SlatorCon Zurich is now SOLD OUT — See you on October 4th!
Stay on top of how technology shapes the language, translation, and localization services industries
The Fairwork Project releases its 2023 report ranking the labor practices of 15 digital work platforms; some, including data-for-AI leader Appen, adopt the report's recommendations.
Aiding project management and communication, large language models can empower project managers with better efficiency and new ways of handling work.
New paper shows OpenAI’s generative model outperforms in English-Hindi translation, but requires more work in Kannada and Telugu; future research may compare Apple, Google.
Meta’s latest large language model, Llama 2, is trained on two trillion tokens. Yet, with 90% English training data, it “may not be suitable for use in other languages,” Meta warns.
Just two months after Bard’s limited release in May 2023, the large language model is now available in more than 40 languages —10x more than its original English, Japanese, and Korean.
The translation and localization resource, originally set to expire on June 30, 2023, takes a new form, but with much of the same information. Professionals praise Microsoft’s decision.
Google updates privacy policy. Spokesperson clarifies that the latest version not only covers Google Translate and other existing services now but also new AI co-pilots like Bard.
Publicis Groupe is the first advertising company to join the Coalition for Content Provenance and Authenticity’s steering committee, whose members include Adobe and Intel.
Meta AI introduces Voicebox, a multilingual, multitasking generative AI model for speech, trained on 50,000 hours of audio and 20 times faster than previous models.
At Apple’s annual online conference, BERT features in a new update to Create ML app, giving developers access to 27 languages that use Latin, Cyrillic, and certain Asian scripts.
Apple WWDC 2023: users can choose a way to be addressed in their language; apps adapt to gender-based preferences and part of speech for grammatical agreement.
AI startup SambaNova built 176-billion-parameter BLOOMChat on HuggingFace’s own large language model, BLOOM. Like competitors, translation abilities are inconsistent.
Microsoft and Amazon make similar new machine translation feature announcements within a day of each other, both intended for enterprise users.
Meta says its new system, which offers text-to-speech and speech-to-text for more than 1,100 languages, has half the word error rate of Whisper. Critics say, Apples and oranges.
Zapier now allows users to transcribe and translate audio from 98 languages into English text. Developers experiment with song lyric extraction and subtitle creation.
Initially launched in a few select countries, Google has plans for its “creative, helpful collaborator” to support — and translate between — 40 languages, up from the current three.
Google releases the PaLM 2 large language model. Technical report accompanying the launch finds that the LLM outperforms Google Translate in machine translation across all locales.
Subject to “guardrails” and innovative watermarks, available only to authorized partners, to prevent misuse; the system delivers speech-to-speech translation with lip matching.
OpenAI enforces branding policies and sends messages with tight deadlines through a third party for companies using GPT as part of their name or product marks to change them.
Since its launch in 2009, Microsoft’s multilingual online dictionary has become a “go-to” resource for IT translation; translators doubt the long-term usefulness of legacy terminology.
Slator Weekly: Join over 15,800 subscribers and get the latest language industry intelligence every Friday
Tool Box: Join 10,000 subscribers for your monthly linguist technology update.
Your information will not be shared with third parties. No Spam.
This will close in 0 seconds