The most pressing issue right now is referred to by them as "entanglent truncation under high-dinsional modalities."
This is also the reason why Qiao Yu has been keeping a close eye on Zheng Zefeng. There’s not much to worry about when only running classical computing tasks.
But once quantum computing tasks are involved, problems such as phase distortion, broken correlations, and cumulative error exponentiation arise.
This is clearly unacceptable to Qiao Yu.
His theory is definitely sound, otherwise the quantum simulation computer could not have been developed.
...
On the way back, Qiao Yu called Zheng Zefeng, Zhang Xiaoli, Luo Qixun, and Hu Xingyao, the three professors, into a car.
Three years ago, core team mbers were transferred to Yanbei University after achieving so results, each bearing the title of high-level talent.
Subsequently, Yanbei International Mathematics Research Center even prepared two floors specifically for the team as a regular and office location.
Since the two projects have been on track, besides internal etings, representatives from the labs often co here for etings.
Most of these are technical etings. Although each eting isn’t long, they occur quite frequently.
Mainly because the two projects are interfacing with many different research units: academic institutions, research bodies, corporate units. Each mber of the team averages interfacing with about a dozen or twenty units.
The daily busyness is imaginable, and everyone frequently goes on business trips.
Theoretically sound things always present strange problems when implented in the labs. So can be solved remotely, but those that can’t are discussed during etings, or the related person travels onsite to handle them.
So technically speaking, Qiao Yu wasn’t wrong. Considering the busyness over the past few years, Qiao Yu really doesn’t match up to these professors. But relatively speaking, his work is definitely more important.
After all, all issues ultimately converge to him during the weekly group etings.
Though he doesn’t need to delve into the details, he must provide a major direction for resolution.
Today aligns with such a situation.
Qiao Yu gathered a professor specialized in mathematical theory, one in quantum theory, and another in algorithm engineering, essentially a small eting.
"The situation was already ntioned in the eting earlier, it’s not that I’m anxious to do sothing, the key is that the Quantum Simulation Alliance is definitely in a hurry. Over the past several years, more than forty companies have collectively invested over thirty billion.
Now our Sky Shield deterrence project phase one has already passed acceptance. The quantum simulation computer still not being market-ready is indeed hard to justify.
So my idea is, we must seize the ti in the coming months to tackle technological problems and strive to resolve issues regarding algorithms. The National Day tribute was successful, and we aim to offer another gift before New Year’s Day.
So let’s discuss properly today, strive to unravel the current issues, so that you all can proceed smoothly. Professor Wang, why don’t you start by sharing recent feedback from the lab?"
Sitting in the spacious business vehicle, Qiao Yu directly comnced the small eting.
While other team mbers can just focus on optimization, their work is crucial.
The current engineering prototype, if rely treated as a standard computer, is roughly akin to a six to seven thousand yuan mid-to-low end personal classical computer.
Watching videos, playing gas, all are possible.
But a computer priced at fifty thousand yuan naturally should see those additional functionalities utilized.
Frequent errors while using functions exclusive to quantum simulations are naturally unacceptable.
Having collaborated for more than four years, Wang Xingyao is accustod to Qiao Yu’s style, and having been called upon, he promptly began his report.
"Based on the updated algorithm, yesterday’s latest simulation on a 100 Quantum Bit Hubbard Model base state energy indicated that when the modal dinsion base point is below 30, the compression rate can reach 80%, the entanglent fidelity is at 99.2371%, and the theoretical computation error is below 0.5%.
When the modal base point is between 30 to 70, the compression rate is 67%, the entanglent fidelity is 87.637%, and the computation error ranges from 8% to 12%.
When the modal base point exceeds 70, the compression rate only remains at 52%, while the entanglent fidelity drops to below 43%, with an error value close to fifty percent.
Conclusively, the updated algorithm enhances low quantum modal operations by at least 30% compared to the previous iteration, but when transitioning to high modal operations, the gap isn’t significant from the previous algorithm."
After the report, Wang Xingyao glanced at Zheng Zefeng.
The new version can be said to be a joint effort by the two, although results didn’t et expectations, they both tried their best, hence now Old Zheng spoke up.
It must be ntioned, Wang Xingyao doesn’t enjoy these impromptu mini-etings.
Being too close to Qiao Yu, the pressure is a bit high.
Even though Qiao Yu remains young, those eyes staring at him always give a creepy feeling.
"Ahem, so, let say a few words. I’ve looked through the experintal report too, and have conducted an analysis. I think the major issue might be that the compressed tensor network structure doesn’t sufficiently align with the current Q-RAM fractal storage physical addressing mode, causing access conflicts.
Since I’ve been occupied with the Sky Shield exercise, I haven’t had a chance to bring up this issue. Hence, I’m considering discussing it with engineers from Hisilicon, Changjiang, and Changxin, still needing a simultaneous update with Q-mFlex.
Especially concerning the fractal 3D stacked resistive mory, I think there are still so fundantal design issues in the present Q-RAM Core, the support for dynamic mapping of modal paraters isn’t sufficient."
Zheng Zefeng understood Wang Xingyao’s gaze, proactively taking over the conversation.
Actually, it’s not that he wanted to step up but couldn’t help it, he has even more pressure.
After all, he’s the only number theory expert in the group directly liaising with Qiao Yu. Whenever anything goes wrong, Qiao Yu invariably thinks of him first.
"Co on, Old Zheng, we’re not supposed to shirk responsibility! You can’t just change hardware design willy-nilly, chip designers have a tough job!
It’s easy for you to say, but altering Q-mFlex’s total four layers ans if Q-RAM Core is moved, the other three layers require redesign!
Attempting hardware redesign now, forget about three months, I doubt whether it can be completed by next New Year’s Day. Hence, let’s not consider changing the mory architecture now.
If a change is necessary, let’s wait for the second generation. Right now my demand is solving the issue via core algorithm design. We are indeed conducting research, but we can’t entirely approach problems from a pure research perspective.
We also need to consider the industry’s viewpoint. Now altering hardware is too costly. To demolish everything and start anew requires another rerun validation, are you paying for it?"
Zheng Zefeng gave an awkward smile.
Though he isn’t short of money now, researching chips is a different story...
Maintaining a chip research engineer’s salary, various benefits, insurance, provident fund, and so on, averages out to at least six hundred thousand per person for enterprises.
Especially in researching simulation quantum chips, finding suitable personnel is challenging. Key firms responsible for mory design and manufacturing account for at least four to five thousand engineers.
Simply maintaining manpower for a year costs close to three billion...
He surely can’t afford it.
"Let’s quickly roll out the first generation product, allowing upstream and downstream enterprises to recover funds, while simultaneously showing foreign tech firms a taste.
You must realize the quantum simulation computer project complents the Sky Shield exercise. The Sky Shield exercise is to show where the safest place in the world is to foreign capital.
The quantum simulation computer symbolizes the current world’s most advanced productivity, letting them know where is most worthy of value investnt.
The two complent each other! So let’s not change the hardware, if we can resolve 90% of the issues, it’s a victory! This is the ti for us to stand up!
Actually, I’ve previously reviewed the analytical report from the lab seriously and ford an idea. By inserting a virtual mapping layer between the Q-RAM Core’s physical fractal address and the compressed tensor logical address.
For instance, utilizing EU_xyz coordinates, and dynamically adjusting the mapping relationship via the quantum state topology perception algorithm. Simultaneously partition the physical Q-RAM Banks asymtrically, allocating frequently accessed compressed tensor headers to low-conflict Bank groups. What do you think?"
Reviews
All reviews (0)