Google News Initiative’s Post

Google announced many things at their AI conference yesterday but the one that will have the most impact is Search 🔎 Videos and language translations are not everyday habits like search, so improving their search engine before others fully jump into this space is key. This video 👇🏾 shows you what to expect in phase 1. #googleai #search #generativeai

Search is Evolving: Beyond Keywords! Remember the days of typing out precise keywords for a simple search? Get ready for a revolution! 1. Rabbit R1: This innovative device goes beyond words. It uses voice commands and interacts with apps to take action for you. Book a flight? Done! Checkout online? No problem! 2. Google's Voice Search 2024: Google's not sitting still! They're constantly improving voice search with the following: Natural Language Understanding: Talk naturally and ask complex questions – voice search gets it! Multimodal Search: Combine voice with images or touch for an even more intuitive experience. Personalized Search: Get results tailored based on your preferences and context. The future of search is here, and it's interactive, personalized, and beyond just keywords! **What are your thoughts on this search evolution? Share in the comments! ** #search #futureofsearch #voicetech #innovation #rabbitR1 https://buff.ly/49U8gZY

In customer conversations I’m seeing new enterprise global content use cases made possible by Generative AI. 1. #HR L+D video dubbing (with cloning) – Employees are more used to video based training modules, and copyright issues associated with voice cloning aren’t prohibitive for this use case. 2. Global call center transcript summarization – Tell me the top three customer issues in English given this set of Korean transcripts. 3. #Contentmarketing content generation – Write me a blog post and a Google ad campaign in Japanese given this English product description. The key to making these use cases work well are: 1. An LLM with a balanced set of multilingual training data. 2. Context-specific multilingual data for that enterprise. This can usually be found anywhere that content localization is happening (e.g., in marketing). Curious to learn more? Book a meeting to chat with one of our solutions architects: https://loom.ly/e58iBLw #generativeai #ai #enterpriseai

Pause on this…. “TextFX is an AI experiment that uses Google's PaLM 2 large language model. These 10 tools are designed to expand the writing process by generating creative possibilities with text and language.” “Google’s TextFX includes 10 tools and is powered by the PaLM 2 large language model via the PALM API. Meant to aid in the creative process of rappers, writers, and other wordsmiths, it is part of Google Labs. Simile: Create a simile about a thing or concept. Explode: Break a word into similar-sounding phrases. Unexpect: Make a scene more unexpected and imaginative. Chain: Build a chain of semantically related items. POV: Evaluate a topic through different points of view. Alliteration: Curate topic-specific words that start with a chosen letter. Acronym: Create an acronym using the letters of a word. Fuse: Find intersections between two things Scene: Generate sensory details about a scene. Unfold: Slot a word into other words or phrases.”

Gen-AI is amazing for helping us improve our communications. We can all now craft those delicate emails, using some simple prompts. As the Developer on system ABC, create a polite email to the rest of the team, with the following points. * great presentation by Greg last week on changes within Tax regulation. * looking forward to the catchup on integration of our systems into SystemZ next week. * need some clarification on Alpha before we start spec'ing it out. And the response : Super. We get a nice long 5 paragraph email politely worded, but still requiring a little attention around the edges. But wait, hadn't you written the minimal information, required. And now you are asking AI to write a long version of this? which you're sending to your colleagues. So instead of reading 3 sentences, they need to read 5 paragraphs. Well, that's not a problem, because this company has 2 way translations. Yes, the staff can access summary versions of the emails. 5 paragraphs are automatically condensed to 3 lines. This is super, no harm no foul. The original message is written by a human in computer. The computer translates it to a human email. The email is sent and read by the #generativeai and translated back from human to computer. Finally the human then reads the computer. Ok, no harm no foul.... Hmmm, if I send that email to 10 people, it's doing 11 translations, or 11 GPT queries. no problem we are getting value from communicating politely to each other, and they are getting value by finding the salient points in emails. I'm confused.. Where's the value? 😉

For everyone asking! With the trilogue of the AI Act over, a few people asked: “When can we see a text?” Let us try to explain what typically happens with such text in before and after trilogue. By Malte Beyer-Katzenberger, "Technical teams of the Commission, the Council and the EP use a text file table called the four-column-document cutting the legal text into its component parts (recitals, articles, paragraphs and sub paragraphs). One column for the Commission proposal, one for the amendments by Council, one for those by EP and one for the compromise found. In between the trilogue meetings that attract media attention, the technical teams are tasked to agree on non-contentious elements. After the last trilogue with the “deal” the textual agreements found are introduced by the Council or EP secretariat. For some parts only oral agreements where found that first need to be translated into text. Disagreements on what was agreed may occur. Such disagreements need a space of confidence to be sorted out. After that, lawyer linguists look at the text and see that it is (a) proper English, (b) compliant with general drafting standards of EU legislation and (c) accompanied with annotations allowing for translation. Also in this process disagreements between institutions can occur and again there is a need for a private space of confidence to agree on a fin text. All that makes also for a text that is in a state of flux for a few weeks normally. In sum: There is no stable text for a while and there is still need to come to final agreement on textual formulations that truly represent the compromise in a confidential setting. This final text is then voted by both EP and Council and only then signed into law and published in @EURLex. And translations to all languages need to be done before publication. That takes a while too." For more information about the proposed AI Act, please follow Luca Bertuzzi here: https://lnkd.in/etr4ysZ2.

"The political agreement reached in trilogue does not in fact include a definitive legal text, nor is it set in stone. This makes upcoming “technical meetings,” which will finalize the text before it is officially adopted by both the Parliament and Council, incredibly important, especially considering some countries still do not consider it a done deal." Daniel Leufer. Some elements on the « clarifications » demanded by France on foundation models, around training data transparency and the systemic model's threshold. Please read the three quick lessons from the process written by Dan Nechita here: https://lnkd.in/e5Y-tda3. I am sure you can. Dan adds, "Good negotiators secure what they want long before getting to the negotiation table. Always think about the people you’re negotiating with. There are unsung heroes and unsung villains in every outcome." I agree with Peter Craddock that the proposed AI Act, which has not been adopted, "is far more to that in terms of legislation, from GDPR to intellectual property rules, with confidentiality requirements, ethics, labor law and unfair commercial practices in the mix among other things."

Our Enterprise Language Hub now offers access to the latest LLMs from Anthropic and Google. Integrate these powerful models into your translation workflows to make them quicker, more accurate, and context-sensitive. If you want to do more and build intelligent solutions for content generation, summarization, data analysis, or tailoring content to your style, you can create custom AI skills using these LLMs that are also available in our GenAI Portal. Check it out now ➡️ https://hubs.la/Q02xWfq80 Available #Anthropic models: • Claude-3-opus-20240229: Best for complex tasks, in-depth analysis, and handling long documents up to 150,000 words in English. • Claude-3-sonnet-20240229: Ideal for enterprise workloads and business applications requiring high-quality output. • Claude-3-haiku-20240307: Perfect for fast-paced, real-time applications and customer support tasks needing quick responses. • Claude 2.1: Optimal for processing and analyzing large documents up to 75,000 words in English. Available #Google models: • Palm2-chat-bison-002 and Palm2-chat-bison-32k-002, capable of processing 6,000 and 32,000 words, respectively, ensure safe and non-toxic output in over 100 languages. These models were developed specifically for multi-turn chats, ensuring user interactions remain seamless and efficient.

Voice Writer Update: Our latest feature now allows for grammar corrections while preserving formatting, including headers, lists, bold text, and more. Write with your voice and let the AI handle the details. Try voice writer here: https://lnkd.in/gz6riAdE

What makes BeyondWords the first choice of news publishers choosing an AI voice partner? 🤔 🎯A major advantage of choosing BeyondWords is our advanced pre-processing libraries. By processing text prior to speech synthesis we allow publishers to fine-tune pronunciations, expand on symbols, abbreviations and acronyms, and offer their audiences the best possible listening experience. Read more: https://lnkd.in/eu6XtcJe