Man v Mashina: Mumkin Kompyuterlar Kuk, Yozish va bizdan yaxshi bo'yoq?

Man v Machine: Can Computers Cook, Write and Paint Better Than Us?

Endi bir o'yinni g'alaba qozonish mumkin sun'iy aql, yuzingizni, hatto mashinalar chipta shikoyat. Lekin bu narsalar, albatta, mumkin, hatto insonlar qiyin topish?


Guardian.co.uk byNomli Ushbu maqola “Man v mashinasi: kompyuterlar pishirishni mumkin, yozish va bizdan ko'ra yaxshiroq bo'yoq?” Leo Benedictus tomonidan yozilgan, Guardian Shanba kuni 4 iyun uchun 2016 08.00 UTC

One video, Men uchun, o'zgartirilgan hamma narsa. Bu eski Atari o'yin rasmlari bo'ldi Qutilib chiqishga urinmoq; tarqamoq, Agar belkurak chap va o'ng ekranning pastki bo'ylab siljitish bir, Ulardan bir to'p o'ynayotgan tomonidan g'isht halok harakat. Siz o'yin futbolchi haqida o'qib bo'lishi mumkin: tomonidan ishlab chiqilgan bir algoritm Deepmind, Kimning AlphaGo dasturi Britaniya sun'iy aql Kompaniya, shuningdek hech futbolchilar borib buyuk biri mag'lub etdi, Li Sedol, Shu yil.

Balki siz kompyuter kompyuter o'yinlari yaxshi bo'lishini kutyapmiz? ular nima bilamiz bir marta, Ular, albatta, har qanday inson ortiq izchil tezroq va ko'proq buni. DeepMind ning Breakout futbolchi hech narsa bilar, ammo. Bu qanday o'yin ishlar bo'yicha ko'rsatmalar bilan dasturlashtirilgan emas edi; u hatto boshqaruvlari qanday foydalanish aytgan edi. u bor edi, barcha imkon qadar ko'p ball olish uchun sinash uchun ekran va amri tasvirni edi.

Watch video. Boshida, Ketma-unutilib to'p tomchi beradi, yaxshi bilish. Oxir-oqibat, faqat haqida mucking, u qaytib to'pni turadi, a g'isht yo'q qiladi va bir nuqtaga oladi, Bas, bu tan va yana tez-tez uni qilsa. ikki soat amaliyot so'ng, yoki haqida 300 o'yinlar, u jiddiy yaxshi bo'ldi, siz yoki men hech bo'ladi ko'ra yaxshiroq. so'ng, keyin haqida 600 o'yinlar, narsalar qo'rqinchli olish. algoritm Shu nuqtaga qaratilgan boshlanadi, yana va yana, orqasida kosmosga g'isht orqali in uchun. bor bir marta, har qanday Breakout futbolchi biladi sifatida, to'p, bir muncha vaqt uchun atrofida pog'ona bo'ladi, bepul ochko to'plab. Bu kompyuter o'z bilan keldi yaxshi strategiya bo'ldi.

"Bizning tadqiqotchilar ko'rgach, bu, bu aslida ularni larzaga,"DeepMind rahbari, Demis Hassabis, Parijda texnologiyasi anjumanida bir tomoshabin berdi. Siz .. qila olasiz; siz ... mumkin uning namoyish tomosha, ham, mashina uning Burrowing strategiyasini amalga raqamlar qachon va Kulgi va qarsaklar eshitish. kompyuter aqlli aylandi, biz kabi bir oz.

"Sun'iy aql" faqat eng qadimiy va barcha hisoblash ning Buzz iboralar eng asabiy haqida. g'oya birinchi tomonidan jiddiy buni zimmasiga oladi deya edi Alan Turing ichida Hisoblash mashinalari va razvedka, The 1950 Qog'oz unda sifatida tanildi nima taklif Turing test: mashina inson ekanligini suhbat orqali sizni ishontirish mumkin, agar, u, albatta, o'ylab qilingan isbotlash uchun har qanday inson mumkin qadar qilayotganini. Lekin muddatli AI odatda qadar ishlatilgan emas edi 1955, qachon Amerika matematik Jon McCarthy mutaxassislari uchun konferentsiya taklif. Bu quyidagi yil o'tdi, va shundan buyon faoliyat mani va umidsizlikka bir taxminan ikki-o'n yil tsikli ishlaydigan bo'ldi. (moda chiqib, uning sehr tasvirlash uchun - "AI qish" - Tadqiqotchilar hatto yangi muddatni bor. 1970 va 1990 ayniqsa qattiq edi.)

Bugungi kunda yangi mania bor, boshqalardan farq ko'rinadi qaysi: bu sizning cho'ntagingizda mos. A Telefon shaxmat bo'yicha jahon chempioni mag'lub mumkin, Sizning bolalar radio qo'shiq va suratga tan, va boshqa tilga tarjima ovozingni. bilan bu erda tasvirlangan Nao robot Yotam Ottolenghi ikki oyoq ustida yurishimiz mumkin, gapirish, to'pni va hatto raqs topish. (Bu robot bo'ldi, garchi, emas AI: u bir menyu loyihalashtirish mumkin emas.)

AI o'zgarishlar haqida eshitib,, Agar xursand bo'lishi sizga mutaxassis shart emas, yoki qo'rqib. Siz faqat his olish boshlash: razvedka erda. Shubhasiz Google tuyg'u bor, ham, u bir mish-mishlarga ko'ra $ 650m uchun DeepMind sotib, chunki. Yilda 2013, Facebook o'z loyihasini ishga tushirdi, sayt uchun yuz va tabiiy til tan rivojlantirish rejalari bilan. Dasturchilar allaqachon aqlli Chatbots ustida ish boshladi, Facebook foydalanuvchilari uning Rasuli xizmatidan foydalanib, chaqirish mumkin bo'ladi qaysi.

Shu paytgacha, hozirgacha, Kompyuterlar da "aqlli" bo'lmagan, yoki faqat tor, shuning uchun. Ular bizni ajablantirmasin oson vazifalarni yaxshi edi ayting, Bunday matematika sifatida, lekin, biz qadriga yetmoq o'sha da yomon, jiddiy qiyin bo'lishi qaysi. yurish qilmish narsa zamonaviy robotlar bilan kurash hali Kichkintoylar kabi o'rganish va; asosiy oyalanarak vazifalari uzoq orzu qoladi. "Bir misol, siz yoki men birovning oshxonada bir piyola choy qilish mumkin bo'lgan qulaylik,"deydi Professor Alan Winfield, G'arbiy Angliya universiteti roboticist. "Buning mumkin sayyorada bir robot bor emas."

bo'lish inson shunday qiyin nima uchun tushunish, Agar fotosuratlar odamlarni tan kompyuter olish mumkin qanday haqida o'ylayman. AI holda, Agar avval uni o'zingiz qanday bilish kerak, kompyuter dasturiy uchun. Siz to'plash va barcha mumkin bo'lgan naqsh haqida o'ylash kerak, ranglar va yuzlari shakllari, va ular qanday yorug'likda va turli rakurslarda o'zgartirish - va nima muhim va nima bilish kerak optikasi faqat loy bo'lgan. AI bilan, Agar tushuntirish shart emas: Agar faqat kompyuterga real ma'lumotlar tog 'berish va uni o'rganish qilaylik. Qanday o'quv dasturini loyihalashtirish bir ezoterik masala bo'lib qolmoqda, bir necha harakat-keyin kompyuter olimlar viloyat, lekin ular miyada tuzilmalar ustida erkin asoslangan ma'lumotlar qayta ishlash tuzilmalarni yaratish bilan g'olibga bor ayting tushunarli. (Bu "chuqur o'rganish" deb ataladi.) real ma'lumotlar tog'larida kelsak, yaxshi, bu nima Google, Facebook, Amazon, Uber va barcha qolgan atrofida yolg'on bo'lishi sodir.

Bu bosqichda, Biz hali eng yaxshi chiqadi AI foydalanadi qaysi bilmayman. Josh Newlan, Shanxay ishlaydigan Kaliforniya coder, cheksiz Konferentsiya qo'ng'iroqlar tinglash bilan zerikib qolibdi, shunday unga tinglash uchun, ba'zi dasturiy ta'minot qurilgan. Hozir, qachon Newlan ismi zikr qilingan, uning kompyuter zumda unga so'nggi yarim daqiqada bir transkript yuboradi, kutib 15 soniya, keyin unga bir qayd o'ynaydi, "Kechirasiz, Men mikrofon soqov edi tushunishmadi. "O'tgan yili, Josh Browder, Britaniya o'smir, qurilgan to'xtab chipta qarshi murojaatlari bepul sun'iy advokati; U chet ellik yuridik tizimlari orqali qochqinlarni hidoyat yana bir qurishni rejalashtirmoqda. imkoniyatlari Xo'sh ... bor, balki bir algoritm imkoniyatlarni hisoblash mumkin.

Shunday qilib, mashina aqllari bir kun o'z o'zuvchidir qiladi? Men uchun gapirish tadqiqotchilar ehtiyotkor bo'ladi, va ularning mashinalari qila olmaydi nima ta'kidlashni Zahmatlar. Lekin men sinov uchun AI qo'yish qaror qildi: u ovqat, shuningdek, Ottolenghi rejalashtirishingiz mumkin? bu mening portret chizish mumkin? texnologiya hali sun'iy aqlli emas - yoki aqlli bo'lishi boshlanadi, haqiqatdan?

ovqat pishirish test

yaxshi, Men u dahshatli emas derlar. Odamlar meni yomon xizmat. haq nomi IBMning Chef Watson, bu taom beradi-da ("Tovuq Jigar Savoury SOS") u loyiq, taxminan, deb ishtahani bo'ladi.

Watson Chef adolatli bo'lishi uchun, va Guardian Dam olish kunlari o'z chef-yozuvchi Yotam Ottolenghi uchun, Men ularga juda vazifasi belgilangan edi. Men bir-biriga yaqin, hech bir joyda tegishli tuyulardi to'rt moddalar asosida bir piyola so'radi: tovuq jigar, yunon yogurt, Vasabi va tekila. Ular yoqdi har qanday boshqa qo'shishingiz mumkin, lekin o'sha to'rt tayyor taom bo'lishi kerak edi, Men ovqat pishirishni va taom bo'lardi. Oshpaz Watson tortinmaydi, bir zumda menga ikki makaron SOS berib. Ottolenghi ko'proq ehtiyot edi. Men muammo bor "deb o'yladim, "Bu ish qilmoqchi emas,"U menga.

Men bir xil deb o'yladi. Yoki kamida Men ularning ashyolardan qaramay OK bo'lishi muvaffaq ikki ovqatlarni eb yakun deb o'yladim, aksincha, chunki ularning ko'ra. Aslida - va siz menga urinyapti deb o'ylayman olaman, lekin shunday, nima - Ottolenghi ning retsepti nozil bo'ldi: jigar va piyoz va tekila kamaytirish, olma bilan xizmat, rediska, qandlavlagi va sachratqi Hammayoqni Salatası, a vasabi va yogurt echinish bilan. ovqat qog'ozda oz mantiqiy bo'lishi mumkin, lekin men har bir element tegishli bir likop to'la narsa his yutib. (Va salat yogurt va Wasabi o'rniga xantal bilan quyuqlashgan: jiddiy, Bir harakat qilib.) Ottolenghi retsept duyurulabilir bir mo'ylov qisqa menga.

narsa, bu taom takomillashtirish, uni va uning jamoasi uch kun o'tdi. Ular tatib va ​​lazzatlari muhokama qilish imkoniyatiga ega bo'ldilar, tezlatish, ranglar, haroratlar, bir tarzda, bu Watson olmaydi - kelajakda bir fikringiz mexanizmini qo'shib haqida "muhokama" bor edi-da,, Watson ning etakchi muhandis chef, Florian Pinel, meni aytadi. "A retsept bunday murakkab narsa,"Ottolenghi deydi. "Meni ham bir kompyuter uni yaqinlashishga qanday tushunish uchun qiyin."

Yotam Ottolenghi va Chef Watson idishlari
Yotam Ottolenghi va Chef Watson idishlari Fotosurat: Guardian uchun Jay Brooks

Watson birinchi IBM tomonidan qurilgan televizor gameshow xavf g'alaba qozonish uchun! ichida 2011. Ba'zi jihatdan, u bir chalg'ituvchi qiyin bo'ldi, bir kompyuter uchun bir viktorina qattiq qismi savollarga tushunish, chunki, javob bilmay; insonlar uchun, u boshqa yo'li. Lekin Watson qo'lga kiritdi, va uning texnologiyasi, boshqa joyda qo'llaniladi boshladi, a oshpaz sifatida, shu jumladan,, yangi retseptlar yaratish asosida 10,000 olingan real misollar Yoqimli ishtaha jurnali.

Birinchi dasturi bu ta'riflari "yutmoq" kerak edi, Watson jamoasi uni qo'yish kabi. hisoblash, bir poda ingredientlar nima tushunish kirdi, Ular tayyorlandi qanday, ular uchun pishirilgan edi qancha vaqt, maqsadida yangi taomlar ularni qanday foydalanish tushuntirish imkoniyatiga ega bo'lish uchun. (jarayon hali ham teskari borish mumkin. Hatto hozir ham Chef Watson "yumuşakça" deb nomlangan moddani tavsiya, qaysi u oydinlik tushuntiradi "oltinchi to'liq uzunligi O'ylamoq tomonidan albom".)

A katta muammo Mashina ta'mi hissini berish uchun harakat qilindi. kompyuter bir yangi kombinatsiyasini yaratish uchun "oson yetarli,"Pinel deydi, "Lekin qanday qilib bir baholash mumkin?qaysi minglab bor - - "Watson xos lazzat birikmalar kombinatsiyasi sifatida har bir tarkibni ko'rib o'rgatgan va keyin umumiy birikmalar edi moddalar birlashtirish. (bu tamoyil, oziq-ovqat xaritada, yaxshi insonlar orasida belgilanadi.) nihoyat, dasturiy ta'minot, inson oshpaz uchun mantiqiy qadam-baqadam ko'rsatmalarni hosil. e'tibor amaliy taom rejalashtirish o'rniga kutilmagan haqida. "Chef Watson sizga ilhom uchun, albatta, bor,"Pinel tushuntiradi. Har bir retsept "o'z ijodiy va hukm foydalanish" uchun eslatma bilan keladi.

Va men kerak. birinchi qadam "Sog'ligingiz tekis-yaproq petrushka" iborat, qaysi faqat yaxshi fikr emas. Men qilaman, samarali, sekin-sekin pishirilgan boyitilgan cho'chqa va mol go'shti Ragu, mening barcha to'rt moddalar, shu jumladan,, hali Watson g'alati, shuningdek, bodring o'z ichiga oladi va "bosh qalampir daraxti bilan mavsum" meni aytaman tutadi, Men tamoyiliga qilish rad qaysi. Oxirida, Men The Farmyard juda yaqin bir lazzat bilan boy sous bor, lekin yeyilmaydigan emas. Men vasabi yoki tekila tatib mumkin emas, Men xursandman qaysi.

Nao robot bilan Yotam Ottolenghi
Nao robot bilan Yotam Ottolenghi Xaber boshlang'ich maktab odobli qarz, London. Fotosurat: Jay Brooks. Styling: Li Flude

Watson aqlli va vazifa qiyin, lekin men bu oziq-ovqat Ilmga qiziqqanlar uchun kulgili bir oz ko'p bo'lmagan, deb aytish uchun tayyor emasman, Ottolenghi meni to'xtaydi qadar. "Men go'sht bir oz karaciğerlerine sekin-ovqat pishirish g'oyasi katta, deb o'ylayman," u aytdi. "Bu lazzat kuchaytiradi. Hamma narsa birga keladi. Men bu retsept bilan yangidan boshlash kerak edi, agar, aniq, albatta yogurt mos emas - lekin men u erda, to'q sariq teri tark edi, ziravorlar bir necha. Menimcha, bu juda yomon retsept deb o'ylamayman. Bu ish mumkin. "

Hukm Watson moddalar soğukluk maxfiy qilsangiz, lekin Ottolenghi ularni kuylash qiladi.

yozuv test

oz qo'yish So'z ustasi IBM va Google dahshatli mashinalari yonida, va u cho'ntak kalkulyator sifatida hisoblash ilg'or ko'rinadi. Hali Watson uning ishlaydilar orqali fumbles esa, So'z ustasi ishda allaqachon. Agar Associated Press fond bozori hisobotlarni o'qisangiz, yoki Yahoo sporti jurnalistika, Agar ular bir kishi tomonidan yozilgan O'ylab ko'raman yaxshi imkoniyat bor.

So'z ustasi sun'iy yozuvchi. Avtomatlashtirilgan Insights deb nomlangan Shimoliy Karolina kompaniya tomonidan ishlab chiqilgan, u ma'lumotlar eng qiziqarli Nuggets plucks va maqola tarkibiga, ularni foydalanadi (yoki elektron pochta, yoki mahsulot listing). Bu, albatta, katta yangiliklar bo'ylab kelganda, u ko'proq hayajonli tilini foydalanadi. Uning ish yana ukish qilish Diksiyonu va sintaksisini o'zgaradi. Hatto bir qo'pol robot oshpaz o'z qo'llanishini bo'lishi mumkin, lekin inson O'quvchilarning uchun yozma silliq bo'lishi kerak. ovozli-tan qurilma kabi bog'laganman Amazon Echo, So'z ustasi ham bir og'zaki inson savolga javob berishi mumkin - o'z sarmoyasi bajarish haqida, say – with a thoughtfully spoken answer, announcing what’s interesting first, and leaving out what isn’t interesting at all. If you didn’t know the trick, you’d think Hal 9000 had arrived.

The trick is this: Wordsmith does the part of writing that people don’t realise is easy. Locky Stewart from Automated Insights gives me a tutorial. You write into Wordsmith a sentence such as, “New ABC figures show that the New York Inquirer’s circulation rose 3% in April.” Then you play around. The 3% has come from your data, so you select the word “rose” and write a rule, known as a “branch”, which will change the word “rose” to the phrase “shot up” if the percentage is more than 5%. Then you branch “rose” to become “fell” if the percentage is negative. If the percentage is -5% or lower, “rose” becomes “plummeted”.

Then you feed it synonyms. So “plummeted” can also be “fell sharply by”. “The Inquirer’s circulation” can be “circulation at the Inquirer”. “Shot up” can be “soared” and so on. Then you add more sentences, perhaps about online traffic, or about which days’ print copies sold best, or about comparisons year-on-year. Then you get clever. You tell Wordsmith to put the sentences with the most newsworthy information first, defined perhaps as those that feature the greatest percentage changes. Maybe you add a branch to say that a result is “the best/worst performance among the quality titles”. Hell, you can even teach it some old Fleet Street tricks, so that if circulation plummets the piece begins “Editor Charles Kane is facing fierce criticism as”, but if circulation has “shot up” this becomes “Charles Kane has silenced critics with news that”. Insert “more” or “again” or “continues” if you get the same thing two months in a row.

“The artificial intelligence is actually the human intelligence that is building the network of logic,” Stewart says, “the same network you would use when writing a story. It could have been developed 10 yoki 15 yil avval, in code, but to make it work at this scale has only been possible lately.” Clearly it takes longer to prepare an article on Wordsmith than to write one conventionally, but once you’ve done so, the computer can publish a fresh newspaper circulation story every month, on every newspaper, within seconds of receiving the information. It can publish millions of stories in minutes – or publish only some of them, if the data doesn’t reach a given threshold of newsworthiness. Thus it becomes an automated editor, ham, with adjustable tastes in thoroughness, frequency and hysteria.

For Wordsmith’s task, I suggest football: it’s a field that produces a lot of data and has a readership that wants personalised articles. Guardian football writer Jacob Steinberg volunteers to take on the computer, and I provide a table of facts from the recent Premier League: last season’s league position and this season’s position at Christmas and at the end, goals scored and conceded, top scorer’s name and total, value of summer transfers and a quote from the manager.

Working solely from this data, computer and human must each write a review of the season for a given club. Steinberg chooses Leicester City on the basis that its numbers should contain a story that anyone would see. Wordsmith doesn’t need to choose. It will do all 20.

And in fact both computer and human quickly produce quite similar work:

Leicester City footballer Jamie Vardy

Both Steinberg and Wordsmith deliver dramatic first sentences. Perhaps keen to sound authentic, Automated Insights use some clever tricks to put feeling into the latter’s article, astutely guessing that Leicester were “hoping to finish in the top 10 after a 14th place finish last season”. I look through Wordsmith’s other articles and Southampton, having finished seventh last season, have “eyes on a European spot”, while Manchester City “began the season dreaming of a league title after finishing second”.

Conversely, Steinberg digs more meaningfully into the numbers, showing that Jamie Vardy not only scored 24 maqsadlar, but that this was a higher percentage of his team’s goals than was managed by all but two other players. Knowing how Wordsmith works, albatta, one could easily set it up to do the same. In fact looking through it, Steinberg’s entire article could have been created by a skilled Wordsmith programmer – with the exception of one line. “It’s a magical season,” he quotes the Leicester manager as saying, before adding, “justifiably so, given that a summer expenditure of £26.7m on transfers made them the eighth lowest spenders”. That “justifiably so” shows a writer who actually understands what he is writing.

Hukm Steinberg is a much better writer, unless you want 20 data-heavy articles in 10 daqiqa.

The painting test

A laptop wants me to smile. “It’s in a good mood," Simon Colton deydi. He knows because he’s the scientist who programmed it. We are in the Science Museum in London, where the Painting Fool, as it is called, is giving a public demonstration. It’s important that I don’t show my teeth, Colton says, because something about the light makes them look green to the Painting Fool.

From my toothless smile the laptop creates a “conception” of what it would like to paint, based on its mood. The mood comes from a “sentiment analysis” of recent Guardian articles, as it happens (on average reading the Guardian is a downer, shubhasiz, apart from the stuff about gardening). Yesterday the Fool was in such a bad mood that it sent someone away unpainted; today it is feeling “positive”.

Next the Fool attempts to paint with a simulated brush and a simulated hand (actually, an image of Colton’s hand) on the screen behind me. It learned to reflect its mood from the work of Dan Ventura, another computer scientist, at Brigham Young University in Utah, who trained a neural network to recognise the emotional attributes of images by sitting thousands of people in front of tens of thousands of paintings and asking them to tag each one with whatever adjectives came to mind. The Fool now knows that bright colours reflect a good mood, and “pencils with tight hatching” create a picture that is “cold”. When it is done, it prints out a page with a typed self-critique. “Overall, this is quite a bright portrait,” it says. “That’s OK, but my style has lowered the level of bright here. So I’m a bit annoyed about that.”

Here along with us, intrigued but too busy at her easel to watch, deb Sarah Jane Moon, an artist who exhibits with the Royal Society of Portrait Painters. She doesn’t want to see my teeth, yoki. “We paint from life," u aytadi, “and you can’t hold a smile for sitting upon sitting. That’s why all the traditional portraits show quite relaxed features.”

The Painting Fool is a special machine, and even slightly famous, but I can’t deny that Moon is almost all of why I’m excited to be here. The feeling of being painted by a real person, having them look at you and think about you, is exciting and flattering. Sentiment analysis and training data, boshqa tarafdan, don’t add up to anything whose view of me I care about, and the finished portraits do not change my mind. Moon’s is a lovely, real thing, which feels straight away like one person seen by another. The Fool’s three efforts have qualities I like, but mostly they look like photographs that have gone through some kind of software filter. Colton insists the Fool is here “to learn to be better” but I look and think: so what?

Painting of Leo Benedictus by Sarah Jane Moon
Leo Benedictus as seen by Sarah Jane Moon…
Painting of Leo Benedictus by the Painting Fool computer
…and as imagined by the Painting Fool laptop. Fotosurat: Murray Ballard

Then I think some more. For one thing, it turns out that art is more mechanical than I’d realised. “I try to look at Leo as an abstract set of shapes, shakllari, ranglar, tones,” Moon tells Colton, “to get away from the fact that that’s a nose. Because when you start to do that, you get caught up in what you think looks like a nose.”

“What the software does is break it down into colour regions,” Colton says.

"Ha,, aniq,” Moon agrees. “I think that’s what the best painters do. It’s transcribing.” Afterwards she tells me she felt a kind of “kinship” with the software as they worked side by side.

Eng muhimi, I realise that what matters isn’t how the machine paints; it’s how I see. Moon I understand, O'ylaymanki. She’s a person and I know how that feels, so I care about her picture. But what does it feel like to be the Painting Fool? Is that what its portraits are trying to tell me?

Hukm Moon’s painting is far richer; the Fool is still learning and has centuries of practice to go.

The translation test

Google Translate was the first piece of proper science fiction to come true, va it’s already a decade old. In many ways it typifies where AI has got to. Useful, ishonch; impressive, without question; but still clunky as hell, despite big improvements.

If you haven’t used it, it works like this: enter text or web links in any of 103 supported languages and you get a rough translation seconds later in any of the others. The app on your phone will transcribe what you say and then speak it back, tarjima (32 languages supported); it can replace the text of a foreign language sign or menu wherever you point the camera. No explanation is needed of how cool that is (and it’s free).

Globally, half a billion people use Google Translate each month, mostly those who don’t speak English (which is 80% of people) but who want to understand the internet (which is 50% Ingliz tili). “Most of our growth, and actually most of our traffic, comes from developing or emerging markets such as Brazil, Indoneziya, Hindiston, Tailand,” says Barak Turovsky, head of product management and user experience at Google Translate. It’s surprisingly popular for dating, ham, he adds. “Things like ‘I love you’ and ‘You have beautiful eyes’, that’s very prevalent.”

The software has always used a form of statistical machine learning: scouring the internet for already translated text – UN declarations, EU documents – and mapping the likelihood of certain words and phrases corresponding to one another. The more data it gathers, the better it gets, but the improvement levelled off a couple of years ago. yaqinda, Turovsky says, they will deploy new deep learning algorithms, which will produce much more fluent translations.

Shunday bo'lsa ham, there are limits, and some seem fundamental when you talk to a human translator and realise how subtle their work is. Ros Schwartz va Anne de Freyman volunteer for this task. Both are professional French/English translators, and I need two because, in order to judge how good the translation is without being fluent in both languages, we need to translate twice – once out of English into French, once back again. Google Translate keeps no memory of the original and can do the same thing.

I choose a short passage of distinctive but not especially wild or ambiguous prose from the beginning of Herzog by Saul Bellow. Translators normally require context, so I tell Schwartz and De Freyman that it comes from a famous mid-century American novel.

Within a few days, Schwartz and De Freyman return a very smooth facsimile of the original text. Here and there some nuances have not survived, but the passage remains a pleasure to read, and the main meanings come across exactly.

Google Translate takes only a few seconds, and the result is both impressive and inadequate, weirdly good in places, in others weirdly bad – turning “he” into “it” and concocting the idea that Herzog is in love. Miraculously, it keeps “cracked” as a description of the hero. French has no word that combines the sense of “broken” and “mad” that cracked coveys in English, so De Freyman makes it “cinglé”, which comes back from Schwartz as “crazy”.

“Google Translate would look at statistical probability and say, what does ‘cracked’ mean?” Turovsky explains. “And statistically, it will try to decide whether it means ‘cracked’ or ‘crazy’ or whatever. ekan, for a machine, is a non-trivial task.” Nor is it simple for a human, even though we find it easy. You’d have to ask whether Bellow could have meant that Herzog was “cracked” as in physically fractured. Then you’d have to assume not, because human bodies don’t generally do that. So you’d wonder what he did mean and assume instead, if you were not already familiar with the usage, that he must mean “crazy”, because you understand the rest of what you’ve read. But to do all this, wouldn’t Google Translate have to be pretty much conscious, I ask? Turovsky laughs. “I don’t think I’m qualified to answer that question.”

Hukm Some bullseyes and howlers from Google Translate, while Schwartz and De Freyman are fluent and exact.

guardian.co.uk © Guardian Yangiliklar & Media Limited 2010

Haqida Maqolalar