Sekėjai

Ieškoti šiame dienoraštyje

2024 m. birželio 5 d., trečiadienis

Asia Chip Giants Aim to Keep Edge Over U.S.


"In early April, South Korean President Yoon Suk Yeol didn't mince his words when describing why chips are paramount to the country's economic survival.

"The competition over semiconductors unfolding now is an industrial war," Yoon told government and industry officials. "An all-out war between countries."

South Korea has a massive war chest prepared for future semiconductor manufacturing: roughly $450 billion in private investment alone. That is roughly the same amount earmarked currently for chip production in the U.S., according to a recent industry estimate. With a plan sketched out to nearly 2050, Yoon's government plans to support the creation of the world's biggest chip-making cluster in South Korea, encompassing 37 factories, spanning eight cities and creating more than three million jobs.

Asia's chip-making champions in South Korea and Taiwan are U.S. allies, often seeing eye-to-eye on security and politics. But on semiconductors, they are increasingly friendly competitors -- and Seoul and Taipei, with clear advantages, aren't sitting idle as Washington looks to muscle back into production.

They offer lower costs, faster construction times and the benefits of an established supply chain. While a handful of Taiwanese and Korean companies are expanding their manufacturing footprints into the U.S., the most advanced technology is getting rolled out first back home.

Taiwan Semiconductor Manufacturing Co., commonly known as TMSC, is expected next year to introduce the world's smallest but most powerful chip in Taiwan -- the type of breakthrough needed for cutting-edge artificial intelligence and smartphones. Samsung Electronics and SK Hynix are set to make their next-generation memory chips in South Korea, with the government orchestrating ready access to engineers, natural resources and suppliers.

The stakes are high for these chip-dominant economies. Semiconductors represent about one-fifth of South Korea's exports. Unlike America's Silicon Valley, Taiwan relies on a so-called Silicon Shield -- the belief that its chip sector is so vital to global trade that it can deter a Chinese attack.

"They have an advantage because they have a broader ecosystem of suppliers across all facets of the supply chain," said John VerWey, a nonresident research fellow at Georgetown University's Center for Security and Emerging Technologies, a think tank.

To be sure, American companies such as Nvidia, Qualcomm and Apple design the world's best chips. But in recent decades, the ability to manufacture them has overwhelmingly resided in Asia.

The U.S.'s future gains as a chip maker could be amplified by extra rounds of congressional funding or a major leap from Intel, the sole American company in the three-way race with TSMC and Samsung to produce advanced logic chips. Asia's lead could also be imperiled should a military conflict break out in the Taiwan Strait or on the Korean Peninsula.

Though the U.S. share of global production of advanced logic chips is expected to grow significantly, the two Asian chip powerhouses will retain strength in key areas, including advanced logic-chip production and memory chips, according to a recent forecast through 2032.

Governments in both places are staying aggressive. In May, South Korea's government announced a $19 billion support package to bolster its chip industry -- more than double what had been floated just weeks earlier. Last year, Taiwan enacted government incentives, including a 25% tax deduction on research-and-development expenses, and these funds are now open for local companies to apply for.

Adding to the competition, China in recent days pledged roughly $48 billion to its national semiconductor fund, the largest installment ever. 

But with U.S. curbs in place, China's ability to quickly compete on the high end with Taiwan and South Korea is likely limited despite the influx of government funds.

Five decades ago, Taiwan pivoted toward semiconductors as part of its industrial transformation. Chips represented roughly two-fifths of the island's exports last year. Taiwan's economic affairs ministry, in written remarks to The Wall Street Journal in early May, called the industry the "divine mountain that guards the nation."

Chips made using TSMC's next-generation 2-nanometer process are on track to be mass-produced in Taiwan first in 2025, with U.S.-based production of such chips expected in 2028. TSMC's dominance stems from its singular focus on chip-making, substantial investments, and collaboration with top tech companies on their most advanced chip-design blueprints, according to industry experts.

Roughly the same geographic size as Maryland, Taiwan packs together industrial hubs connected by high-speed railroads, science parks and chip factories. Engineers stand ready 24 hours a day to troubleshoot problems at a moment's notice, according to Taiwan's Ministry of Economic Affairs. 

In the U.S., TSMC has pointed to several disadvantages from talent recruitment to work culture. As a result, the company plans to charge more for its chips made at its Arizona facilities.

TSMC also dominates in advanced packaging -- a key technique in which chips are stacked atop one another, and which is used to produce chips that power chatbots such as OpenAI's ChatGPT. Currently, all of TSMC's advanced packaging facilities are situated in Taiwan, and the company hasn't disclosed any plans to build them anywhere else.

"If TSMC has a problem, basically the whole global electronics industry has a problem," said Handel Jones, chief executive of International Business Strategies, a chip-industry consulting firm.

On top of its $19 billion support package for the chip industry, South Korea has already attracted research-and-development facilities from top foreign partners for its enormous semiconductor cluster that is under way. That includes the Netherlands' ASML Holding -- the only company making advanced lithography machines that can produce cutting-edge chips.

The government plans to inject another roughly $7 billion toward supporting R&D efforts for AI.

That includes further advances in the type of specialized memory chip critical for AI computing called high-bandwidth memory [1], or HBM, an area dominated currently by SK Hynix, which is currently Nvidia's main partner for the chips.

The R&D efforts could happen anywhere, but such advances would need to be located close to facilities churning out the latest high-end memory, said Yun Doo-hee, who oversees the AI chip initiative at South Korea's Ministry of Science and ICT. And those factories, he added, "are mostly here."

The South Korean government has promised faster land approval, guaranteed access to electricity and expanded tax credits for infrastructure investments. Support programs are under way for areas perceived as relative weak points, such as logic chips. Top local universities, in partnership with companies and regulators, have created "semiconductor majors" at both the undergraduate and graduate levels.

The perk: free tuition and a guaranteed engineering job at Samsung or SK Hynix upon graduation." [2]

  Even in the field of chips, where it is necessary to compete with "friends", the United States and the European Union are unable to compete. No staff, no infrastructure, no money, no required research.

1. "What is a high bandwidth memory?

HBM is a new type of memory chip with low power consumption and ultra-wide communication lanes. It uses vertically stacked memory chips interconnected by microscopic wires called "through-silicon vias," or TSVs. A large number of GDDR5 chips are required to reach high bandwidth."

2. Asia Chip Giants Aim to Keep Edge Over U.S. Sohn, Jiyoung.  Wall Street Journal, Eastern edition; New York, N.Y.. 05 June 2024: B.1.  

Dirbtinį intelektą kuriantys, darbuotojai bijo pareikšti susirūpinimą --- Grupė teigia, kad OpenAI, kitos firmos turi sukurti ataskaitų teikimo kanalus darbuotojams

„Dirbtinio intelekto (AI) pramonės darbuotojų grupė teigė negalintys išreikšti susirūpinimo dėl AI grėsmės žmonijai dėl konfidencialumo susitarimų, informatorių apsaugos trūkumo ir baimių dėl keršto.

 

 Antradienį paskelbtame laiške daugiau, nei tuzinas, esamų ir buvusių OpenAI, Google DeepMind ir Anthropic darbuotojų teigė, kad dirbtinio intelekto įmonės turi sukurti ataskaitų teikimo kanalus, kad darbuotojai galėtų saugiai pareikšti savo įmonėse ir visuomenei rūpimus klausimus. Jie teigė, kad konfidencialumo susitarimai neleidžia jiems viešai aptarti klausimų.

 

 „Žmonės, labiausiai išmanantys, kaip veikia AI sistemos ir su jų diegimu susijusi rizika, nėra visiškai laisvi kalbėti“, – sakė, laišką pasirašęs, buvęs OpenAI darbuotojas, Williamas Saundersas.

 

 Be Saunderso, laišką pasirašė dar šeši buvę OpenAI darbuotojai. Keturi dabartiniai „OpenAI“ darbuotojai ir vienas buvęs ir vienas dabartinis „Google“ AI tyrimų laboratorijos „DeepMind“ darbuotojas taip pat pasirašė savo vardus. Šeši iš pasirašiusiųjų buvo anonimiški.

 

 Trys pagrindiniai dirbtinio intelekto ekspertai parėmė laišką: AI mokslininkas Stuartas Russellas ir Yoshua Bengio bei Geoffrey Hintonas, kurie yra vadinamieji AI krikštatėviai dėl savo ankstyvų proveržio tyrimų. Hintonas paliko „Google“ praėjusiais metais, kad galėtų laisviau aptarti technologijos riziką.

 

 Hintonas ir kiti pastaraisiais metais skambino pavojaus varpais dėl būdų, kaip dirbtinis intelektas gali pakenkti žmonijai. Kai kurie dirbtinio intelekto tyrinėtojai mano, kad ši technologija gali tapti nekontroliuojama ir tapti tokia pat pavojinga, kaip pandemijos ir branduolinis karas. Kiti nerimauja, tačiau mano, kad dirbtinis intelektas turėtų būti labiau reguliuojamas.

 

 OpenAI antradienį atsakė į laišką, kad sutinka, kad turėtų būti vyriausybės reguliavimas.

 

 „Didžiuojamės savo pasiekimais, teikiančiais pačias pajėgiausias ir saugiausias AI sistemas, ir tikime savo moksliniu požiūriu į rizikos mažinimą“, – sakė „OpenAI“ atstovė. „Sutinkame, kad griežtos diskusijos yra labai svarbios, atsižvelgiant į šios technologijos reikšmę, ir toliau bendradarbiausime su vyriausybėmis, pilietine visuomene ir kitomis bendruomenėmis visame pasaulyje."

 

 „DeepMind“ ir „Anthropic“, kurią remia Amazon.com, antradienį iš karto nepateikė reakcijos dėl prašymų komentuoti.

 

 2015 m. įkurtas startuolis „OpenAI“ 2022 m. išleido „ChatGPT“ visuomenei. Pokalbių robotas tapo vienu iš labiausiai virusinių dirbtinio intelekto produktų, padėdamas „OpenAI“ perkelti į milijardus dolerių kainuojančią įmonę. Samas Altmanas, OpenAI vadovas ir vienas iš AI revoliucijos architektų, pareiškė norintis, kad ši technologija būtų kuriama saugiai.

 

 „News Corp“, „The Wall Street Journal“ savininkas, bendradarbiauja su „OpenAI“ turinio licencijavimo srityje.

 

 Antradienį pasirašytuose laiškuose teigiama, kad esami ir buvę darbuotojai yra vieni iš nedaugelio žmonių, galinčių patraukti korporacijas tapti atskaitingomis, nes vyriausybinės institucijos dirbtinio intelekto kompanijoms nevykdo plačios priežiūros.

 

 Jie teigė, kad jų susirūpinimą kelia tai, kad žmonės gali prarasti autonominių AI sistemų kontrolę, o tai savo ruožtu gali priversti žmones išnykti.

 

 Pasirašiusieji taip pat prašo įmonių leisti darbuotojams anonimiškai pranešti apie susirūpinimą, nekeršyti informatoriams ir neversti jų pasirašyti susitarimų, kurie galėtų juos nutildyti. Jie nori, kad dirbtinio intelekto įmonės būtų skaidresnės ir daugiau dėmesio skirtų apsaugos priemonėms.

 

 „OpenAI“ antradienį pranešė, kad turi anoniminę vientisumo karštąją liniją ir neišleidžia technologijų, kol nesukuria būtinų apsaugos priemonių.

 

 Buvęs „OpenAI“ darbuotojas Danielis Kokotajlo, pasirašęs laišką, teigė, kad įmonės neatsižvelgia į AI riziką, lenktyniaujant, kuriant šią technologiją.

 

 „Nusprendžiau palikti OpenAI, nes praradau viltį, kad jie elgsis atsakingai, ypač siekdami dirbtinio bendro intelekto“, – sakė jis.

 

 „Jie ir kiti pasirinko „greitai judėk ir sulaužyk dalykus“ metodą, o tai yra priešinga tam, ko reikia tokiai galingai ir menkai suprantamai technologijai.“ [1]

 

1. AI Workers Fear Voicing Concerns --- Group says OpenAI, others need to create reporting channels for employees. Lukpat, Alyssa.  Wall Street Journal, Eastern edition; New York, N.Y.. 05 June 2024: B.1.

AI Workers Fear Voicing Concerns --- Group says OpenAI, others need to create reporting channels for employees


"A group of employees in the artificial-intelligence industry said they can't voice concerns about AI's threat to humanity because of confidentiality agreements, a lack of whistleblower protections and the fear of retaliation.

In a letter released Tuesday, more than a dozen current and former employees of OpenAI, Google's DeepMind and Anthropic said AI companies need to create reporting channels for employees to safely voice concerns within their companies and to the public. They said confidentiality agreements block them from publicly discussing issues.

"The people with the most knowledge about how frontier AI systems work and the risks related to their deployment are not fully free to speak," said former OpenAI employee William Saunders, who signed the letter.

In addition to Saunders, six other former OpenAI employees signed the letter. Four current OpenAI employees and one former and one current employee from Google's AI research lab DeepMind also signed their names. Six of the signees were anonymous.

Three leading AI experts endorsed the letter: AI scientist Stuart Russell and Yoshua Bengio and Geoffrey Hinton, who are so-called godfathers of AI because of their early breakthrough research. Hinton left Google last year so he could more freely discuss the risks of the technology.

Hinton and others have been sounding the alarm in recent years over the ways AI could harm humanity. Some AI researchers believe the technology could grow out of control and become as dangerous as pandemics and nuclear war. Others are more tempered in their concerns but believe AI should be more regulated.

OpenAI said in response to the letter Tuesday that it agrees there should be government regulation.

"We're proud of our track record providing the most capable and safest AI systems and believe in our scientific approach to addressing risk," an OpenAI spokeswoman said. "We agree that rigorous debate is crucial given the significance of this technology and we'll continue to engage with governments, civil society and other communities around the world."

DeepMind and Anthropic, which is backed by Amazon.com, didn't immediately return requests for comment Tuesday.

OpenAI, a startup founded in 2015, released ChatGPT to the public in 2022. The chatbot became one of the most viral AI products, helping vault OpenAI to a multibillion-dollar company. Sam Altman, OpenAI's leader and one of the architects of the AI revolution, has said he wants the technology to be developed safely.

News Corp, owner of The Wall Street Journal, has a content-licensing partnership with OpenAI.

The letter signees on Tuesday said current and former employees are among the few people who can hold the corporations accountable because there isn't broad government oversight of AI companies. 

They said one of their concerns is that humans could lose control of autonomous AI systems that could in turn make people go extinct.

The signees are also asking companies to let employees anonymously report concerns, to not retaliate against whistleblowers and to not make them sign agreements that could silence them. They want AI companies to be more transparent and to focus more on safeguards.

OpenAI said Tuesday it has an anonymous integrity hotline and that it doesn't release technology until it has created necessary safeguards.

Former OpenAI employee Daniel Kokotajlo, who signed the letter, said companies are disregarding the risks of AI in their race to develop the technology.

"I decided to leave OpenAI because I lost hope that they would act responsibly, particularly as they pursue artificial general intelligence," he said. 

"They and others have bought into the 'move fast and break things' approach and that is the opposite of what is needed for technology this powerful and this poorly understood."" [1]

1. AI Workers Fear Voicing Concerns --- Group says OpenAI, others need to create reporting channels for employees. Lukpat, Alyssa.  Wall Street Journal, Eastern edition; New York, N.Y.. 05 June 2024: B.1.

 

Kaip Lietuvos verslas gali konkuruoti pasaulyje, turėdamas tokias, atsiprašant, "kadrų kalves"?

"Kasmetiniame pasauliniame universitetų reitinge „QS World University Rankings 2025“ Vilniaus universitetas (VU) šoktelėjo per 34 pozicijas ir atsidūrė 439 vietoje, pagerėjo ir Vytauto Didžiojo (VDU) bei Kauno technologijos (KTU) universitetų pozicijos.

„Džiugina tai, kad padarėme ryškią pažangą absolventų įsidarbinimo rodikliuose bei išskirtinai gerai įvertintas studentų ir akademinio personalo santykis. Tai parodo, kad mūsų įsipareigojimas kurti žmogišką ir šiltą santykį bendruomenėje pamatuojamas ir reitingų sėkme“, – VU pranešime cituojamas universiteto rektorius prof. Rimvydas Petrauskas.

Pernai VU reitinge užėmė 473 vietą, metus prieš tai – 400.

Pranešime teigiama, kad gerėjo VU absolventų įsidarbinimo rodiklis, lyginant su pernai metų duomenimis, jis kilo per 220 pozicijų iki 105 vietos. Šis rodiklis atspindi institucijų gebėjimą užtikrinti aukštą absolventų įsidarbinimo lygį ir jų įtaką savo profesinėse srityse. 

Lyginant su pernai, reitinge kiek pakilo VDU ir KTU reitingai. VDU reitinguojamas tarp 741–750 pozicijų, pernai – 801–850.

„Be galo džiaugiamės šiuo universiteto įvertinimu. Pasaulinė universitetų konkurencija dramatiškai auga, o šiemet QS reitinge apskritai buvo vertinama gerokai daugiau institucijų nei praėjusiais metais, (...) dėl to VDU pakilimas reitinge yra itin reikšmingas pasiekimas“, – VDU pranešime cituojamas universiteto rektorius prof. Juozas Augutis.

KTU šiemet daugiau nei 50-čia pozicijų pagerino savo reitingą ir yra tarp 751–760 pasaulio universitetų, pernai – 801–850.

„Kasmet konkurencija su stipriais universitetais iš viso pasaulio tik didėja, o taip pat į reitinguojamųjų sąrašus per metus patenka mažiausiai 20 naujų, stiprių institucijų. Tad mažų šalių ir sąlyginai nedideliems universitetams iššūkiu tampa net ir siekiai išlaikyti turimas pozicijas“, – KTU pranešime cituojama Strategijos ir kokybės valdymo direktorė dr. Lolita Jurkšienė.

Reitinguose nukrito Vilnius TECH arba Vilniaus Gedimino technikos universitetas, pernai buvęs tarp 801–850, šiemet universitetas reitinguojamas tarp 851–900 pozicijų.

Tuo metu Mykolo Romerio universitetas išlaiko tas pačias pozicijas ir užima vietą tarp geriausiai įvertintų 1001–1200 pasaulio universitetų.

Sudarant QS pasaulio universitetų reitingus, aukštojo mokslo institucijos iš viso vertinamos pagal aštuonis kriterijus: akademinę reputaciją, reputaciją tarp darbdavių, mokslinių darbų citavimo rodiklį, studentų ir dėstytojų skaičiaus santykį padaliniuose, užsienio dėstytojų dalį, universiteto tarptautinį tyrimų tinklą, užsienio studentų skaičių ir tvarumą.

Šiųmetiniuose QS reitinguose įvertinta 5.663 institucijų iš 106 pasaulio valstybių, iš jų 21 institucija reitinguotos pirmąkart.

Aukščiausias pozicijas šiame reitinge kaip įprastai užima JAV ir Jungtinės Karalystės aukštosios mokyklos: Masačusetso technologijos institutas, Londono imperatoriškasis koledžas, Oksfordo, Harvardo, Kembridžo ir Stanfordo universitetai."