Umit Ozaydin’s Post

I am yet to make any personal experiences with being "corrected" by AI, but this is what happens. [insert Surprised Pikachu face here] #translation #localization #xl8 #t9n #l10n #ArtificialIntelligence #AI #GenerativeArtificialIntelligence #GenerativeAI #GenAI #QualityAssurance #QuabityAssuance #CreedBratton #Useless #LITranslators

Today, it finally happened. Today 𝐈 𝐰𝐚𝐬 𝐜𝐨𝐫𝐫𝐞𝐜𝐭𝐞𝐝 𝐛𝐲 𝐚𝐧 𝐀𝐈. I knew it would happen sooner or later, but I hoped I'd take it better. Perhaps because I hoped it would be a 𝐬𝐥𝐢𝐠𝐡𝐭𝐥𝐲 𝐥𝐞𝐬𝐬 𝐬𝐭𝐮𝐩𝐢𝐝 AI proofreading me. The 🤡𝘤𝘭𝘪𝘦𝘯𝘵🤡 used an AI that: 🚀 Doesn't know the basic punctuation rules of French (the famous unbreakable space before certain punctuation marks) 🚀 Thinks that "P" is a vowel (therefore suggests a specific article for words starting with a vowel ➡ "𝐀𝐜𝐪𝐮𝐢𝐬𝐭𝐚 𝐮𝐧' 𝐩𝐫𝐨𝐛𝐚𝐛𝐢𝐥𝐢𝐭à" LMAO) 🚀 Detects non-existent spaces at the end of sentences 🚀 Suggests removing accents from words that need them 🚀 Proposes changes in prepositions at its own taste 🚀 Flags my non-existing word Zampastico (transcreation for Pawsome) but doesn't bat an eye at "Miaoravigioso" (transcreation for Purr-fect) (and the client didn't even bother thinking "Hey, there's a made up word in the source, 𝐦𝐚𝐲𝐛𝐞 it's fine in the target too?") 🤬 Three linguists in three different languages had to justify, to a PM who doesn't know any of these languages, why the flags raised by the AI were nothing but BS. 🤬 Three linguists who wasted precious minutes doing so (and a PM having to deal with useless BS) (And me, who spent an hour creating this post, but you get it, I needed to vent/share). 𝐈𝐧𝐬𝐮𝐥𝐭𝐢𝐧𝐠. I can't find any other words. My calm response, on the right. Oh, BTW, it's a 𝐡𝐮𝐠𝐞 𝐛𝐫𝐚𝐧𝐝, a leader in its sector for gaming apps and board games. Just to give you some context. Just great. #Gameloc #Localization #AI #GameOverForGreed

I don't work with translation agencies. I don't use generative AI. I don't *trust* generative AI, and that has nothing to do with a lack of knowledge: I studied LLMs at university over a decade ago. But this tale, expertly told by Chiara, explains exactly why I don't use generative AI, and advise my clients to not even think about checking the work of humans using generative AI. Naturally, they should also avoid translation agencies that lean heavily into generative AI and machine translation, too. #aiaintcomingforme #aiaintcomingforchiaraeither #litranslators #generativeai

English word of the year 2023? It's "hallucinate", a fairly common word, but with a new meaning and use (According to the Cambridge Dictionary lexicographers). Familiar with the new use? Slightly late, but I didn't know until today, so here it goes. As you might know before, Hallucinate means "to seem to see, hear, feel, or smell something that does not exist, usually because of a health condition or because you have taken a drug". Now it also is used in a very up-to date way, describing one of the problems with AI (Artificial Intelligence). Today you can use the word hallucinate like this: "When an artificial intelligence produces wrong information." The explanation for this new use of the word is interesting. "The Cambridge Dictionary team chose hallucinate as Word of the Year 2023 as it recognised that the new meaning gets to the heart of why people are talking about AI. Generative AI is a powerful tool, but one we’re all still learning how to interact with safely and effectively – this mean being aware of both its potential strengths and its current weaknesses.” And now you know too. Me, I thing it's fascinating how our languages and words are affected by time, trends and everything. And English is probably the language thats transforms faster than any other today. As so many non-natives use it on an everyday basis. More facts then. The word hallucinate is first used in English in the 1650s, coming from the Latin word alucinatus (later hallucinatus), past participle of alucinari "wander (in the mind), dream; talk unreasonably, ramble in thought," probably from the Greek word alyein. (source: Etymology online) #copywriter #wordsmatter The image? Well, it's an AI image, created with the prompt "An AI that hallucinates", by Wepik AI image generator.

Writing in the times of AI. One often overlooked but very important aspect of meaning acquisition in language is its two-mode semantics: reference and inference. Reference is how words (and other linguistic entities) latch on things in the outside world. The meaning of an apple is partially contained in our experience of that red (or any other color for that matter), round, plant-based object that we eat. Inference, on the other hand, is how the meaning of a word is encoded in its relationships with different words or linguistic entities. The meaning of an apple is related to the words 'fruit', 'red color', 'round', 'plant', 'food', etc. AI, and large language models, in particular, excel in the latter mode of semantics. They contain extremely large and complex networks of relations between linguistic entities. That is why they sound like very fluent speakers. They are very good 'word matchers' for the lack of a better term. When we write (or use language in general), we strive to engage the reader (or the communication receiver) in these two modes of meaning acquisition. We try to invoke experiences of real or imagined objects while at the same time taking advantage of the relationships of the vast network of words and sentences we learned throughout our lives and believe we share with other humans. As writers in order to thrive in the age of AI while taking advantage of such technology, we need to double down on making our writings more immersive and evocative while making use of the vast semantic networks provided by LLMs. In relation to brands and content in general this means that now all brands can sound fluent, but only the personal ones will stand out. In other words, only the brands with a human face will thrive and garner the sought-after engagement. I believe we already reached a saturation level of generic content and the need for differentiation and fine-grained personalization is more pressing than ever. As a closing remark, the current status quo reminds me of the beautiful opening paragraph in Dickens's masterpiece 'A Tale of Two Cities': "It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness, it was the epoch of belief, it was the epoch of incredulity, it was the season of light, it was the season of darkness, it was the spring of hope, it was the winter of despair."

Both Bard and Gemini are large language models (#LLMs) developed by #Google, but they have some key differences: #Bard: Originally called Meena, Bard is focused on conversational AI. This means it's designed to hold engaging and informative dialogues, answer your questions in a comprehensive and informative way, and even follow your instructions and complete your requests thoughtfully. #Gemini: Rebranded from Bard, Gemini is primarily focused on content creation. It excels at generating different creative text formats, like poems, code, scripts, musical pieces, email, letters, etc., based on your instructions and preferences. Bard: Bard excels at tasks like providing summaries of factual topics, translating languages, writing different kinds of creative content, and answering your questions in an informative way. However, it may not be as proficient in generating certain creative formats or understanding highly technical topics. Gemini: Gemini shines in generating various creative text formats, translating languages, writing different kinds of creative content, and understanding highly technical topics. However, it may not be as engaging or informative in conversational settings as Bard. Choose Bard if you need a conversational AI for dialogue, summaries, and creative writing. Choose Gemini (when it becomes available) if you need a content creation AI for generating specific text formats or understanding complex topics. It's important to note that both models are constantly evolving, and the lines between their capabilities may continue to blur. Choosing the right one depends on your specific needs and preferences. #AI #aiadvancements Shankar Sahai

Translation will now change forever. GPT-4o is 50% cheaper and 2x faster than GPT-4-turbo: And this is perhaps one of my favourite capabilities. If it becomes socially acceptable to hold a phone up to a stranger, this will make conversations when travelling super easy. My concerns sit around the future of: • Google translate • Interpreters • Duolingo My takeaway: I don't think you'll ever beat conversing with someone in their native language to build deep, meaningful relationships. But this does mean the Duolingo Bird is starting to watch its back a little more often. If you enjoy insights like this, follow me Alex Banks for more on AI. P.S. Join 40,000+ others in my newsletter to stay on the cutting edge of AI: https://lnkd.in/dPNxW8-Y

✨ Ever feel like your AI-generated copy misses the mark? The sentences might be grammatically correct, but they just don't flow naturally. That's where dependency grammar comes in - a linguistic framework that focuses on connecting words based on their relationships. I actually discovered it in the recent episode of the Lex Fridman show with Edward Gibson, where he shared insights on language, grammar and syntax. What stood out to me most was the ⭐ Dependency Grammar Framework. By leveraging the framework in your prompts, you can guide any AI to craft copy that's easier to read and comprehend. HOW DEPENDENCY GRAMMAR CAN HELP YOUR COPY: - Creates more natural-sounding sentences - Improves readability and comprehension - Enhances the persuasive impact of your copy --- 🧩 THE DEPENDENCY GRAMMAR FRAMEWORK 🧩 Dependency grammar analyzes how words in a sentence depend on each other to convey meaning. By prioritizing these connections in your prompts, you can steer AI to generate copy that flows logically from one idea to the next. How the Dependency Grammar Framework helps: - Identifies the core elements of a sentence - Connects related words and phrases - Ensures a clear, logical flow of ideas - Here's the Dependency Grammar Framework: - Identify the main verb or action - Connect the subject to the verb - Link modifiers to the words they describe - Ensure a clear relationship between clauses - Maintain a logical flow from one idea to the next Ready to put this framework into action with a powerful AI prompt? Keep reading to discover the mega-prompt that will transform your copy! 🥳 ⚙️ HERE’S THE GRAMMAR PROMPT ⚙️ Use the dependency grammar linguistic framework rather than phrase structure grammar to craft a [ARTICLE/POST/EMAIL/ETC.]. The idea is that the closer together each pair of words you’re connecting are, the easier the copy will be to comprehend. Here is the topic and additional details: [DETAILS] ❓ HOW TO USE THE PROMPT ❓ STEP 1: Replace [ARTICLE/POST/EMAIL/ETC.] with the type of copy you're writing STEP 2: Fill in the [DETAILS] section with your topic and any additional information STEP 3: For best results, keep your [DETAILS] section concise and to the point 💥🚀 Happy Prompting! Follow Me For More 👍 #AI #Prompts #prompting #promptengineering #artificialiIntelligence

My dear network, I'm pleased to share a new article on my blog, following up on the previous one about massively automating text translation. This new article aligns perfectly with the next step: transforming translated text into human-like speech. It details the process of massivelly converting text to speech (#TTS) using realistic AI voices via the #Elevenlabs API, regardless of the translated language. Sounds exciting, isn't it ?! 😁 https://lnkd.in/gPJacFrK

Have you ever considered how a Content Designer's skills could enhance Large Language Models (LLMs)? 🧐 📲 How do we ensure these AI tools understand and accurately complete tasks? The key is clear communication - much like working with humans. 👥 In this blog post, Alex Robertson shares how Content Designers here at Justice Digital are leveraging their expertise to extract insights from complex documents like service assessment reports. 🚀📈 👉 Read the full post and learn how content design can help unlock the full potential of AI: https://bit.ly/JD-AI-CD