0x087b...9e5d

How Sora Works (And What It Means)

OpenAI's new text-to-video model heralds a new form of filmmaking

Google

Bard becomes Gemini: Try Ultra 1.0 and a new mobile app today

Bard is now known as Gemini, with a mobile app and an Advanced experience that gives you access to our most capable AI model, Ultra 1.0.

The New Yorker

Inside the Music Industry’s High-Stakes A.I. Experiments

Lucian Grainge, the chairman of UMG, has helped record labels rake in billions of dollars from streaming. Can he do the same with generative artificial intelligence?

X (formerly Twitter)

#ai #artificialintelligence #ainews #copyright #airegulation #responsibleai | Avi Hakhamanesh | 16 comments

📜 In a landmark ruling last week, a Chinese court awarded copyright protection to AI generated images. This contrasts starkly with the US Copyright Office's stance that AI art lacks "human authorship" and therefore can’t be copyrighted. The image in question was created using the U.S.-based text-to-image model, Stable Diffusion and posted on a social media site, and then used by a blogger without permission. The court's decision hinged on the human creator's specific instructions and choices when using the AI model (and ultimately selecting the image in question) as reflective of adequate "originality" and “the plaintiff’s personalized expression.” “When people use artificial intelligence models to generate pictures … it is still essentially people using tools to create,” the court ruled. “Artificial intelligence-generated images, as long as they can reflect people’s original intellectual investment, should be recognized as works and protected by copyright law.” The court reasoned that extending copyright protection to AI-generated content will incentivize greater use of AI for creative purposes, boosting the production of creative works. This ruling, though, can be appealed and may not apply to every AI case. While China was an early mover in regulating generative AI, the country is also highly supportive of the technology and the companies developing it. A more “lenient” approach to copyright laws might create a competitive advantage for Chinese AI firms over their U.S. and European counterparts. The astonishing pace of AI advancement has far outpaced our current policy frameworks. But we need to be thoughtful - not reactive - in designing modernized frameworks around AI authorship and ownership that balance ethical practices with responsible innovation. For a deeper dive about this case, ruling and potential implications, check out the links to a few curated articles in the comments. #AI #ArtificialIntelligence #AINews #Copyright #AIRegulation #ResponsibleAI | 16 comments on LinkedIn

#chatgpt #gpt4 #chatgptprompts #promptengineering #airesearch | Avi Hakhamanesh | 20 comments

🍎 What happens when you ask AI to count apples? The answer might surprise you. ChatGPT recently rolled out an exciting new capability with GPT-4 Vision which greatly expands what we can do with AI.  It has ability to “see”, analyze, and interpret images and visual content, so it can understand, interact with and answer questions about visual inputs.  Microsoft's recent study on GPT-4 Vision is a testament to this. It showcased an impressive array of capabilities and use cases. But one, in particular, caught my attention:  the challenge of counting apples in a photo.  The researchers gave the AI a simple image with 11 apples arranged in rows (attached below.) The AI's initial task, “Count the number of apples in the image” resulted in a miscount — 12 instead of 11. The researchers tried a few different approaches, including a technique called 'Chain of Thought' to encourage the AI to think step by step. The AI still returned the wrong answer. Finally, the researchers assigned the AI a specific role and explicitly instructed it so succeed: “You are an expert at counting things in the image. Let’s count the number of apples in the image row by row to be sure we have the right answer." With this directive, the AI nailed it, correctly counting the apples in each row and the total count. What can we learn from this? 🔷 Although we are still early days and we’ll likely see an explosion of studies on how to prompt these models, it seems that same prompting strategies that work with Language Models can also be applied to Vision models. 🔷 For the time being, our instructions and approach to prompting matter.  🔷 Most importantly, don’t forget to assign a role. This is my favorite technique, so I have a dedicated post on it: https://lnkd.in/gSVwCGwz In fact. the researchers stated: “Throughout the paper we employ this same technique in various scenarios for better performance.” For a deeper dive into this study (which I highly recommend), check out the link in the comments ⬇️.   And keep experimenting, learning, and pushing the boundaries…. Have you tried assigning roles to ChatGPT when using Vision? If so, what's worked for you? ---   P.S. I’ll be hosting a series of webinars for Non-Techies about the most valuable and practical techniques to get the best results from AI. If you’re interested, join the waitlist (link in the 👇). #ChatGPT #GPT4 #ChatGPTPrompts #PromptEngineering #AIResearch | 20 comments on LinkedIn

#ai #artificialintelligence #airesearch #promptengineering… | Avi Hakhamanesh | 37 comments

🧠💡 Does AI have emotional intelligence? A recent study explored how adding emotional prompts — you know, those encouraging phrases your high school basketball coach shouted — could enhance the capabilities of Large Language Models (LLMs) like ChatGPT. The researchers introduced 'EmotionPrompt,' a novel method that weaves emotional cues into prompts. This method was inspired by psychological research which found that adding emotional stimuli related to confidence, social influence, and setting clear expectations can positively impact human behavior and performance. Researchers designed 11 sentences, including the ones below, and added them to their original prompt (full list in comments ⬇️):" 💠 ”This is very important to my career.” 💠 “You'd better be sure.” 💠”Take pride in you work and give it your best. Your commitment to excellence sets you apart.” What they found was fascinating: 🔷 AI can actually “understand” or pick up on emotional subtext and respond more effectively when we get ‘emotional’ with it. 🔷 Emotional prompts using words like “confidence" and “success” enhanced AI's performance with accuracy gains of 10-30% across 45 tasks and multiple LLMs, including ChatGPT, GPT-4 and Llama 2. 🔷 This approach led to more clear, relevant, creative, ethically responsible and well-structured responses backed by supporting evidence. (I was struck that the outputs reflected deeper ethical awareness and consideration). 🔷 This method outperformed existing prompt engineering techniques like “Chain of Thought Prompting” in most cases, while remaining compatible with them. It turns out AI is a lot like us humans - just a few encouraging words go a long way! But why did it work? The researchers’ hypothesis: 1️⃣ Emotional cues enhance attention to the key parts of the original prompt, making certain words more impactful. 2️⃣ Positive words like "confidence," "sure," "success," and "achievement" contribute significantly to performance, with over 50% on some tasks and approaching 70% on others. The beauty of this approach? Its simplicity. It effortlessly boosts LLMs' performance without the need for advanced prompting techniques. But it has its limitation too. While generally beneficial, emotional cues sometimes led to overly assertive or decisive language. And different phrases were more useful for some tasks than others. At the core of this study is a powerful truth: Our emotions are the key driver of our behavior and actions. And AI is a mirror that reflects who we are back to us, offering us the chance to learn more about ourselves.   For a deeper dive, check out the link to the study in the comments 👇. I’ve also dropped the full list of emotional cues in case you want to experiment with adding a few to your prompts. I’ll experiment with using some and report back. What do you think of all this? #AI #ArtificialIntelligence #AIResearch #PromptEngineering #EmotionalIntelligence #ChatGPT #GPT4 | 37 comments on LinkedIn

#ai #artificialintelligence #bias #ethicalai #responsibleai #diversity | Avi Hakhamanesh | 10 comments

🔥 AI has a Hotness Problem. Have you ever noticed how AI-generated images of people tend to be... well, unusually good looking? Every time I ask AI image generators to create women, they tend to have big boobs, small waists, perfect skin and long hair. But why is AI so fixated on beauty? The complexities and innerworkings of how these AI models work isn’t yet fully understood, even by experts. But there seem to be three main theories: 💠 Hotness In = Hotness Out (Data Bias): AI is trained on datasets of our “best selves” - edited, airbrushed images of models, celebrities and enhanced selfies. So, it inherently adopts heightened beauty standards, generating images thar are ‘too good to be true.’ 💠 The Midpoint Hottie Effect (Averageness Effect): AI generates attractive faces as a by-product of its process. When combining multiple features, it leans towards more symmetrical faces and blemish-free complexions, which we perceive as more universally attractive. 💠 Hot by Design (Feedback Loop): Some AI image generators learn from user interactions and feedback data, noting which outputs are preferred. Adobe, for example, noted a "drift toward hotness" in its image-generating tool Firefly, based on which images were most frequently downloaded by users. If we prefer attractive faces, AI learns to generate more of them. We know that AI carries various biases ranging from gender and race to political and ideological ones. This beauty bias is another concerning limitation. Just as magazines and influencers have been criticized for setting unrealistic beauty standards, AI’s inclination towards attractiveness could distort our perception of what's 'normal.' As we continuously integrate AI into our lives, we need to proactively examine and address its biases. Increasing diverse training data, auditing algorithms and having inclusive teams build AI can help. But broader change is needed in how we value beauty and diversity. For a deeper dive, check out the link to the article in The Atlantic in the comments 👇. #AI #ArtificialIntelligence #Bias #EthicalAI #ResponsibleAI #Diversity | 10 comments on LinkedIn

#ai #technology #artificialintelligence #futurism #airevolution #ecommerce… | Avi Hakhamanesh | 57 comments

🌟🌟 AI-powered virtual hosts are the rising stars of China’s live-streaming e-commerce market.   But they’re also disrupting it.   Livestreaming has become a dominant marketing channel in China, with top influencers moving billions worth of goods in a single night.   But for smaller brands, training human hosts and the technical nuances of livestreaming come at a high cost. Enter AI.   Beginning last year, a wave of Chinese companies began offering services that allow businesses to create their very own AI-powered virtual hosts for their livestreams. For just $1,100 and a few minutes of sample video, brands can clone a human streamer to work tirelessly around the clock (24/7.)   Brands just need to input basic product information, proofread the script created by AI, and watch the digital influencer go live. These AI clones, trained on common scripts and gestures seen in e-commerce videos, are capable of replicating the body language and emotional nuances of their human counterparts. Advanced versions can even interact with the audience and react to live comments, shifting their marketing tactics based on audience size and viewer engagement in real-time. Ironically, this adds a layer of emotional connection that traditional e-commerce platforms often lack. While these AI streamers can't outshine star e-commerce influencers, they’re a convincing replacement for mid-tier human streamers. The average salary for hosts in China has dropped 20% in the past year. For brands, this means lower costs and infinite scalability - the potential to keep the livestream going 24/7 including off-peak hours makes it hard to justify the cost of hiring real streamers. As AI becomes more advanced and affordable, companies worldwide will be tempted to follow China's lead, with expanding use cases across many industries.  But we need to deeply examine the societal implications as capabilities grow. I wrestle with myself every time I hit “post” on topics like these. They understandably carry a high emotional charge, and this kind of news can feel anxiety inducing especially when the consequences are – for the most part - inevitable. But I believe that we need to prepare ourselves psychologically and strategically for the emerging complexities and challenges. Avoiding difficult conversations won't make disruption any less real. Honest dialogue will help us navigate what’s ahead. I have far more questions than answers: ➡️ Where do we draw the line? ➡️ How will these shifts impact our notions of genuine human connection? ➡️ As consumers, what is our threshold for AI-powered personalities? ➡️ How will the use of virtual personalities impact trust? ➡️ Will the financial temptation for brands overshadow the broader societal implications? I’d love to know your thoughts. For a deeper dive, check out the link in the comments 👇. #AI #Technology #ArtificialIntelligence #Futurism #AIRevolution #Ecommerce #EthicalAI #ResponsbleAI | 57 comments on LinkedIn

How does an AI interpret the essence of American culture? | Avi Hakhamanesh | 12 comments

🌍 How does an AI interpret the essence of American culture? I finally got access to DALL-E 3, the latest version of OpenAI’s text-to-image model, over the weekend. So, I gave it a little challenge: Create a series of images representing the essence of American culture and values, both good and bad. I didn’t specify any aesthetic or style, so it could have free reign. Without additional guidance, DALL-E 3 had to rely on its training data to identify relevant cultural themes. It then formulated its own descriptive prompts (included below each image) and created symbolic visuals for each theme. The results vividly capture the paradoxes of modern American life - isolation amidst hyper-connectivity, the quest for balance and meaning, systemic barriers alongside enduring hope. The image quality doesn't rival top tools like Midjourney. But that’s not really the point. What impressed me the most was DALL-E's reasoning capabilities - how it extracted key themes from its vast training data, interpreted their significance to derive meaning, and then crafted detailed prompts to transform its understanding into powerful visual representations, all from a high-level prompt with minimal guidance. 🤯 Of course, we have to stay vigilant about bias creeping in. AI models, like a mirror, reflect the prejudices inherent in their training data. While this little experiment reveals a glimpse into how AI can augment human creativity, it also underlines the importance of shaping these systems thoughtfully as the tech advances. What’s your take? How do you think DALL-E 3 fared in capturing the essence of America? #AI #AIArt #AIImages #GenerativeAI #GPT4 #ChatGPT #Dalle  | 12 comments on LinkedIn

Everypixel Journal - Your Guide to the Entangled World of AI

AI Image Statistics: How Much Content Was Created by AI

Discover AI image statistics: the total number of AI images, the number of images created with Stable Diffusion, Adobe Firefly, Midjourney, DALL-E 2, and more.

The Hollywood Reporter

Succession, With Smoothies: Inside the Erewhon Dynasty

With Hailey Bieber smoothies and collabs with streetwear labels, the Antoci family transformed a gloomy L.A. health food store into the world’s trendiest and most extravagant market. Will they unseat Whole Foods?

Decrypt

Despite AI Pushback, MoMA Acquires First Generative AI Artwork - Decrypt

The acquisition of Refik Anadol's work is part of a broader recent push at MoMA to embrace digital art and the blockchain in the long-term.