Elena Sinel FRSA’s Post

Check out this super-relevant article from Gus Bekdash!

OpenAI promised a reasoning AI, & #chatGPT-4o1 just came. Next in their roadmap is #agency. What are the implications? Here’s a model that makes reasonable predictions relevant to 𝐨𝐮𝐫 𝐥𝐢𝐯𝐞𝐬 𝐚𝐧𝐝 𝐛𝐮𝐬𝐢𝐧𝐞𝐬𝐬. I developed a model for AI predictions that worked well so far. Here's a little piece of it. Natural Intelligence (𝐍𝐈) & especially Human Intelligence (𝐇𝐈) have: 𝐀𝐠𝐞𝐧𝐜𝐲: Capacity to make autonomous decisions and goals and act on them 𝐂𝐮𝐫𝐢𝐨𝐬𝐢𝐭𝐲: Drive to explore and seek new information and experiences for no reason 𝐀𝐦𝐛𝐢𝐭𝐢𝐨𝐧: Desire to achieve specific and challenging goals 𝐅𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐚𝐥 𝐂𝐮𝐫𝐢𝐨𝐬𝐢𝐭𝐲: goal-directed exploration 𝐄𝐱𝐞𝐜𝐮𝐭𝐢𝐨𝐧: Process and capacity to carry out decisions and achieve results 𝐁𝐨𝐬𝐬: Most NI and practically all HI have external entities that direct or limit their actions. The equivalent for AI is the 𝐀𝐈 𝐨𝐩𝐞𝐫𝐚𝐭𝐨𝐫. It’s a model, which means it is a simplified version of reality. Keep that in mind. My dog, cat, and parakeets all have those in various degrees. But let’s use the human standard as the reference point. Humans are 100% in all these areas. #AI4Leaders #AI #AiEvolution 𝟔 𝐩𝐚𝐠𝐞𝐬 / 𝟐 𝐦𝐢𝐧𝐮𝐭𝐞 𝐫𝐞𝐚𝐝

Check this out - over 20 cheat sheets to help you learn about AI. YouTube recommendations, tools, prompt tips, free courses, etc. It is worth downloading the document for reference later on.

Is your favorite AI tool experiencing brain drain? Here’s why quality matters. Many AI chatbots are using Wikipedia, X/Twitter and Reddit to train their AI tools. Some are even resorting to synthetic data (whatever that means).That’s prioritizing data quantity over quality. Your goal isn’t to make ANY decisions — it’s to make BETTER decisions. Enquire’s Ayda copilot does just that, learning from the interactions of our global network of experts and clients. Better yet, you are encouraged to connect to the sources who shaped the answers, networking and collaborating together. AI insights should help connect you to the human expertise you need. If you agree, get started today: https://enquire.ai/pro/ #decisionmaking #ai #artificialintelligence #copilot #chatgpt #claude #grok

With the current state of AI development, we are like the people on the bus. No one knows where it is driving us, some people are yelling turn right, while we are heading left, some are frightened and screaming to stop the damn bus, one part of them is applauding and cheering and the rest of them are scratching their heads in disbelief. The common denominator for most people is that they don't understand the big picture or underlying technology. We can do better, learn, understand, upskill. _______ If you like my posts, please follow me Terezija Semenski, and hit the 🔔 on my profile to get a notification for all my new posts. Grow with me 🚀 #ai #chatgpt #machinelearning #datascience #ml #techwithterezija

🚨 𝗔𝗜 𝗔𝗹𝗲𝗿𝘁: 𝗛𝗮𝘃𝗲 𝘆𝗼𝘂 𝗲𝘃𝗲𝗿 𝗵𝗲𝗮𝗿𝗱 𝗼𝗳 𝗠𝗼𝗱𝗲𝗹 𝗖𝗼𝗹𝗹𝗮𝗽𝘀𝗲? Generative AI models such as #ChatGPT are trained on vast amounts of human-generated text. However, with the rise of AI-generated content (I mean let's be honest, we all use it), the internet is becoming a mix of human and AI-generated text. What happens next, is that subsequent (newer) #GenAI models are increasingly being trained on their own outputs. 😰 𝗪𝗵𝗮𝘁 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝘀𝗮𝘆𝘀 𝗮𝗯𝗼𝘂𝘁 𝗶𝘁 𝗶𝘀 𝗽𝗿𝗲𝘁𝘁𝘆 𝗱𝗶𝘀𝘁𝘂𝗿𝗯𝗶𝗻𝗴.. Shumailov et al. (2023) were curious to understand the consequences of AI-generated content 'poisoning' newer GenAI models. What they found was pretty disturbing. When AI models are trained on data that includes previous AI outputs, they become progressively less accurate. This phenomenon, that they called ‘Model Collapse’, leads to models that have a distorted understanding of reality. The models begin to “contaminate” themselves with each new iteration, degrading the quality of their outputs over time. 🌐 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀 Imagine a scenario where the tools we rely on for accurate information and decision-making start producing flawed and misleading content. This could impact everything from search engines (SearchGPT was only recently announced) to the entire educational system. The effects will be profound. ⚠️ 𝗟𝗼𝗼𝗸𝗶𝗻𝗴 𝗮𝗵𝗲𝗮𝗱 As we continue to integrate AI into more aspects of our lives, it's crucial to be aware of the potential pitfalls. Ensuring the quality of (human-generated) training data and incorporating robust safeguards will be key to preventing Model Collapse and maintaining the integrity of AI systems. Especially considering the fact that some model pollution has already taken place.. 𝗥𝗲𝗳𝗲𝗿𝗲𝗻𝗰𝗲: Shumailov, I., Shumaylov, Z., Zhao, Y., Gal, Y., Papernot, N., & Anderson, R. (2023b, May 27). The curse of recursion: training on generated data makes models forget. arXiv.org. https://lnkd.in/dHTcuMGT #AI #ArtificialIntelligence #ModelCollapse #GenAI #TechTrends #Innovation

The image highlights a significant gap between how many companies view AI and the reality of its complexity. At the top, we see a simplified version where businesses often think of AI as a magical black box—just feed it data, and it spits out value. But the bottom half breaks down the truth: AI involves far more steps than most expect. From selecting and sourcing the right data, to rigorous processes like cleaning, normalizing, and feature engineering, each stage demands attention to detail. Beyond that, there are numerous critical steps involved in model training, evaluation, and tuning before even getting to the "value" stage. And this all happens under the shadow of legal, ethical, historical biases, and security concerns. AI isn't just about algorithms; it's about navigating through these layers of complexity to ensure it works efficiently and responsibly. This breakdown should remind companies that AI requires deep preparation, not just a desire for instant results. Image source: SKILLSLASH #ai #artificialintelligence #genai #generativeai #chatgpt #data #datascience

Do you work with older people? Help empower them with artificial intelligence (AI) knowledge! Join our free webinar next week on Wednesday 20 November at 1pm GMT! Register your place: https://lnkd.in/eyHbDH3u Our expert AbilityNet Digital Skills trainer, Steve Mersereau, will guide you through the fascinating world of AI. Here’s what we’ll cover: - What AI is and what it isn’t: Clear up common misconceptions and understand the true nature of AI. - Demonstration of Popular Free AI Tools: See AI in action with tools like Gemini, ChatGPT, CoPilot, and Perplexity. We’ll show you how these tools can generate text and images. - What’s Next with AI: Explore the future of AI and what it means for you. This webinar is perfect for anyone looking to stay informed and empowered in the digital age. Don’t miss out on this opportunity to learn from the best and ask all your burning questions! #AI #ArtificialIntelligence #OlderPeople #Seniors #TechForGood

AI, e.g., ChatGPT, can't do what we are doing in data analytics NOW. European Parliament's AI Act has identified a practical list of high-risk AI applications. This Act has been approved by European Council 🇪🇺. 關於 AI 人工智能的一些事實 (Chinese version): https://lnkd.in/eeVbdcza Some fact findings about AI (English version): https://lnkd.in/gwcPNUPP 其中，用一個簡單的是非題就能顯示我們受著作權保護的中英對照元數據 (metadata) 能做到現今 AI 人工智能無法做到的數據分析工作。 A simple go/no-go test shows that, with our intellectual property (IP), a copyrighted Chinese-English multilingual metadata, we can do what artificial intelligence (AI) can't do in data analytics, NOW.

Super useful: Another interesting GPT is the GPT Builder assistant. Very helpful in giving the necessary prompts to create a well functioning custom GPT.