(Fig.1) Today's quantum computers cannot correct errors. → Reject error results postselectively and illegitimately without error correction = still error-prone, useless.
The 1st, 3-5th paragraphs of this hyped news (11/24/2024) says
"Microsoft and Atom Computing,... creating and entangling 24 logical qubits on a commercial quantum machine." ← Just 24 bitstring (= one logical qubit can take only 0 or 1 state, so still Not a computer. "commercial" is just misleading hype, this 1st,4th-paragraphs )
" These logical qubits were entangled in a cat state, also known as a Greenberger-Horne-Zeilinger (GHZ) state (= where all 24 bitstring are just made to be simply 0 or 1 states )"
"losses of neutral-atom qubits... error rate of 10.2%, which is 4.1 times better than the baseline physical error rate of 42%." ← hopelessly too bad error rates
But that’s not all – when errors were detected and ( atoms' ) losses were detected and corrected, the error rate improved to 26.6%," ← error rates of just 24 qubits were 10 ~ 42 %, which is too bad, completely useless ( this 6~9th-paragraphs ).
↑ Their neutral atoms (= each atomic energy states are used as each qubit's 0 or 1 ) unstably trapped in laser light are fragile, too slow, impractical, too easily lost to be a practical quantum computer.
This site's 3rd, 14th paragraphs say
"At present, errors make it difficult to perform even the simplest computations" ← Today's error-prone quantum computers cannot conduct even simple calculations.
"with each qubit being a single atom, there was an appreciable error rate. By detecting errors and discarding those calculations where they occurred (= cheating, Not error correction ),.. this doesn't eliminate errors" ← Their error correction failed.
↑ This research paper about Microsoft atomic qubits ↓
p.2-Fig.1-lower says "In one case (middle = error rate 26.6% ), we reject trials with a detected error, but attempt to correct for lost qubits (= just add new atomic qubits for lost atoms without correcting qubit state's errors ). In the other case (right = error rate 10.2% ), we reject (= discard ) trials with any lost qubit or detected error (= just ignore results with lost atoms and errors without error correction ). "
p.8-Table III shows just making 24-bitstring all 0 or 1 states (= very simple task ) caused more than 90% error rate or atomic loss , which were illegitimately rejected without error correction (= accept rate was only 8.4% ~ ). ← still error rates were impractically high (= 12 ~ 26 % ) even after rejecting (= ignoring ) erroneous qubits illegitimately.
p.9-Fig.10~p.10-Table IV shows just conducting n = 27 CNOT two-qubit operations in just 28 logical qubits (= Bernstein-Vazirani algorithm ) caused a lot of errors, out of which about 90% results were discarded as irreparable errors (= accept rate was only 8.6 ~ 17.5 % in n = 27 ), but error rates were still too high = 22 ~ 31%
↑ As a results, today's quantum computers are completely useless, too error-prone, unable to correct errors, far from a practical quantum computer that will need millions of qubits (= this overhyped Microsoft used only impractical 256 atomic qubits divided into just 24 logical qubits = just 24 bitstring, this 6th-paragraph ).
This 2nd-last paragraph mildly says
"In the end, the current hardware has an error rate.. not good enough that we can perform the sort of complex operations that would lead quantum computers to have an advantage in useful calculations" ← quantum computers are useless, error-prone, No quantum advantage after all.
2024/11/24 updated. Feel free to link to this site.