Recently Nihon Keizai Shimbun electronic version reported
the following:
「超計算」人類の手中に グーグル実証か
科学&新技術
2019/10/18 18:00日本経済新聞 電子版
グーグルが開発したとみられる新しい量子チップ
人工知能（AI）などに続く革新的技術として期待される量子コンピューターが「スーパーコンピューターを超える日」が近づいてきた。米グーグルは、理論上の概念だった性能を実証し、最先端のスパコンで1万年かかる問題を瞬時に解く実験に成功したもようだ。米IBMなども研究に力を入れる。急速な進歩はいずれ人類にこれまでにない計算パワーをもたらす。AIの活用や金融市場のリスク予測などを通じ、社会にディスラプション（創造的破壊）を起こす可能性を秘める。
グーグルが「量子超越」を達成したもようだ――。英フィナンシャル・タイムズは9月、こう報じた。日本経済新聞が入手した資料によると、最先端のスパコンでおよそ1万年かかる計算問題を、同社の量子コンピューターが3分20秒で解いたという。
量子超越は、従来のコンピューターでは困難な計算問題を量子コンピューターが解く性能を指す。理論上はスパコンを上回るとされた計算性能を、グーグルは世界で初めて実証したとみられる。同社は「コメントできない」としているが、事実なら「教科書に載るレベル。歴史に残る成果」（科学技術振興機構の嶋田義皓フェロー）だ。近く正式発表するもようだ。
量子コンピューターは「量子力学」という物理法則に従って動く。従来のコンピューターは「0」か「1」で情報を表すが、量子力学の世界は「0であり、かつ1でもある」という特殊な状態が起こりえる。
この仕組みを利用した「量子ビット」と呼ぶ計算単位を使うことで、膨大な情報もまとめて処理できる。計算の回数が大幅に減り、時間が劇的に短くなる。グーグルは今回、53個の量子ビットを実現し、乱数をつくる計算でスパコン超えの性能を実証したようだ。
グーグルなどが量子コンピューターの研究に乗り出したのは、半導体の微細加工による従来のコンピューターの性能向上に限界が見え始めたためだ。AIなどの登場を受け、膨大なデータを扱えるコンピューターが求められている。
50～100量子ビットに到達し、開発は「NISQ」と呼ぶ中規模の量子コンピューターに移りつつある。まだ幅広い計算に使えるわけではないが、経済や産業、社会を変えると期待が膨らむ。
計算能力が足りないために、解決しない難題は多い。例えば都市部の渋滞解消。現在は無数の車がそれぞれの都合で走り、渋滞を招く。1台ずつが進む道を短時間に計算するのは困難だ。量子コンピューターを使い、車ごとに「渋滞を起こさない最適ルート」を指示できれば解消に役立つ。
AIによる画像や言語などの処理も短時間、省エネになる。計算力を生かし、個人の体質に合わせて薬を作り分けるような新たな医療の誕生も後押しできる。
量子コンピューターの「使い道」の開拓に力を入れるのがIBMだ。16年に量子コンピューターを外部の利用者にクラウド経由で公開した。世界で15万人を超す登録利用者のほか独ダイムラーや米JPモルガン・チェースなど80近い企業などと研究を進める。
日本では慶応義塾大学に連携拠点があり、銀行や化学大手が参加する。画期的な薬や材料の開発、金融市場のリスク予測などの研究が熱を帯びる。
ただし、量子コンピューターがもたらすのは「光」だけではない。革新的技術は時に脅威となる。ささやかれるのが、ネット社会が根底から揺らぐリスクだ。
現在は通信の際にパスワードなどの情報を暗号化している。最新のスパコンでも解読に時間がかかることから「安全」とみなす。量子コンピューターはこの暗号を破る恐れがある。新しい暗号技術の検討も進む。
IBMのメインフレーム（汎用機）の発売は1964年。従来のコンピューターもその前に20年ほどの黎明（れいめい）期があった。日本IBMの森本典繁執行役員は「量子コンピューターもそういうフェーズにある」と指摘する。
コンピューターの歴史で、およそ70年ぶりに起き始めた革新の動き。本格的な量子コンピューターの実用化には課題が多いが、米インテルや中国のアリババ集団なども開発に参入し、今後もブレークスルーが生まれる見通しだ。
Translation
New quantum chip that seemed to have been developed by
Google
The quantum computer, which was expected to be an innovative
technology following artificial intelligence (AI), was approaching the “day
beyond the supercomputer”. Google seemed to have succeeded in demonstrating a
performance that was a theoretical concept, and succeeded in instantly solving
a 10,000-year problem with a state-of-the-art supercomputer. The US IBM and
others were also focusing on research. Rapid progress would eventually bring
unprecedented computational power to mankind. Through the use of AI and
financial market risk predictions, it had the potential to cause disruption
(creative destruction) in society.
It seemed that Google had achieved “Quantum Transcendence”.
The UK Financial Times reported in September: according to materials obtained
by the Nihon Keizai Shimbun, the company's quantum computer solved a
computational problem in 3 minutes and 20 seconds that took about 10,000 years
with a state-of-the-art supercomputer.
Quantum transcendence referred to the ability of quantum
computers to solve computational problems that were difficult with conventional
computers. It seemed that Google had demonstrated the world's first computing
performance that was theoretically superior to supercomputers. The company said
it had “no comment”; but if it was true, it was "a level that can be recorded
in textbooks. A record that stands in history" (Yoshiaki Shimada, fellow
of the Japan Science and Technology Agency). It seemed that it would be
officially announced soon.
Quantum computers operated according to the physical law of
“quantum mechanics”. Conventional computers represent information by “0” or
“1”, but in the quantum mechanics world “it's 0 and it's also 1” special states
could occur.
By using a calculation unit called “quantum bit” using this
mechanism, it was possible to process a large amount of information
collectively. The number of calculations would be greatly reduced and the time might
dramatically shorten. This time, Google realized 53 qubits, and it seemed that
it had demonstrated the performance exceeding a supercomputer in the
calculation that made random numbers.
The reason why Google and others had started research on quantum
computers was that the limits to improving the performance of conventional
computers by micro-fabrication of semiconductors had begun to appear. With the
advent of AI and the like, computers that could handle enormous amounts of data
were required.
Reaching 50-100 qubits, development was moving to a
medium-sized quantum computer called "NISQ". It was not yet usable
for a wide range of calculations, but expectations increased when the economy,
industry and society changed.
There were many unsolvable problems due to lack of computing
power. For example, eliminating congestion in urban areas. Innumerable cars now
ran for their own convenience, causing traffic jams. It was difficult to
calculate in a short time the path that each vehicle took. If you could use a
quantum computer to indicate the "optimal route that does not cause
traffic jams" for each car, it would help.
Processing of images and languages by AI might also save
energy and time. Utilizing computational power, it might be possible to support
the birth of new medical treatments that made medicines according to the
individual's body feature.
IBM was committed to developing “ways of utilization” for
quantum computers. In 2016 it released quantum computers to external users via
the cloud. In addition to over 150,000 registered users around the world,
research would be conducted with nearly 80 companies such as Daimler in Germany
and JP Morgan Chase in the US.
In Japan, Keio University had a cooperation base, and banks
and major chemical companies had participated. Research on innovative drug and
material development, financial market risk prediction, etc. was in an exciting
and passionate atmosphere.
However, quantum computers did not only bring “light”.
Innovative technology could be a threat sometimes. What was whispered was the
risk that the Internet society would be shaken from the ground up.
Currently, information such as passwords was encrypted
during communication. Even the latest supercomputers were considered
"safe" because it took time to decipher. Quantum computers could
break this code. The study of new encryption technology could also proceed.
IBM's mainframe (general-purpose machine) was released in
1964. Conventional computers also had a dawning period of about 20 years before
that. Norishige Morimoto, executive officer of IBM Japan, pointed out that
"quantum computers are in that phase".
In the history of computers, after about 70 years, innovative
movement began to occur for the first time. Although there were many issues in
putting a full-scale quantum computer into practical use, the US Intel and the
Alibaba group in China were also entering the development, and breakthroughs were
expected to continue.
So, it
seems that a revolutionary change in the field of computer is happening.
沒有留言:
張貼留言