Home page
Quantum computer is useless
(Fig.1) Today's noisy (= error-prone ) quantum computers cannot outperform noiseless (= errorless ) quantum simulators (= classical computers ). But the media lied.
Despite today's useless hopeless quantum computers that cannot factor even the simplest 21, the mainstream media spread fake news saying HSBC-IBM ( error-prone ) quantum computer showed advantage in predicting some bond trading.
↑ This fishy HSBC-IBM quantum computer's advantage is just overhyped fake news ( this-3rd~4th-paragraphs ).
In this HSBC's fake quantum advantage experiment, they used IBM's error-prone 109-qubit quantum computer to cause random meaningless noise and errors in the original bond-trading datasets.
These trading datasets were given to a classical computer-machine-learning estimator to predict some bond-trading values called fill probability empirically.
That's all. No quantum computation nor quantum advantage nor quantum mechanical prediction.
This site summarizes this research ↓
2nd-paragraph says -- No quantum computation
"HSBC recently published a paper claiming improved bond trading predictions using a 100-qubit quantum computer, but what they actually did was much simpler—and not quantum computing in the usual sense. They took classical trading data, encoded it into a quantum circuit on IBM hardware, measured the circuit’s outputs, and used those measured values as extra input features for a classical machine-learning model running on ordinary computers. The quantum device was never used to run an algorithm, optimize parameters, or perform any kind of end-to-end computation; it just generated a nonlinear, noisy transformation of the data."
3rd-paragraph says -- No quantum advantage
"it's not evidence of quantum advantage or practical quantum computing, just a creative way to use a noisy device (= error-prone quantum computer ) as a fancy random number generator."
Errors and noise randomly generated by the error-prone IBM quantum computer accidentally gave slightly better results than a noiseless (= errorless ) ideal quantum simulator run on a classical computer.
↑ The ideal noiseless (= errorless ) quantum simulators (= just classical computers ) should always give more accurate and better results than today's error-prone (= noisy ) quantum computers in the same tasks.
So these seemingly-better results by datasets modified by the error-prone quantum computer are just accidental and irrelevant to quantum computer's advantage or quantum computation.
This-site debunking this HSBC research ↓
The 1st~2nd-paragraphs say -- No quantum computer advantage
"headlines worldwide: “Quantum computers beat classical models in real-world finance."
"To cut to the chase, they did not"
6th-paragraph says -- Quantum errors' artifact
"transforming classical market data via a quantum algorithm, then feeding that into classical machine learning models. Compared to common approaches, this resulted in better fill predictions: 'but only when the data came from real, noisy (= error-prone ) quantum hardware', reads the abstract. When the same quantum transformation, it continued, was run on a noiseless simulator (i.e., an idealised quantum system run on a classical computer ), the advantage vanished"
"If quantum advantage disappears when the quantum computer works perfectly (= in the noiseless or errorless classical computer's simulator ) and only appears when it does not, that is not an advantage. That’s an artifact."
↑ The noisy (= error-prone ) quantum computer giving better results than the ideal noiseless (= errorless ) quantum simulator (= run on a classical computer ) means these seemingly-better results were just accidental artifact irrelevant to quantum computer's advantage.
This-4~5th-paragraphs say -- No quantum advantage
"Yet HSBC conceded in its paper that the observed effect was purely empirical, caused by 'inherent noise' in the process with no theoretical foundation. That means the much-touted result is most likely a cherry-picked example of selection bias"
This-4th-paragraph says -- Ideal errorless quantum simulator
"Simulating an ideal quantum computer, with no noise and no errors."
This-p.4-right-A says -- Error-prone noisy quantum computer
"With agreement found using a quantum simulator in
the idealized noiseless case, we turn our attention to simulation of the circuit on IBM's noisy simulator with a
noise profile mimicking that of IBM quantum devices.
We utilize a noisy simulator to understand how our approach would behave in a realistic noisy intermediate
scale (NISQ) quantum computer, which is affected by
multiple sources of noise from the environment and is
prone to errors during computations4"
In spite of this fact, the media, HSBC (= company ) and IBM falsely claimed their (= error-prone ) quantum computer could predict real-world bond trading values.
↑ This research paper ( this-last-link ) on fake quantum advantage ↓
p.2-Figure 1. shows -- Classical machine-learning
Empirical bond trading data was sent into a (error-prone noisy ) quantum computer, and then used as datasets in a classical computer's machine-learning called fill probability estimator.
p.2-right-2nd-paragraph says -- Quantum computer modified data
"current market
information, pass them through a particular quantum algorithm
executed on a quantum computer to generate a corresponding
set of transformed trade event vectors, then use those instead
to train a machine learning model for estimating fill probabilities, and finally, compare with actually realized outcomes. "
p.5-right-2nd-last-paragraph says -- Just empirical, No theory
"Therefore, we are adopting a purely empirical approach
and design our experimental tests using backtesting"
p.7-Figure.4 says -- Quantum computer's random noise
" Here, Unitary represents random single-qubit unitaries to set the initial state of each qubit"
p.8-right-3.2-2nd-paragraph says -- Error-prone quantum computer
"Since today's (quantum computer) devices are still inherently noisy (= error-prone ) and may alter the generated feature outputs, we
also use a noiseless (= errorless ) quantum simulator running on a classical
computer (= superior classical computers ) to compare with"
p.9 says -- still Not a quantum computer
This quantum computer had only 109 qubits (= one qubit takes only 0 or 1 value ), which is useless, far from useful millions of qubits ( this-4th-paragraph ). ← They used deceptive "hybrid (= almost classical )" computer, the useless quantum computer alone cannot calculate anything.
p.9-left-3.2.2. says -- classical computer needed
"Since the
quantum hardware is susceptible to noise (= errors ),.. In this study, we apply two error mitigation methods" ← Classical computers were used to empirically mitigate errors of their useless quantum computers ( this-4th-paragraph, this-p.1-right-2nd-paragraph )" ← so No quantum computer's advantage.
p.14-left-2nd~3rd-paragraphs say -- Errors' accidental artifact
"The model
performance results for respectively generated fill probability... show
significant uplifts of out-of-sample test scores for models with
noisy (= error-prone ) quantum hardware-generated inputs over those with noiseless (= error-less ) quantum simulation or classical input" ← false results
"the results are still subject to many open questions. For instance, it is Not understood how exactly quantum hardware noise affects our particular quantum circuit, and it is unclear how resulting noise-encoded feature vectors may benefit the analysis of noisy financial observables. This requires further investigation." ← No quantum advantage
So this HSBC research's noisy (= error-prone ) quantum computer hardware accidentally happened to make better prediction due to unpredictable errors (= No real quantum advantage ) or the artificial illegitimate error-mitigation trick conducted by a classical computer's post-processing ( this-p.1-right-2nd-paragraph ).
As a result, No quantum computer advantage in bond trading, contrary to overhyped fake news just trying to make today's error-prone hopeless quantum computers look "promising".
Feel free to link to this site.