DEV Community: ITransBook

What Exactly Is the China-U.S. AI Path Debate Really About?

ITransBook — Tue, 15 Sep 2026 02:15:59 +0000

Currently, global AI is at a critical juncture of accelerating technological paradigm shifts and geopolitical restructuring. The debate between open-source and closed-source paths in China and the U.S. is not merely a divergence in technological evolution paths in the ordinary sense, but a strategic extension of the China-U.S. geopolitical tech rivalry into underlying technological architectures, industrial control, and global rule-making power. The clash of these two technological paths will profoundly reshape the underlying logic of the global intelligence era and the geopolitical competition landscape.

In recent years, Silicon Valley giants represented by OpenAI, Anthropic, and Google have built a "technology tower" centered on closed-source business models, leveraging massive capital investment and supercomputing clusters. Meanwhile, open-source forces led by Chinese tech companies (such as Alibaba's Qwen series, DeepSeek, and Zhipu GLM) as well as some open factions in the U.S. (such as Meta's Llama series) have risen rapidly through efficient algorithm architectures and open ecosystems.

Overall, top U.S. AI labs adhering to the closed-source path is essentially a deep entanglement of Washington's desire to maintain technological hegemony with Silicon Valley's high-investment, high-compute, high-premium capital logic. The American closed-source model, represented by OpenAI's latest reasoning model GPT-5.6 Sol, Anthropic's Claude Opus 5, and Google's Gemini 3.5/3.6 series, relies on computing walls built from tens of thousands of top-tier GPUs and their data engineering, attempting to monopolize frontier exploration. The U.S. encapsulates model capabilities in cloud APIs (Application Programming Interfaces), on the one hand creating a highly controllable commercial monetization loop, and on the other turning them into a geopolitical tool. By bundling "national security" with closed-source controls, Washington has formed a "security narrative," arguing that cutting-edge AI models pose "potential national security risks" and must be centrally monopolized by a handful of Western companies. This reliance on high API subscription fees and cloud binding not only raises the financial threshold for small and medium-sized enterprises worldwide to access high-end AI, but also traps countries lacking native computing support in severe "digital dependency" and even strategic passivity.

Facing U.S. chip export controls and computing containment, China's tech forces have taken open source as a strategic breakthrough, exploring an asymmetric game path of extreme algorithmic efficiency and large-scale industrial deployment. In terms of algorithms, DeepSeek achieved performance comparable to Silicon Valley's closed-source peak at lower cost, directly breaking the iron law that "breakthroughs require billions of dollars in high-energy consumption stacking." In terms of ecosystem, large models like Alibaba's Qwen 2.5 rank first globally in open-source community downloads, making them one of the most favored foundational bases for developers worldwide. A report by the U.S.-China Economic and Security Review Commission (USCC) under Congress points out that China has two different competitive advantages: "digital circulation" and "physical circulation." Open-source models lower the barrier to AI deployment, quickly entering massive physical scenarios such as manufacturing, logistics, and robotics, and the data and demand generated by these scenarios in turn drive continuous model iteration. The "flywheel effect" of efficient algorithms, open ecosystems, and industrial validation reinforcing each other gives China's AI strong survival and catch-up capabilities even under strict computing blockades.

The open vs. closed source debate between China and the U.S. has accelerated the "rebalancing" of global tech geopolitics, providing strategic opportunities for the "Global South" and neutral economies (such as Southeast Asia, the Middle East, Europe, and Latin America) to escape the squeeze between major powers and reshape their technological sovereignty. For European and Middle Eastern countries, transferring core industrial or national government data to Silicon Valley giants' cloud servers is an unacceptable compliance and security risk; developing hundred-billion-parameter models entirely from scratch faces astronomical resource thresholds. Under these circumstances, high-performance open-source models represented by Qwen, DeepSeek, and Llama have become the "digital new infrastructure" for these countries to build localized AI infrastructure. After fine-tuning open-source bases with local languages and deploying them privately, local institutions in France, Singapore, Saudi Arabia, and other countries can obtain highly cost-effective, autonomous and controllable models. Open source is transforming AI from a "strategic embargoed good" controlled by a few countries into global "public infrastructure."

The open vs. closed source debate has also evolved into a contest over global AI safety, governance, and compliance paradigms. The closed-source camp has long treated the "black box model" as the best defense against malicious use of models (creating bioweapons, cyberattacks, etc.), using this as a justification for maintaining blockades and imposing high-threshold administrative reviews. However, as open-source model capabilities reach critical thresholds, the narrative of achieving security through obscurity is being deconstructed by academia and the international community. More and more international standardization organizations and legislative bodies recognize that excessive closed-source review can easily become a barrier to geopolitical monopoly, and that "transparent, auditable, multi-stakeholder" open-source governance is the best way to build globally trustworthy AI systems.

Looking ahead, the global AI landscape will not see a one-sided "complete blockade" or "full replacement," but rather a dual-track competition. The closed-source path represented by some U.S. labs will continue to rely on absolute scale of capital and computing to conduct high-risk exploration on the unknown frontier toward Artificial General Intelligence (AGI); while the open path represented by China and the global open-source community, relying on high cost-effectiveness, good scenario matching, and a foundation of millions of developers, will become the main engine driving AI technology into all industries. The strategic game between China and the U.S. on open vs. closed source paths has objectively broken through barriers of technological monopoly and accelerated the pace of global technology inclusiveness. A multipolar, resilient, and multi-stakeholder global AI new order is rapidly taking shape.

This article was translated using the translation software ITransBook. Readers interested in AI translation software can visit www.itransbook.com for more information.

Institution: The AI boom is masking the rising risks in the US economy

ITransBook — Tue, 08 Sep 2026 10:32:48 +0000

Nomura, an international financial institution, said in a recently released research report that the large-scale development of artificial intelligence (AI) has become an important engine of US economic growth, but it is also putting pressure on the US economy in multiple ways. If AI development suffers a setback, it may expose the rising risk premium in the United States.

The report said the risks brought by the US AI boom include rising chip and electricity prices pushing inflation higher, a surge in imports of chips and technology equipment widening the trade deficit, and large cloud computing companies issuing large amounts of bonds pushing up US Treasury yields.

In addition, due to net capital inflows and abnormal capital gains from the AI-driven rise in the US stock market, the net liability position of US international investment (that is, dollar assets held by the rest of the world) has increased significantly. According to calculations by Nomura analysts, the ratio of the net liability position of US international investment to the total net asset position of all net creditor countries has soared to 80%.

The report pointed out that because US stock market valuations are high and economic fundamentals are weakening, if AI development suffers a setback, it may trigger a major adjustment in US capital markets. Given foreign investors' large exposure to US stocks and the leverage and circular financing within the AI ecosystem, such an adjustment could turn into a "global risk-off event."

The report said some views hold that dollar assets are "irreplaceable" due to the large size of the US economy, the depth and liquidity of its capital markets, and the status of the dollar as the global reserve currency. However, the pillars on which the dominant position of dollar assets rests are weakening, and external trust in US fiscal and trade policies, security assurance policies, and the independence of the Federal Reserve has all declined.

The report pointed out that the United States still relies heavily on foreign capital inflows to make up for its continuously expanding fiscal and current account deficits, but the massive twin deficits cannot be sustained. The mounting debt burden is like a "ticking time bomb," and the related pressure is accumulating.

The report argued that given the surge in the net liability position of US international investment and the high concentration of other countries' international investment assets in the United States, moderate de-risking by foreign investors would be "enough to trigger a depreciation of the dollar."

This article was translated using the translation software ITransBook. Readers interested in AI translation software can visit www.itransbook.com for more information.

Robots don't need visas

ITransBook — Tue, 01 Sep 2026 14:44:37 +0000

This August, Beijing's E-Town is buzzing. The 2026 World Robot Conference held its first-ever procurement day, with buyers coming from Southeast Asia, the Middle East, and Latin America. A Thai farmer wanted to check out crop-protection robots, a Saudi contractor was eyeing inspection robot dogs, and a Latin American plantation owner was asking about the price of harvesting robots. Their questions were all down-to-earth: Can this machine help me get the work done?

Around the same time, Washington slapped a new label on robots: a threat to national security. That same day, the U.S. Federal Communications Commission updated its list of regulated equipment, adding new Chinese-made humanoid robots and quadruped robots. Of course, being cautious about new technology is understandable. When the spinning jenny first appeared, weavers feared the iron contraption would eat people and stormed the factories to smash it to pieces; when the camera first came out, plenty of people ducked away from the lens, worried that a flash would steal their souls. Every invention that changed the world has gone through a stage of being doubted. Security concerns are real and deserve serious attention.
But caution and shutting the door are two different things. Robots have eyes, legs, and internet connectivity—but by those three criteria, smartphones check all the boxes, electric cars check all the boxes, and even the robot vacuum cleaner at home checks all the boxes. A better approach to new things might be like airport security: every passenger gets screened, but the airport doesn't refuse anyone boarding based on nationality. Giving robots a check-up—setting standards, issuing certifications, and auditing data—is far more effective and reassuring than simply rejecting them at the door.

Besides, the world's need for robots right now may be greater than many imagine. Much of the globe is aging, and fewer young people are around to work. Factories in Vietnam can't fill their shifts, construction sites in the Middle East hit 50 degrees Celsius in summer, and plantations in Latin America are short of hands every harvest season. Robots aren't here to steal jobs; more often, they're here to fill the gaps that no one wants to fill—or that no one is left to fill. A robot sent down a mine shaft replaces a person risking their life to go down there; a robot working in a high-temperature workshop takes over the labor that humans can't endure.

China's experience is a case in point. China's robotics industry chain wasn't built behind closed doors; it was forged in openness. Over the decades, foreign brands set up factories in China, global components gathered here, and fierce competition pushed Chinese companies to learn, step by step, how to make robot joints and brains—and to make them better and better. Today, Chinese companies account for more than half of all humanoid robot models worldwide, and for every ten quadruped robots sold, nearly seven come from China. More importantly, costs have come down, and automation is no longer a privilege of wealthy countries—factories and farms in developing nations can afford it too. Openness didn't crush China's industry chain; it fed it into the most complete one on the planet. China is willing to share that experience, along with these robots, with the world.

At the end of the day, robots are by nature a product of global cooperation. Japan's precision joints, Europe's motors, America's chips, China's assembled machines and application scenarios—none of it works without the others. When a river meets a dam, the water doesn't disappear; it just goes around, and someone eventually has to pay the cost of that detour. Maintaining the network of cooperation serves everyone's interests.

Robots don't need visas. They just want to go where people need them to work. Security can be safeguarded through standards, and competition can be about who builds better, cheaper, and more capable solutions—and in that kind of competition, the whole world wins. Next time a robot shows up at the door, maybe the real question isn't where it came from, but what it can help us do.

This article was translated using the translation software ITransBook. Readers interested in AI translation software can visit www.itransbook.com for more information.

The Second World Humanoid Robot Sports Games Opens in Beijing

ITransBook — Mon, 24 Aug 2026 07:29:40 +0000

On August 22, 2026, the opening ceremony of the Second World Humanoid Robot Sports Games was held at the National Speed Skating Oval in Beijing.

On August 22, a robot freestyle fighting demonstration took place at the opening ceremony.

On August 22, before the opening ceremony, robots performed by playing soccer.On

August 22, before the opening ceremony, robots performed by playing tennis.

This article was translated using the translation software ITransBook. Readers interested in AI translation software can visit www.itransbook.com for more information.

How to Fix the 'Unexpected exception when initializing view part Fragment Matches' Error in Trados 2021 (Fragment Matches) – A 3-Step Solution

ITransBook — Mon, 17 Aug 2026 14:26:56 +0000

When using Trados 2021 or other versions, if you encounter the error dialog "Unexpected exception when initializing view part Fragment Matches’ could not be started.", translators may not know how to resolve it. I will first introduce the solution I found as follows:
1.Enter the following directory on your computer:
C:\Users\%USERNAME%\AppData\Roaming\SDL\SDL Trados Studio\Studio16
2.Delete the following files
BaseSettings.xml
Plugincache.xml
UserSettings.xml
3.Restart the Trados software
This error is generally caused by abnormal settings during the use of Trados, and translators often encounter it. If you encounter this error, you can resolve it using this method.

By the way, I will introduce a useful translation software, ITransBook. It not only supports local AI models but also mainstream AI APIs. Except that the ITransBook file translation feature requires a fee, AI conversation, text translation, word translation, and shortcut key translation are all free. Translators can use its word translation and shortcut key translation features in CAT software such as Trados, memoQ, and Wordfast. After becoming a paid member of ITransBook, translators can use ITransBook to translate Trados and memoQ format files. The ITransBook download website is www.itransbook.com. You can also search for the keyword ITransBook in the Microsoft Store to download it.

How to Fix the 'Interfaces.ITelemetryService could not be started' Error in Trados 2021 (ITelemetryService) – A 3-Step Solution

ITransBook — Wed, 05 Aug 2026 14:47:35 +0000

When using Trados 2021 or other versions, if you encounter the error dialog "Service ‘Sdl.Desktop.Telemetry.Interfaces.ITelemetryService’ could not be started.", translators may not know how to resolve it. I will first introduce the solution I found as follows:
1.Enter the following directory on your computer:
C:\Users\%USERNAME%\AppData\Roaming\SDL\SDL Trados Studio\Studio16
2.Delete the following files
BaseSettings.xml
Plugincache.xml
UserSettings.xml
3.Restart the Trados software
This error is generally caused by abnormal settings during the use of Trados, and translators often encounter it. If you encounter this error, you can resolve it using this method.

How to Fix the 'A pane’s docked state cannot be determined' Error in Trados 2021 (DockAreaPane) – A 3-Step Solution

ITransBook — Fri, 31 Jul 2026 00:44:31 +0000

When using Trados 2021 or other versions, if you encounter the error dialog "A pane’s docked state cannot be determined until it has been associated with a DockAreaPane", translators may not know how to resolve it. I will first introduce the solution I found as follows:
1.Enter the following directory on your computer:
C:\Users\%USERNAME%\AppData\Roaming\SDL\SDL Trados Studio\Studio16
2.Delete the following files
BaseSettings.xml
Plugincache.xml
UserSettings.xml
3.Restart the Trados software
This error is generally caused by abnormal settings during the use of Trados, and translators often encounter it. If you encounter this error, you can resolve it using this method.

How to Fix the 'Requested Service Is Not Available' Error in Trados 2021 (IFilesViewService) – A 3-Step Solution

ITransBook — Sun, 26 Jul 2026 13:21:19 +0000

When using Trados 2021 or other versions, if you encounter the error dialog "The requested service is not available: Sdl.TranslationStudio.Api.ProjectManagement.IFilesViewService", translators may not know how to resolve it. I will first introduce the solution I found as follows:
1.Enter the following directory on your computer:
C:\Users\%USERNAME%\AppData\Roaming\SDL\SDL Trados Studio\Studio16
2.Delete the following files
BaseSettings.xml
Plugincache.xml
UserSettings.xml
3.Restart the Trados software
This error is generally caused by abnormal settings during the use of Trados, and translators often encounter it. If you encounter this error, you can resolve it using this method.

A common error I encounter in Trados during translation, and how I resolve it.

ITransBook — Wed, 15 Jul 2026 15:15:55 +0000

Recently translated an Excel file with nearly 100,000 rows, as shown in the figure.

After adding it to Trados 2021, opening the translation file prompted an out-of-memory error.

After adding it to Trados 2024, opening the translation file still showed an out-of-memory error.

Fortunately, I have ITransBook, which runs this 100,000-row file without any pressure, as shown in the figure.

After translation, as shown in the figure below:

Attempt to open this translation file with Trados 2024.

You can see that the target column contains Deepseek translations, which are very accurate, with no misalignment or misalignment issues.

By the way, I would like to introduce ITransBook. It not only supports local AI models but also mainstream AI APIs. Except for the ITransBook file translation feature, which requires payment, AI chat, text translation, hover translation, and shortcut translation are all free. Translators can use its hover translation and shortcut translation features in CAT software such as Trados, memoQ, and Wordfast. After becoming a paid member of ITransBook, translators can use ITransBook to translate Trados and memoQ format files. The ITransBook download website is www.itransbook.com, and you can also search for the ITransBook keyword in the Microsoft App Store to download it.

Translator's Self-Help: Fix Trados Startup Errors in 4 Steps

ITransBook — Thu, 09 Jul 2026 13:57:58 +0000

When using Trados 2021 or other versions, encountering the error box "Unable to start service 'Sdl.TranslationStudio.Api.ProjectManagement.IProjectManagementService'", translators often do not know how to resolve it. The solution I found is presented below:

1.Close Trados Studio or the error message
2.Go to C:\Users\%USERNAME%\Documents\Studio 2021\Projects on your system
3.Delete the projects.xml file
4.Start Trados Studio

This error is generally caused by having too many projects in Trados, and translators often encounter it. If you encounter this error, you can resolve it by following this method.

By the way, I would like to introduce a useful translation software, ITransBook. It not only supports local AI models but also mainstream AI APIs. Except for the ITransBook file translation feature, which requires payment, AI chat, text translation, hover translation, and shortcut translation are all free. Translators can use its hover translation and shortcut translation features in CAT software such as Trados, memoQ, and Wordfast. After becoming a paid member of ITransBook, translators can use ITransBook to translate Trados and memoQ format files. The ITransBook download website is www.itransbook.com, and you can also search for the ITransBook keyword in the Microsoft App Store to download it.

Translator's Self-Help: Fix Trados Startup Errors in 3 Steps

ITransBook — Thu, 09 Jul 2026 08:21:57 +0000

When using Trados 2021 or other versions, if you encounter the error box "Unexpected exception when getting view part instance of view part translation result", and don’t know how to solve it, here is the solution I found:
1.Enter the following directory on your computer:
C:\Users\%USERNAME%\AppData\Roaming\SDL\SDL Trados Studio\Studio16
2.Delete the following files
BaseSettings.xml
Plugincache.xml
UserSettings.xml
3.Restart the Trados software
This error is generally caused by abnormal settings during the use of Trados, and translators often encounter it. If you encounter this error, you can solve it by following this method.

DeepL vs Claude: Which Delivers More Accurate Chinese-English Translations? A Rigorous Side-by-Side Comparison

ITransBook — Thu, 02 Jul 2026 07:56:23 +0000

When translating a file, the first thing we think of is that the translation must be very accurate. There are many file translation service providers on the market, including not only human translation services but also machine translation services, and in the last 2-3 years, AI translation services as well. Today, I will compare DeepL and Claude translations to see which one is more accurate.

The following is my process of comparing DeepL and Claude translations using ITransBook.
1.Open our ITransBook translation software.

2.Click the File Translation icon on the left.
3.Click the + sign in the upper left corner to create a new project.

Enter "Comparison of DEEPLE and Claude engines" for the project name. Select "English (American)" for the source language, choose "Target (1:N)" for the target option, select "Chinese (Simplified, 简体中文)" for the target language, and choose DeepL and Claude as the translation engines.
4.Click "API Settings" on the right.

I have already purchased a credit package, so I don't need to configure the API keys for DeepL and Claude. I can choose the default model, and here I will select "claude-sonnet-4-6".

Click the "Save" button.

5.Drag the file to be translated into the file translation area.

This is a screenshot after the translation is completed.

Click the "More" button, then click the "Export multilingual document" button to export the comparison translation.

Check the three displayed options and click the "Export" button.

Click the "OK" button to proceed to the comparison translation folder.

We double-click the comparison translation file to open it and begin comparing the DeepL and Claude translations.

Both DeepL and Claude translated "token" incorrectly. On March 25, the China National Committee for Terminology in Science and Technology officially issued an announcement, preferentially recommending "词元" as the standard Chinese term for "token" in the field of artificial intelligence. In translating "companies need to have both human capital and in-house AI capabilities...", Claude's translation "需要同时具备" is more specific and accurate than DeepL's translation "需要...". For this passage, I personally think Claude's translation is better.

DeepL translated "firms across the economy" as "各行业企业", which is easy to understand, and "workers" as "员工", which fits the context better, superior to Claude translating "workers" as "劳动力". For "a lot of displacement", both DeepL and Claude translated it poorly, with DeepL's translation being further from the original meaning; it should be translated as "大量岗位变动" instead. For this passage, I personally think DeepL's translation is better.

For the translation of the word "narratives", DeepL's translation as "说辞" is better, carrying an emotional tone, but for the subsequent "just narrative", DeepL translated it as "叙事", showing an inconsistency in terminology translation. For these two passages, I personally think DeepL's translation is better.

For "reduce prices for customers", DeepL translated it as "为客户降低成本", while Claude translated it as "为客户降低价格". DeepL's translation is more accurate and flexible; the passage revolves around costs, so it should be reducing costs. For this passage, I personally think DeepL's translation is better.

Overall, DeepL was more accurate in translating this interview article, while Claude still has room for improvement in terms of word choice.