On 16 October 2024, the Alibaba Group’s international commerce division introduced Marco MT, a self-developed large language model (LLM) specifically tailored for translation.
This technology, essential for generative artificial intelligence (GenAI) applications, promises to enhance communication in cross-border e-commerce and everyday interactions.
Zhang Kaifu, the vice president and head of the AI initiative at Alibaba International Digital Commerce Group (AIDC), stated,
“Marco MT will seamlessly integrate into Alibaba International’s overarching business strategies, fortifying the operational backbone of our diverse cross-border e-commerce platforms such as AliExpress, Lazada, Trendyol, and more.”
This launch comes as the company seeks to remain competitive amidst the global expansion of budget-shopping platforms like Shein and Temu.
The functionality of Marco MT extends to integrating seamlessly with systems that manage product listings, customer service, and search capabilities.
Merchants can leverage the LLM to translate product details — titles, descriptions, and image captions — into the languages of their target markets.
It also handles the translation of search keywords and dialogues.
Currently, the model, powered by Alibaba's own large language model Qwen, supports translations in 15 languages, including Chinese, English, French, German, Italian, Japanese, and Korean.
Zhang expressed plans to broaden language support, improve translation quality, and reduce usage costs, stating,
“We plan to expand the number of languages supported, further enhance translation quality across a broader range of scenarios, and consistently reduce the cost of usage for our users,” indicating a commitment to enhancing user experience.
Zhang Kaifu, the vice president and head of AI initiatives at Alibaba International Digital Commerce Group (AIDC).
AIDC has ambitious plans to extend Marco MT’s capabilities beyond Alibaba's platforms, aiming to benefit the global merchant community and individual users.
This reflects Alibaba's recognition of the vast potential in the cross-border e-commerce market, as the company navigates challenges posed by competitors.
Zhang observed that existing translation tools often struggle with “navigating the intricacies of culturally nuanced and idiom-laden expressions,” which has driven the development of Marco MT.
According to Alibaba, Marco MT has already surpassed leading AI translation competitors, including Google and DeepL, based on evaluations from the Flores machine-translation benchmark.
Zhang noted that Marco MT leverages deep cultural insights and an understanding of commercial regulations across various regions, enabling it to deliver contextually accurate translations.
For instance, while interpreting the popular Chinese phrase "YYDS," which means “eternal god” or “the greatest of all time,” Marco MT translated it as “top-notch,” whereas OpenAI's GPT-4 rendered it as “eternal.”
In contrast, Google’s service failed to comprehend the phrase altogether, showcasing the nuanced understanding Marco MT brings to the table.
Since the launch of its initial AI translation tool last year, Alibaba has reported significant uptake, with over 500,000 merchants already utilising the service.
The translation tool is particularly valuable for sellers aiming to create product listings in the language of their target markets.
Zhang emphasised the importance of contextual translation, stating,
“The idea is that we want this AI tool to help the bottom line of the merchants because if the merchants are doing well, the platform will be doing well.”
He recounted an example where a literal translation of a colloquial Chinese term for a slipper would likely deter English-speaking consumers.
“The updated translation engine is going to make Double 11 a better experience for consumers because of more authentic expression,” Zhang explained, referring to the shopping festival set for 11 November.
Alibaba's international division has demonstrated robust growth, with a 32% revenue increase to $4.03 billion for the quarter ending June 2024.
This contrasts sharply with the 1% drop in sales reported by Alibaba’s main e-commerce platforms, Taobao and Tmall, which are primarily focused on the Chinese market.
Analysts from Nomura predict that Alibaba's international revenue growth will slightly slow to 29% year-on-year in the quarter ending September, with operating losses narrowing.
The launch of Marco MT represents a strategic move for Alibaba as it aims to maintain its competitive edge in the rapidly evolving e-commerce landscape.
Zhang anticipates substantial demand for Marco MT from European and American markets, alongside significant interest from emerging markets, where about half of the top 20 active AI tool users on Alibaba.com hail from developing countries.
As Chinese companies increasingly seek growth opportunities beyond their borders, tools like Marco MT could prove crucial for enhancing their market presence and operational efficiency.
Kraken confirmed receiving Mt. Gox’s BTC and BCH funds for creditors, expecting payouts in 7-14 days. Mt. Gox transferred $5.8 billion in Bitcoin to undisclosed addresses, affecting market dynamics with Bitcoin’s price briefly dropping to $63,500.
ZeZhengMt Gox starts moving Bitcoin, signalling potential repayments to creditors.
Huang BoPrivasea uses FHE technology to provide data privacy and security, leveraging AI and distributed network architecture to allow complex data processing and analysis while keeping the data fully encrypted.
JinseFinanceThe core definition of DePIN is to bring real-world users into Web3 and achieve low cost and socialization through decentralized networks.
JinseFinanceAs an emerging technology, decentralized sorter aims to optimize the transaction sorting process of the blockchain network in a decentralized manner to improve transaction efficiency, reduce costs, and solve the MEV problem at the same time.
JinseFinanceSenator Rubio warns BRICS expansion threatens SWIFT system, US sanctions, and accuses China of backing Russia's actions in Ukraine.
Xu LinAt present, the ecological construction of Olympus DAO seems to be relatively simple, mainly relying on the profitability of the credit business sector.
JinseFinanceOne of the core features of Dencun’s upgrade is to significantly reduce the transaction costs of Ethereum L2, improve the transaction throughput of L2, and benefit the L2 ecosystem.
JinseFinanceThese long-awaited reimbursements will be financed through the sale of Mt. Gox's remaining assets, including 142,000 BTC, 143,000 BCH, and 69 billion yen (about $625 million)
DavinThe Mt Gox case had finally reached a settlement agreement back in 2021, and the BTC owed to creditors is ...
Bitcoinist