Is Europe Bound for a Trumpian Economic Reset?

"Having posited last week that the incoming Trump administration eventually would force Europe to shape up its own economy, I still must profess surprise at how quickly this process is unfolding. Witness the new growth strategies unveiled in Brussels and London on Wednesday.

U.K. Chancellor of the Exchequer Rachel Reeves on that day delivered a major speech on economic growth. The backdrop is a stagnating British economy plagued by persistent inflation, surging tax bills and plummeting business confidence. In their desperation to fix things, Ms. Reeves and her boss, Prime Minister Keir Starmer, are making a habit of these set-piece policy "resets" -- you'd think Britain's economy was a faulty Wi-Fi router -- but each time they manage to stumble a little further in the right direction.

In this outing, Ms. Reeves promised to slash through the bureaucratic and permitting red tape that makes it difficult to build anything in Britain.

The marquee project will be a third runway for London's Heathrow airport. That hypothetical stretch of tarmac has been discussed and debated for years by politicians of all stripes and has always faced intractable environmentalist opposition whenever an opportunity arrived to issue a final go-ahead.

Who knows if this time will be any different for Heathrow, but for now what matters is the implicit signal Ms. Reeves sent with this announcement: Economic growth, as embodied by an airport-building project, is more important to Mr. Starmer's Labour Party administration than the climate-change neuroses that so often have thwarted the new runway in the past. This is why the climate left is angry about Ms. Reeves's policy turn despite bones she threw to them in her speech, such as more spending on electric-vehicle charging points and less red tape for offshore wind generation.

Meanwhile across the English Channel, the European Union is trying to get its own economic act together. European Commission President Ursula von der Leyen on Wednesday laid out a new policy agenda called the "Competitiveness Compass."

Most of it, including the voter-friendly summary, reads like characteristic Brussels argle-bargle, but to the extent the project has a bottom line, it's this: Europe's climate policies are ruining the economy and something must be done.

Ms. von der Leyen, only months into her second term, appears to be promising a rethink of the "European Green Deal" that constituted her main accomplishment in her first term. Her new "compass" points toward less regulation, with a heartening if implausible numerical target for reducing the red-tape burden on business. Most important, the strategy hints at an imminent softening of the EU's electric-vehicle mandate, which already is starting to bankrupt Europe's auto industry a decade before it takes full effect.

Now, let's not get ahead of ourselves. Europe's policy turns this week are small in scale compared with the wholesale realignment under way in America. In part the difference is structural. As the holder of a powerful executive office within a relatively unified nation, Mr. Trump enjoys greater latitude to redirect government energies than a European prime minister, and certainly more authority than a European functionary in chief such as Ms. von der Leyen.

European leaders also face a political problem largely of their own making. European voters have been conditioned by decades of preachy politicking from left and right to believe more fervently in the climate agenda than the U.S. electorate ever did. As a result, European capitals are teeming with elected politicians mostly but not exclusively of the left who are eager to thwart any turn from climate obsessiveness and toward economic growth.

Recent elections have started to reverse this trend, but there's still political life in Europe's green left. Mr. Trump, in contrast, has the electoral wind at his back when he cans his predecessor's EV mandates or drills, baby, drills.

Expect that Europe's return to economic and climate sanity, if that's what we're witnessing, will be an iterative process rather than a big bang. Sure enough, "iterative" is what Ms. Reeves and Ms. von der Leyen delivered this week, despite all their hype.

Left unanswered are crucial climate questions, such as what role fossil fuels and nuclear power will play in energy production (the only economically sound answer is "a big one"), and unavoidable economic questions such as from where private investment capital will come given Europe's suffocating taxes and financial regulations.

But do give the Europeans credit for getting the first and most important decision right, so far. This is to follow along in Mr. Trump's pro-economic-growth slipstream rather than trying to swim against the current.

A continent that basked in its liberal-green virtue-signaly opposition to anything Mr. Trump did in his first term now finds it doesn't have the luxury to do it again amid accelerating deindustrialization and the deteriorating living standards of European voters. They didn't beat him the first time. That leaves them to try to join him the second time around." [1]

Switching back to Russian cheap natural gas perfectly matches this turn of events. The world needs functioning German industry to make an affordable transition to a green economy. Nobody wants to buy overly expensive heat pumps. This is why Germany's Scholz and France's Macron are dead in the water right now.

1. Political Economics: Is Europe Bound for a Trumpian Economic Reset? Sternberg, Joseph C.  Wall Street Journal, Eastern edition; New York, N.Y.. 31 Jan 2025: A15.   

 

„DeepSeek“ panaudojo „taktikos margumyną“, „apeidama JAV.“ --- Dirbtinio intelekto paleidimas pasitikėjo nepatyrusiais inžinieriais, eksporto kontrolės spraga

 

 "SINGAPŪRAS. Paimkite jaunų kinų inžinierių komandą, kurią pasamdė viršininkas, niekinantis patirtį. Pridėkite keletą protingų programavimo sparčiųjų manevrų ir JAV taisyklių spragą, leidžiančią jiems gauti pažangius lustus.

 

 Būtent tokią formulę Kinijos „DeepSeek“ naudojo, kad šokiruotų pasaulį savo dirbtinio intelekto (AI) programomis.

 

 Įprastas mąstymas manė, kad, kuriant pirmaujančią dirbtinį intelektą reikia daugybės brangių, pažangiausių kompiuterių lustų, o Kinijos įmonėms kiltų problemų konkuruoti, nes jos negalėtų gauti tų lustų. „DeepSeek“ išradingai nepaisė šios prognozės, dėl ko Volstryte nusiaubė 1 trilijoną dolerių, ir paskatino Silicio slėnį permąstyti savo požiūrį.

 

 Pasak prezidento Trumpo, Kinijos bendrovė pažadino Vašingtoną. Jo administracija artimiausiais mėnesiais nuspręs, ką daryti dėl Bideno eros politikos, ribojančios Kinijos prieigą prie geriausių dirbtinio intelekto lustų.

 

 „DeepSeek“ lyderis Liang Wenfeng įkūrė savo įmonę Hangdžou technologijų centre, tame pačiame mieste, kuriame yra technologijų milžinė „Alibaba“. AI įmonė išaugo iš rizikos draudimo fondo, kurį bendrai įkūrė Liangas, kuris naudoja dirbtinį intelektą pelningiems sandoriams finansų rinkose rasti.

 

 Duodamas interviu vienam Kinijos leidiniui 2023 m., Liangas teigė, kad daugumą techninių pareigybių užima ką tik baigę absolventai arba žmonės, turintys vienerių ar dvejų metų patirtį.

 

 Patirtis, pasak jo, buvo galima kliūtis. „Kai ką nors darydami, patyrę žmonės nedvejodami pasakys, kad reikia daryti taip, o nepatyrę žmonės turės ne kartą tyrinėti ir rimtai galvoti, kaip tai padaryti, o tada rasti sprendimą, atitinkantį esamą faktinę situaciją, Liang pasakė.

 

 Tai, ką jie sugalvojo, dabar tiria Silicio slėnio geriausi ir ryškiausi.

 

 Dar visai neseniai novatoriški AI modeliai, kurie yra už tokių program, kaip „OpenAI ChatGPT“, buvo mokomi dėl daugybės teksto, vaizdų ir kitų duomenų rinkinio. Jie naudojo specializuotus algoritmus, kad surastų modelius, kuriuos pokalbių robotas galėtų naudoti pokalbiui palaikyti.

 

 „DeepSeek“ taktika buvo sumažinti duomenų apdorojimą, reikalingą modeliams parengti, naudojant kai kuriuos savo išradimus ir metodus, kuriuos priėmė panašiai suvaržytos Kinijos AI įmonės.

 

 Įsivaizduokite ankstesnes ChatGPT versijas kaip bibliotekininką, perskaičiusį visas bibliotekoje esančias knygas, sakė Lennartas Heimas, tyrinėjantis dirbtinį intelektą tyrimų centre „Rand“. Kai užduodamas klausimas, jis pateikia atsakymą pagal daugybę perskaitytų knygų.

 

 Šis procesas yra daug laiko ir brangus. Norint skaityti tas knygas, reikia elektros ištroškusių kompiuterių lustų.

 

 DeepSeek pasirinko kitą požiūrį. Jos bibliotekininkas neperskaitė visų knygų, bet yra išmokytas surasti tinkamą knygą, kad gautų atsakymą, kai jam užduotas klausimas.

 

 Sluoksniuota ant to yra kita technika, vadinama „ekspertų mišiniu“.

 

 Užuot bandę surasti bibliotekininką, galintį atsakyti į klausimus bet kuria tema, „DeepSeek“ ir kai kurie kiti dirbtinio intelekto kūrėjai daro kažką panašaus į klausimų delegavimą konkrečių sričių, pvz., grožinės literatūros, periodinių leidinių ir kulinarijos, ekspertų sąrašui. Kiekvienam ekspertui reikia mažiau mokymų, todėl sumažėja lustų poreikis atlikti viską iš karto.

 

 DeepSeek metodas reikalauja mažiau laiko ir galios, prieš užduodant klausimą, tačiau atsakant sunaudojama daugiau laiko ir energijos. Atsižvelgiant į viską, Heimas sakė, kad „DeepSeek“ spartieji klavišai padeda mokyti dirbtinį intelektą už nedidelę konkuruojančių modelių kainą.

 

 „Inžinerija yra apie suvaržymus“, – X. rašė buvęs „Intel“ vadovas Patas Gelsingeris. „Kinijos inžinieriai turėjo ribotus išteklius ir jie turėjo rasti kūrybiškų sprendimų.

 

 Išradingumas paaiškina tik dalį „DeepSeek“ sėkmės.

 

 Kita dalis yra sudėtingas JAV eksporto kontrolės įvedimas, dėl kurio „DeepSeek“ galėjo įsigyti galingų amerikietiškų lustų.

 

 Bideno administracija 2022 metais pradėjo kontroliuoti, į Kiniją eksportuojamus, lustus. JAV bendrovės, kurios norėjo parduoti Kinijai, pirmiausia, turėjo sumažinti lusto funkciją, vadinamą interconnect bandwidth, kuri nurodo duomenų perdavimo greitį.

 

 Reaguodama į tai, „Nvidia“, pasaulyje pirmaujanti dirbtinio intelekto lustų kūrėja, pasiūlė Kinijai naują produktą, kuris atitiko šį parametrą, tačiau kompensavo jį kitais būdais, išlaikydamas aukštą našumą. Taip atsirado lustas, kuris, pasak kai kurių analitikų, buvo beveik toks pat galingas, kaip tuo metu geriausias Nvidia lustas.

 

 JAV pareigūnai viešai ir privačiai išreiškė, kad nors „Nvidia“ nepažeidė įstatymo, tačiau pažeidė jo dvasią. Vyriausybė tikėjosi, kad pramonės lyderiai bendradarbiaus, kurdami veiksmingą greitai kintančių technologijų eksporto kontrolę, sakė buvęs aukštas Bideno administracijos pareigūnas.

 

 „Nvidia“ atstovas pirmadienį sakė, kad „DeepSeek yra puiki dirbtinio intelekto pažanga“, pademonstravus naujovišką AI techniką, naudojant skaičiavimo galią, „kuri yra visiškai eksportui leistina”.

 

Praėjus metams po pradinės kontrolės, vyriausybė sugriežtino taisykles. Vis dėlto „DeepSeek“ turėjo maždaug metus laiko įsigyti galingą „Nvidia“ Kinijos rinkos lustą, vadinamą H800. Gruodį paskelbtame moksliniame darbe „DeepSeek“ teigė, kad naudojo 2 048 šiuos lustus, kad apmokytų vieną iš savo AI modelių.

 

 Nuo tada, kai taisyklės buvo peržiūrėtos 2023 m., „Nvidia“ sukūrė naują su eksporto kontrole suderinamą lustą, skirtą Kinijai, kuris yra žymiai mažesnis, nei H800.

 

 Kai kurie JAV AI pramonės lyderiai skeptiškai vertina tai, kad „DeepSeek“ atskleidė visas savo paslaptis. Jie teigė, kad Kinijos mokslininkai galėjo sukaupti pažangiausių „Nvidia“ lustų atsargas prieš JAV apribojimus arba naudoti sprendimus, pvz., prieigą prie „Nvidia“ įgalintos skaičiavimo galios iš šalių, esančių už JAV ir Kinijos ribų. Bideno administracija paskutinėmis dienomis įgyvendino naujas taisykles, skirtas tokioms akloms zonoms pašalinti.

 

 „DeepSeek“ neatsakė į prašymus komentuoti. [1]

1. DeepSeek Used Bevy of Tactics To Outfox U.S. --- AI startup relied on inexperienced engineers, loophole in export controls. Woo, Stu; Huang, Raffaele.  Wall Street Journal, Eastern edition; New York, N.Y.. 29 Jan 2025: A1. 

DeepSeek Used Bevy of Tactics To Outfox U.S. --- AI startup relied on inexperienced engineers, loophole in export controls

"SINGAPORE -- Take a team of young Chinese engineers, hired by a boss with disdain for experience. Add some clever programming shortcuts, and a loophole in U.S. rules that allowed them to get advanced chips.

That is the formula China's DeepSeek used to shock the world with its artificial-intelligence programs.

Conventional thinking held that developing leading AI required loads of expensive, cutting-edge computer chips -- and that Chinese companies would have trouble competing because they couldn't get those chips. DeepSeek defied those predictions with a resourcefulness that led to a $1 trillion bloodbath on Wall Street and is spurring Silicon Valley to rethink its approach.

The Chinese company delivered a wake-up call to Washington, according to President Trump. His administration is set to decide in the coming months what to do about Biden-era policies limiting China's access to the best chips for AI.

DeepSeek's leader, Liang Wenfeng, built his company in the tech hub of Hangzhou, the same city where tech giant Alibaba is based. The AI company grew out of a hedge fund co-founded by Liang that uses AI to find profitable trades in financial markets.

In an interview with a Chinese publication in 2023, Liang said most technical positions were filled by fresh graduates or people with one or two years of experience.

Experience, he said, was a potential obstacle. "When doing something, experienced people will tell you without hesitation that you should do it this way, but inexperienced people will have to repeatedly explore and think seriously about how to do it, and then find a solution that suits the current actual situation," Liang said.

What they came up with is now being studied by Silicon Valley's best and brightest.

Until recently, the pioneering AI models that lie behind programs such as OpenAI's ChatGPT were trained on a vast compilation of text, images and other data. They employed specialized algorithms to find patterns that a chatbot could use to hold a conversation.

DeepSeek's tactic was to cut down on the data processing needed to train the models, using some inventions of its own and techniques adopted by similarly constrained Chinese AI companies.

Imagine the earlier versions of ChatGPT as a librarian who has read all the books in the library, said Lennart Heim, who researches AI at the think tank Rand. When asked a question, it gives an answer based on the many books it has read.

This process is time-consuming and expensive. It takes electricity-hungry computer chips to read those books.

DeepSeek took another approach. Its librarian hasn't read all the books but is trained to hunt out the right book for the answer after it is asked a question.

Layered on top of that is another technique, called "mixture of experts."

Rather than trying to find a librarian who can master questions on any topic, DeepSeek and some other AI developers do something akin to delegating questions to a roster of experts in specific fields, such as fiction, periodicals and cooking. Each expert needs less training, easing the demand on chips to do everything at once.

DeepSeek's approach requires less time and power before the question is asked, but uses more time and power while answering. All things considered, Heim said, DeepSeek's shortcuts help it train AI at a fraction of the cost of competing models.

"Engineering is about constraints," former Intel Chief Executive Pat Gelsinger wrote on X. "The Chinese engineers had limited resources, and they had to find creative solutions."

Ingenuity explains only part of DeepSeek's success.

The other part is the rocky introduction of U.S. export controls, which gave DeepSeek a window to buy powerful American chips.

The Biden administration in 2022 put in place controls on chips exported to China. U.S. companies that wanted to sell to China first needed to throttle a chip function called interconnect bandwidth, which refers to the speed at which data is transferred.

In response, Nvidia, the world's leading designer of AI chips, came up with a new product for China that complied with this parameter -- but compensated for it by maintaining high performance in other ways. That resulted in a chip that some analysts said was almost as powerful as Nvidia's best chip at the time.

U.S. officials vented publicly and privately that while Nvidia didn't break the law, it broke the spirit of it. The government had hoped that industry leaders would be collaborative in designing effective export controls on fast-changing technology, said a former senior Biden administration official.

An Nvidia spokesman said Monday that "DeepSeek is an excellent AI advancement" that demonstrated an innovative AI technique while using computing power "that is fully export-control compliant."

A year after the initial controls, the government tightened the rules. Still, that left an opening of about a year for DeepSeek to buy Nvidia's powerful China-market chip, called the H800. In a research paper published in December, DeepSeek said it used 2,048 of these chips to train one of its AI models.

Since the rules were revised in 2023, Nvidia designed a new export-control-compliant chip for China that is significantly less powerful than the H800.

Some U.S. AI industry leaders are skeptical that DeepSeek has revealed all of its secrets. They said Chinese researchers could have stockpiled leading-edge Nvidia chips before the U.S. restrictions, or used workarounds such as accessing Nvidia-enabled computing power from countries outside the U.S. and China. The Biden administration in its final days implemented new rules to address such blind spots.

DeepSeek didn't respond to requests for comment.” [1]

1. DeepSeek Used Bevy of Tactics To Outfox U.S. --- AI startup relied on inexperienced engineers, loophole in export controls. Woo, Stu; Huang, Raffaele.  Wall Street Journal, Eastern edition; New York, N.Y.. 29 Jan 2025: A1.