot be simulated on classical computers within the same computational complexity. (Keep in mind, though, that for Shor's algorithm, there is no proof that classical computers cannot execute it efficiently - it might be the case that we have not found a suitable algorithm yet.) There is also the issue of universality to be discussed but let's also leave it for another time.

As you can see, it is very hard to draw the lines between categories and definitely categorize each of the companies. The terminologies used by different companies are not consistent with each other.

Sabine Hossenfelder made a great video about D-Wave's quantum computer and whether we should call it a quantum computer. The analogy she used humorously suggests that for understanding a spoon's metal properties, the spoon itself, observed under a microscope, is the most effective "quantum computer." However, you cannot use the spoon to compute anything else. While this example might seem totally out of the blue, it builds exactly on the same idea as quantum computers build. The motivation behind quantum computers is to use qubits, which mimic atoms' behavior, for calculating atoms' quantum properties, instead of using traditional bits. And there is nothing better to mimic the spoon than the spoon itself.

So, here's what I'm thinking about OpenAI going into quantum computing. Even if all the buzz turns out to be spot-on, it looks like OpenAI's next big move might be crafting some specialized processors (or solvers but this is less likely). These would be way better suited to handling the heavy lifting for transformers and neural network architectures than our current GPUs. Whether they end up calling it an LPU, an NPU (neural processing unit), or something else entirely, one thing's for sure: we're not talking about a quantum computer here.

We need a clear message as a community

On February 21st, 2024, Apple dropped some pretty exciting news: iMessage is getting a major security upgrade with PQ3, a post-quantum cryptography messaging protocol. Honestly, I think this announcement grabbed more eyeballs than any other quantum tech news lately, and not just from the usual tech-savvy crowd.

I've had so many conversations about this, even with folks who usually don't pay much attention to quantum. But one thing felt out of place - people were buzzing about how this is a game-changer for the future of quantum computing. But while it's great hearing that companies are gearing up for the so-called Q-day, ensuring our messages stay safe from the "harvest now, decrypt later" threat, this update, let's face it, isn't really about quantum computer advancements, neither in hardware nor software. It's all about bringing the latest in cryptographic tech into our everyday communication, and that's pretty awesome in its own right.

I've got to say, I'm probably one of the biggest quantum enthusiasts you'll meet. I'm all in on quantum, to the point where I'd happily wire up all the qubits myself if that's what it takes. But we've got to be real about something. We need to clear the air and get our message straight. Putting a "Q" in front of every term (like OpenQAI) isn't doing quantum any favors. Honestly, it risks making the whole field seem less credible, and that's the last thing any of us want.

I live in the San Francisco Bay Are where both OpenAI and PsiQuantum mentioned in the article are located. The topic of AI seems to be more prevalent in discussions here compared to quantum computing. Let's change it together with reliable and clear communication!

#quantum #quantumtech #quantumcomputing #ai #llm #qubits


Disclaimer: The views expressed in this article are solely those of the author and do not reflect the opinions or positions of any entity with which the author is affiliated or employed. The author's statements should be considered personal opinions and are not representative of the official stance or policy of any organization associated with the author.


Jamie Gill

disrupting manufacturing | AMA Quantum Computing | reducing risk/fear (-$) and extracting value (+$) through incisive mathematics

7 个月

Great work on your new blog, Michael Baczyk. I took note you're in the SF bay area... I'm just about to leave in a few minutes after a great last couple of days at Simons. Next time, I'll check to see if you're around... Excellent post putting this highly nuanced field into context for a broad audience. ?? I suspect you'll be able to accomplish this over and over again... Thanks for your insights and looking forward to more to come!

I really enjoy your informative posts, Michael! Having read your article above, here are two points I'd like you to consider: 1. Are you being unduly pessimistic about the funding picture? This article suggests that it's only private venture capital that has taken a dip (and even that mostly in the US); public funding has not. Indeed, many countries seem to be upping their game. https://www.forbes.com/sites/trevorclawson/2024/02/09/changed-times-why-europes-quantum-startups-need-a-path-to-profit/ 2. Even if Open AI has only hired one Quantum engineer...or otherwise just dipped a tiny toe in all things Quantum, don't you think that it makes total sense for them to invest more? And if so, don't you think it's just a matter of time before they do? Personally, I'm only too happy not to have the QT field cornered by the tech bros. But I'd be surprised if they didn't make any significant forays into the field at some point, and probably sooner rather than later.

Rohit Narayan

Director-Digital (Quantum Computing & AI enthusiast)

7 个月

Wondering if Quantum Inspired Algorithms will play a useful role in the interim. Given we can classically simulate upto 40+ quibits, is there merit in focussing on quantum inspired algos for useful solutions? This also provides a good runway for quantum algo developers whilst the fault tolerant quantum hardware is still a few years away. Can this lead to innovation in programming LLM's which may lead to lesser parameters for same performance? It will be a shame if we lose the momentum on quantum given the amount of investment that has flown in this sector over the last 5-7 years!

Joe Spencer

Quantum Technology Specialist | Computing | Sensing | Defence | Space

7 个月

Steve Suarez ? Based on our chat today about OpenAI

Yang Chen

Llama-on-premise/SYCL/Linux Kernel/Infrastructure

7 个月

NVIDIA started hiring quantum engineers this month in the US, while the quantum HW company is building compilers and libraries. quite similiar to 10 years ago when I was interviewing with Intel which asked me about GPU compiler and HPC background in GPGPU.

要查看或添加评论,请登录

社区洞察

其他会员也浏览了