Quantum physics is not for the faint of heart. Some things are super easy to explain (see our Executives Guide to Quantum Computing). But the cutting edge science and advanced engineering taking place in research centers and academia and government can get really deep and complex and is frequently full of terms only used by quantum researchers. I feel so fortunate to have friends and associates in the space who are kind enough to explain things to me. But today’s announcement is happening at a time when breakthroughs in cutting edge AI tools are claiming they can put the power of human reasoning and highly trained LLMs to work for me. I saw this as an opportunity to test two of these latest AI tools, OpenAI’s Deep Research and Grok3’s DeepSearch.

About OpenAI’s Deep Research:

OpenAI’s Deep Research is an AI agent designed to autonomously perform complex, multi-step research tasks by analyzing and synthesizing information from diverse online sources. Powered by the advanced o3 model, it generates comprehensive reports within 5 to 30 minutes, citing relevant data and providing detailed insights. This tool aims to assist professionals in fields such as finance, science, and engineering by streamlining the research process and delivering expert-level analysis efficiently.

About Grok3’s DeepSearch and Think:

Grok3’s DeepSearch is a feature of xAI’s Grok-3 chatbot that enables real-time scanning of the internet and the social media platform X to retrieve pertinent information in response to user queries. This tool is designed to enhance research, brainstorming, and data analysis tasks by providing users with up-to-date information and summaries. DeepSearch aims to offer a more interactive and dynamic user experience by integrating live data into its responses. Grok3’s Think is a specialized mode within the Grok3 AI model that activates advanced reasoning capabilities for complex problem-solving tasks. When enabled, Think mode allows Grok3 to engage in multi-step reasoning processes, carefully analyzing problems, exploring multiple approaches, and self-correcting before delivering a final response. 

Today’s Announcement by Microsoft:

Microsoft has announced the development of Majorana 1, what they call the world’s first quantum processor powered by topological qubits, utilizing a novel class of materials called topoconductors. This breakthrough is claimed to enable the scaling of quantum processors to a million qubits on a single chip, significantly enhancing computational power and stability. Microsoft asserts that this advancement positions them to build a fault-tolerant quantum computer within years, potentially revolutionizing fields such as medicine and materials science by solving complex problems beyond the capabilities of classical computers.

Along with their blog announcement and media blitz they also published a scientific paper in Nature titled Interferometric single-shot parity measurement in InAs–Al hybrid devices. This paper was meant to back up their assertions. This is where the hard science and engineering terminology gets real deep real quick.

This is the paper I decided to put to the test using Grok3 and OpenAI.

I uploaded the paper and reviewer comments to both Grok3 and OpenAI. I gave them both a similar prompt. I asked them to look for inconsistencies of logic.

I paste links to both below for the curious. Both lead me to conclude that Microsoft’s research team has done some real science and some hard engineering here, and are to be commended for the results in the paper published in Nature. But it is impossible to fully substantiate the claims in their blog post. The data they provided backs up the fact that they are measuring things at a level that has never been done before. This kind of measurement is a huge step forward for science and engineering of quantum systems and is a technical milestone. However the paper in Nature appears to be built on circular logic about what is going on and makes many assumptions about what they are really measuring. The paper seems to be overstating results.

It is probably prudent for us to wait for further review by scientists and quantum engineers before we take Microsoft’s blog post at face value, even though we can congratulate them for building a way to measure things that have not been measurable before.

The Grok3 and OpenAI systems did diverge in their conclusions quite a bit. Grok3 was far more believing in the Nature article and Microsoft blog post, even though it also noted inconsistencies. Which raises a big point. None of these AI systems are foolproof themselves.

To read results of the tools see:

• ChatGPT Deep Research Results
• Grok3 DeepSearch and Think Results

I feel like this exercise proved to me that the latest LLM tools can help bring insight to me on complex subjects and can help me learn. But I would much rather be asking a real quantum physicist for thoughts. I believe this conclusion will go for most other domains of knowledge at this point in their evolution. The tools can be great to help with context, but insights from human experts still win out.

Related Resources:

• Executives Guide to Quantum Computing: An explanation of quantum terms for the busy executive
• Quantum Computing and Quantum Security Sensemaking: A reference to key OODA research on all things quantum.
• The Executive’s Guide to Quantum Security: A reference to steps to take now to ensure quantum security including protecting against Harvest Now Decrypt Later attacks.
• Is Quantum Computing Ushering in an Era of No More Secrets?: Context from OODA’s Matt Devost on the very near future of quantum computing.