Paul Bratcher’s Post

𝐊𝐲𝐮𝐭𝐀𝐢 𝐝𝐞𝐦𝐨𝐞𝐬 𝐟𝐢𝐫𝐬𝐭 𝐜𝐨𝐧𝐯𝐞𝐫𝐬𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐦𝐨𝐝𝐞𝐥: 𝐭𝐡𝐞𝐲'𝐫𝐞 𝐠𝐨𝐢𝐧𝐠 𝐟𝐨𝐫 𝐆𝐏𝐓-𝟒𝐨 💥 Non-profit French lab Kyutai was created last November with a huge endowment of 300M€. So I was eagerly waiting for their first results, but nothing was coming! ➡️ Until now. They really delivered today: in their first ever presentation, they revealed their first model named "𝚖𝚘𝚜𝚑𝚒". It's a model that listens and speaks: audio to audio. How is that so important? Generally, AI systems chain 3 steps to perform audio to audio: 1. Listen. 🔊➡️ 💬 2. Think. 💬 ➡️ 💬 3. Speak 💬 ➡️ 🔊 This is because it's easier to train reasoning in a text to text model, since you have huge corpuses of data to train on. But this position of text as a central bottleneck creates problems. On top of losing non-textual nuances like the voice tone, it creates a horrible latency: it's generally considered very hard for such a pipeline to go under 500ms latency, which makes it too slow for a natural conversation ⛔ So 𝗞𝘆𝘂𝘁𝗮𝗶 𝗷𝘂𝘀𝘁 𝗺𝗮𝗱𝗲 𝗮 𝗱𝗶𝗿𝗲𝗰𝘁 𝗮𝘂𝗱𝗶𝗼 𝘁𝗼 𝗮𝘂𝗱𝗶𝗼 𝗺𝗼𝗱𝗲𝗹. 🔊➡️🔊 For this, they have an impressive novel approach: instead of using text for training, they use extremely compressed audio. The model also uses textual thoughts alongside audio in its training, since written text is still a very efficient representation. It has two streams: input audio (listening) and output audio (speaking) going on at all times. ✅ All of this together gives a 𝙧𝙚𝙖𝙡𝙡𝙮 𝙞𝙢𝙥𝙧𝙚𝙨𝙨𝙞𝙫𝙚 𝙘𝙤𝙣𝙫𝙚𝙧𝙨𝙖𝙩𝙞𝙤𝙣 𝙢𝙤𝙙𝙚𝙡, 𝙖𝙘𝙝𝙞𝙚𝙫𝙞𝙣𝙜 𝙪𝙣𝙙𝙚𝙧 200𝙢𝙨 𝙡𝙖𝙩𝙚𝙣𝙘𝙮. ✨ Moshi has a very natural flow, it can even whisper or imitate a french accent. Small limitation: it tends to continue speaking even when people want it to stop (but I do it too) 𝐈𝐧𝐬𝐢𝐠𝐡𝐭𝐬: ➤ 𝗧𝗶𝗻𝘆 𝘁𝗲𝗮𝗺: 8 people ➤ 𝗦𝗢𝗧𝗔 𝗿𝗲𝘀𝘂𝗹𝘁𝘀: 𝗻𝗮𝘁𝘂𝗿𝗮𝗹 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻𝘀, 𝘄𝗶𝘁𝗵 𝗿𝗲𝗰𝗼𝗿𝗱 𝗹𝗼𝘄 𝗹𝗮𝘁𝗲𝗻𝗰𝘆 𝘂𝗻𝗱𝗲𝗿 𝟮𝟬𝟬𝗺𝘀 🏅 ➤ 𝗧𝘄𝗼 𝗺𝗼𝗱𝗮𝗹𝗶𝘁𝗶𝗲𝘀: main is audio, also uses a second text stream to support thought ➤ Two permanent streams: audio input and audio output ➤ Fine-tuned on 100k synthetic audio + text sampltes ➤ 7B model, probably quantized on 8 bits ➤ Model is going to be released in 𝗢𝗽𝗲𝗻-𝗦𝗼𝘂𝗿𝗰𝗲 🎉

Role of a Conversational AI Operator - Part 1 If the role is solely that of a conversational AI operator, surely focusing on engagement and effective communication rather than the advanced technical aspects of AI development, would seem to be a more logical study or upskill path? Here are the first set of 6 key techniques and skills that we have discovered are becoming more prominent in working with not only competent but great AI conversationalists. This is part 1 Role of a Conversational AI Operator - Part 1 1. Active Listening. The ability to actively listen to user inputs is crucial. Understanding user intent, emotions, and context allows the conversationalist to provide more relevant and empathetic responses. 2. Empathy. Developing and expressing empathy in conversations is essential. It involves understanding and acknowledging the user's feelings, concerns, or needs, creating a more human-like interaction. 3. Effective Communication. Clarity and coherence in communication are paramount. The conversationalist should be able to convey information concisely and in a manner that is easy for the user to understand. 4. Adaptability. Being adaptable to different communication styles and user preferences is important. The conversationalist should be able to adjust their tone and style based on the context and user interactions. 5. Problem-Solving Skills. While not necessarily solving technical problems, a conversationalist should have the ability to address user queries, concerns, or issues effectively. This might involve providing information, offering assistance, or guiding users to relevant resources. 6. Cultural Sensitivity. Understanding and respecting cultural differences is crucial in creating inclusive and respectful conversations. A conversationalist should be aware of potential cultural nuances that may impact communication. See part 2

Apple Researchers Propose a Multimodal AI Approach to Device-Directed Speech Detection with Large Language Models https://lnkd.in/d23xW26V Apple has made a major breakthrough in virtual assistant technology by creating a system that eliminates the need for trigger phrases in human-device interactions. This allows for more natural and spontaneous conversations with virtual assistants. The key to this innovation is Apple's use of a multimodal approach, which combines acoustic data, linguistic cues, and outputs from automatic speech recognition systems. This approach has led to significant improvements, with error rate reductions of up to 61% over traditional audio-only models. The impact of this research is significant, as it aims to make interactions with devices more intuitive, similar to human-to-human communication, fundamentally changing our relationship with technology. For businesses looking to leverage AI, practical solutions include identifying automation opportunities, defining KPIs, selecting AI solutions, and implementing them gradually to stay competitive and redefine the way of work with AI. For AI KPI management advice and continuous insights into leveraging AI, businesses can connect with us at hello@itinai.com or follow us on Telegram and Twitter. Additionally, our AI Sales Bot from itinai.com/aisalesbot is designed to automate customer engagement 24/7 and manage interactions across all customer journey stages, redefining sales processes and customer engagement. For more information and free consultation, businesses can visit our AI Lab in Telegram @aiscrumbot and follow us on Twitter @itinaicom. #ai #ainews #llm #ml #startup #innovation #itinai #uxproduct

🔊 Unlocking the Power of Emotional Prosody in AI Voice Applications So what is Emotional Prosody? Wikipedia Definition: Emotional prosody or affective prosody is the various non-verbal aspects of language that allow people to convey or understand emotion. It includes an individual's tone of voice in speech that is conveyed through changes in pitch, loudness, timbre, speech rate, and pauses. In the realm of human communication, the nuances of how we say something often carry as much weight as the words themselves. This phenomenon, known as emotional prosody or affective prosody, encompasses the rich tapestry of non-verbal cues in our speech, such as variations in pitch, loudness, timbre, speech rate, and strategic pauses. These subtle yet profound elements of our voice play a pivotal role in conveying emotions and intentions, transcending the mere textual content of our messages. As we venture deeper into the era of AI, the quest for creating voice applications that resonate with true human-like qualities becomes paramount. The challenge transcends achieving a good-sounding voice; it involves capturing the essence of human emotionality through speech. To truly revolutionize user experience, AI technologies must master the art of emotional prosody. This means programming AI to understand and replicate the natural flow of human speech, including the adept use of pauses, filler words, and, most crucially, the appropriate modulation of voice to express emotions like empathy and enthusiasm at just the right moments. Why is this so critical? Because at the heart of effective communication lies the ability to connect emotionally. In scenarios ranging from customer service bots to virtual assistants and beyond, the capacity of an AI to exhibit genuine affective prosody can make the difference between a sterile interaction and a genuinely engaging, human-like experience. As we develop these advanced AI systems, our goal should not only be to mimic human speech but to embody the emotional intelligence that underpins human interaction. The journey towards imbuing AI with the subtleties of emotional prosody is both a technical challenge and an artistic endeavor. It requires a deep understanding of linguistics, psychology, and cutting-edge AI programming techniques. Yet, the rewards promise to be profound, paving the way for more intuitive, empathetic, and engaging AI applications that truly speak to the human condition. As innovators, developers, and dreamers, we stand on the brink of a new frontier in AI development. Let's embrace the complexity and beauty of emotional prosody, unlocking new dimensions of interaction! #AI #VoiceTechnology #EmotionalProsody #HumanComputerInteraction #Innovation

🌐 Develop a global mindset and make diversity your strength. 🤖 Amazing insight into how #AI algorithms are continually getting better and how the market size for emotion detection and conversational AI is expected to grow to more than $55 billion by 2026. At Learnlight we believe the nature of communication skills can be broken down into language, intercultural, and interpersonal skills #learninganddevelopment #edtech #ailearning

I Think, Therefore I Collaborate Personal Perspective: The allure of conversational AI. Key points 1.Conversational AI enables natural, emotional human-machine communication. 2.This AI unlocks new levels of human-AI collaboration across business and social sectors. 3.It acts as an intuitive "brain interface" bridging human and machine intelligence. The Rise of the Emotionally Intelligent Machine At the heart of this shift is a simple but profound idea: that machines can engage with us on a level that feels natural, intuitive, and emotionally resonant. By leveraging the power of natural language processing and machine learning, conversational AI systems like are able to understand and respond to human communication in ways that feel authentic, empathetic, and deeply personal. And while this task have been on the radar for many years, we seem to at a proverbial inflection point and I believe the implications will ripple across both technology and humanity. A New Era of Collaboration So what does this mean for the future of human-AI interaction? In short, it opens up a world of exciting possibilities for collaboration and co-creation. Imagine a scenario where every student has access to a highly knowledgeable, endlessly patient, and emotionally attuned AI tutor. Or where patients can rely on a friendly, supportive AI companion to help them navigate the complexities of the healthcare system or learn about a disease with the simplicity and compassion uniquely crafted for the user. Bridging the Digital Divide But perhaps the most exciting aspect of conversational AI is its potential to serve as a kind of "brain interface" between humans and machines. By engaging with us on a level that feels intuitive and emotionally resonant, systems like Claude and ChatGPT are creating a bridge across the digital divide, allowing us to communicate with them in a way that feels truly meaningful and authentic. Challenges and Opportunities Of course, as with any transformative technology, there are challenges and risks to consider. How do we ensure that conversational AI systems are developed and deployed in an responsible manner? How do we navigate the complex social and psychological implications of forming deep, emotionally resonant bonds with artificial minds? After all, the power of AI's superhuman persuasion can emerge as one of its greatest tools—for better or worse. I believe that conversational AI represents a major turning point in the history of human-machine interaction. By engaging with us on a level that feels natural, intuitive, and emotionally resonant, systems are opening up a world of possibilities for collaboration, co-creation, and mutual understanding. Whether we're talking about education, healthcare, business, or any other domain of human endeavor, the potential impact of conversational AI is truly staggering—and its at hand.

I published a post last week about how I don’t use AI to write. It got a LOT of attention. But there’s a flipside I didn’t talk about in that post. In my academic studies examining the impact of generative AI on strategic communication, I did find use cases where it made sense to use GenAI for writing. Here’s an excerpt from the literature review I delivered to the communication department at UCCS: “ChatGPT has the potential to even the playing field for communication professionals with strong idea generation skills, but weaker writing skills (Noy & Zhang, 2023). If generative AI tools like ChatGPT can help strategic thinkers get their ideas down into a structured, effectively written format that others can understand, I believe we may see new creative messaging from individuals whose ideas previously went unheard due to disability, neurological divergence, educational background, or societal bias.” Now, we each must make these choices for ourselves, and if we are working within an organization, we must contribute thoughtfully to the conversation around using generative AI. There are certainly ethical and legal concerns with this technology to take into consideration — and there are broader societal impacts as well. Today, using or not using AI is not a binary choice. It’s a series of considerations and decisions we must make, because these tools are now baked into our everyday lives. I choose not to use AI to do my creative work or critical thinking, because I believe that in the communication work I do — from copywriting to authoring books to writing academic papers to speaking on stage — I make a more positive impact on the world doing the writing myself. I would argue that’s probably true for most professional communicators. *But I’m extremely privileged.* I have full use of my neurotypical brain for thinking, my eyes for seeing words on the page, my fingers for typing on a keyboard. I was born into a suburban, middle-class, college-educated family, and I am college educated myself. There are many people in this world I can’t reasonably speak for. And so I wanted to post this flipside of the conversation to help us all see the broad spectrum before us.

I love this post on Generative AI. How are you using it, personally or professionally? I'd love to read your comments.

This year at #bioeuropespring, many conversations centered on the use of AI and generative AI in the pharmaceutical industry. Translation is one of the simplest use cases. Today, while reading a post written by Michael Jackson, I came across the word "stan." My first instinct when encountering a new word is still the old habit I learned 40 years ago: Look it up in a dictionary. But "stan" isn't found in traditional or online dictionaries. As a result, I consulted a generative AI and got this answer: Here's a detailed explanation of "stan" in this context, along with its etymology: Meaning: In contemporary slang, "stan" is a combination of the words "stalker" and "fan." It describes someone who is extremely passionate, bordering on obsessive, about a celebrity or public figure. A "stan" goes beyond regular admiration: They know extensive details about the person's life. They actively defend them on social media. They might see the person as a role model or even an idol. Etymology: The word "stan" originated from Eminem's 2000 song of the same name. In the song, an obsessed fan named Stan becomes progressively more unhinged over his devotion to Eminem. While the song is disturbing, the term "stan" took on a lighter meaning in internet culture. Additional Notes: "Stan" can be used both as a noun ("I'm a Beyoncé stan") and a verb ("I stan Taylor Swift so hard"). The intensity implied by "stan" is often humorous and exaggerated. Let me know if you'd like any other internet slang explained! Generative AI is a vital tool for non-native speakers to gain a deeper understanding of word meanings in conversations. Five years ago, "stan" would have remained a mystery. A business partner might have explained it, but I doubt I'd have remembered in the middle of a meeting. Similarly, during a training session, it's unlikely I'd recall every new word. This is a gap AI can close. Will it replace translators or English trainers? I don't think so. I still need professionals to help me craft important business marketing messages. AI will augment their skills, not replace them. And similarly, going forward, AI will empower me to improve all the messages I write – even those I'd never bother sending to a translator. The same goes for English trainers. I need to practice real-life conversational scenarios, where I may not have access to generative AI during small talk or negotiations. AI will empower people and fill in knowledge gaps, much like the internet and mobile revolutions did. Always remember: Deals are done between humans. Tools empower them, but never eliminate them from the equation. This is a common misconception at the start of every tech revolution. #ai #entrepreneurship #business

🌟 Let's delve into the exciting advancements and trends shaping the future of chatbot development: 1. Voice Assistants Integration: With the proliferation of smart speakers and voice-enabled devices, integrating chatbots with voice assistants like Amazon Alexa, Google Assistant, or Apple's Siri has become essential. This integration enables users to interact with chatbots using natural language, making the experience more intuitive and convenient. 2. AI-powered Conversational Agents: Advancements in natural language processing (NLP) and machine learning have empowered chatbots to understand and respond to user queries more intelligently. AI-powered conversational agents can analyze user inputs, learn from interactions, and provide personalized responses, leading to more engaging and effective conversations. 3. Chatbots for Virtual Events: As virtual events continue to gain prominence, chatbots are playing a crucial role in enhancing attendee experiences. These chatbots can assist with event registration, provide event information, answer FAQs, facilitate networking among participants, and even deliver personalized recommendations for sessions or exhibitors. 4. Multimodal Interfaces: Chatbots are evolving beyond text-based interactions to support multimodal interfaces, including voice, text, images, and even gestures. This enables users to engage with chatbots using their preferred mode of communication, enhancing accessibility and user experience. 5. Hyper-personalization: Chatbots are increasingly leveraging data analytics and user profiling to deliver hyper-personalized experiences. By analyzing user behavior, preferences, and past interactions, chatbots can tailor responses and recommendations to individual users, enhancing engagement and satisfaction. 6. Integration with Messaging Platforms and Social Media: Chatbots are being integrated with popular messaging platforms and social media channels to meet users where they are. Whether it's Facebook Messenger, WhatsApp, Slack, or WeChat, chatbots enable businesses to engage with customers in real-time on platforms they already use regularly. 7. Low-Code/No-Code Development: The emergence of low-code/no-code development platforms has democratized chatbot development, allowing non-technical users to create and deploy chatbots with minimal coding knowledge. This trend is lowering the barrier to entry for businesses looking to leverage chatbots for customer service, marketing, and sales. 8. Conversational AI Ethics and Trust: With the increasing sophistication of conversational AI, there is growing emphasis on ethical considerations and building trust with users. 9.Emotional Intelligence: Future chatbots are expected to possess emotional intelligence, enabling them to recognize and respond to user emotions effectively. 🚀 #ChatbotRevolution #AIChatAdvancements #VirtualEventBots #MultimodalEngagement