- The AI Pulse
- Posts
- 🤖 Perplexity AI CEO Refuses to Define “Plagiarism”
🤖 Perplexity AI CEO Refuses to Define “Plagiarism”
PLUS: Mysterious Text-to-Image AI Model Revealed, A New Technique to Train General-Purpose Robots
Welcome back AI enthusiasts!
In today’s AI Report:
🎙Perplexity AI CEO Refuses to Define “Plagiarism”
🧙Mysterious Text-to-Image AI Model Revealed
🦾A New Technique to Train General-Purpose Robots
🛠Trending Tools
💰Funding Frontlines
💼Who’s Hiring?
Read Time: 3 minutes
🗞RECENT NEWS
TECHCRUNCH DISRUPT 2024
🎙Perplexity AI CEO Refuses to Define “Plagiarism”
Image Source: TechCrunch Disrupt 2024/“Perplexity AI CEO Aravind Srinivas on the rush toward an AI-curated web.”/YouTube/Screenshot
Perplexity AI CEO Aravind Srinivas wouldn’t say how his company defines “plagiarism” during an interview at TechCrunch Disrupt 2024.
Key Details:
Perplexity AI is a free AI-powered Answer Engine that rivals Google Search. You can ask Perplexity AI any question, and it searches the internet for the best answer.
An Answer Engine directly answers questions without requiring you to click through websites.
Srinivas was adamant that Perplexity AI “always cites sources” and doesn’t “claim ownership of any content.”
“It’s surfacing content from websites, summarizing that content into an easy-to-digest format, and showing where the content originated from,” he explained. “Exactly how journalists do their job, academics do their job, and students do their homework.”
“At the end of every sentence, there’s a footnote from where the information is from,” Srinivas added.
AI Plagiarism Detector Copyleaks found that Perplexity AI paraphrased 48% and plagiarized 7% of a Forbes paywall article.
When asked about The Washington Post and The New York Times lawsuits, Srinivas argued that publishers wish Perplexity AI “didn’t exist” and would prefer “publicly reported facts to be owned by corporations.”
Perplexity AI recently developed a “Publishers’ Program” to support the vital work of media organizations.
When Perplexity AI earns revenue from an interaction where a publisher’s content is referenced, that publisher will also earn a “double-digital percentage.”
🚨Watch Srinivas’s interview at TechCrunch Disrupt 2024 here.
🩺 PULSE CHECK
Do you agree with Srinivas’s stance that Perplexity AI doesn’t commit “plagiarism”?Vote Below to View Live Results |
RECRAFT
🧙Mysterious Text-to-Image AI Model Revealed
Image Source: “Recraft Introduces a Revolutionary AI Model That Thinks in Design Language”/Screenshot
Recraft, a company that enables designers to create and perfect their visuals, unveiled a text-to-image AI model called “Recraft V3” that outperforms everyone.
Key Details:
“Recraft V3” delivers across-the-board improvements in generating high-quality images with dramatically improved human features like hands, eyes, hair, and skin.
“Recraft V3” boasts improved Text Rendering, allowing developers to create images that include long texts, as “opposed to just a couple of words.”
Over the past four days, “Recraft V3” participated in the Artifical Analysis Text-to-Image AI Model Leaderboard, securing the top spot with an 1172 Arena Elo and 72% Arena Win Rate. In comparison, OpenAI’s DALL-E 3 achieved a 970 Arena Elo and 50% Arena Win Rate.
Why It’s Important:
“Recraft V3’s” improved Text Rendering enables the creation of more detailed images with complex text layouts, expanding creative possibilities.
Developers can create stunning visuals in seconds with AI-assisted tools, accelerating the design process, enhancing design quality, and reducing technical barriers.
AI RESEARCH
🦾A New Technique to Train General-Purpose Robots
Image Source: MIT, CSAIL/META, FAIR/“Scaling Proprioceptive-Visual Learning With Heterogenous Pre-Trained Transformers”/Screenshot
In a classic cartoon called “The Jetsons,” Rosie the Robotic Maid seamlessly switched from vacuuming the house to cooking dinner to taking out the trash. To this day, training a general-purpose robot is still a major challenge.
To tackle this challenge, MIT CSAIL Researchers and Meta FAIR Scientists developed “Heterogenous Pre-Trained Transformers (HPT),” a technique for training AI models on a diverse dataset of robotic tasks to make robots more versatile.
“HPT” works by training a Large Neural Network (NN) on a diverse dataset of robotic tasks. This Large NN recognizes patterns between different robots performing different tasks in different environments. Recognizing these patterns helps train AI models that enable robots to generalize their skills to new situations.
Traditional robots are often highly specialized, excelling in specific tasks but struggling with new situations. “HPT” aims to overcome this specialization barrier by allowing robots to learn from the experiences of other robots and apply it to new situations.
🚨Watch videos of “HPT” helping real robots complete real tasks here.
🛠TRENDING TOOLS
⚙️Aidelly is an AI-powered marketing assistant.
🗣Aptitude conducts conversations with hundreds of customers.
⏱TimeOS is a productivity companion that summarizes your day.
🥦SpinachAI turns meetings into notes, results, actions, and insights.
🎓Quillminds supercharges your learning with AI-assisted study tools.
🔮Browse our always Up-To-Date AI Tools Database.
💰FUNDING FRONTLINES
💼WHO’S HIRING?
Pinterest (Remote): Information Technology {IT} Software Engineering Intern, Summer 2025
Waymo (Mountain View, CA): Snow Water Equivalent {SWE} Simulation Intern, Summer 2025
General Motors {GM} (Mountain View, CA): Software Engineering Intern, Commercial Software, Summer 2025
Forward (San Francisco, CA): Software Engineer, New College Grad 2025
Nvidia (Santa Clara, CA): Applied Large Language Model {LLM} Engineer, AI and Chip Design, New College Grad 2025
🤖PROMPT OF THE DAY
BUSINESS PLAN
📊Six Month Business Plan
Help me create a six-month business plan for [Venture] with [Product or Service] in [Industry] with [Target Audience] and [Budget Size].
Provide as much detailed information as possible relevant to my situation, including the “3 C’s”: “Customer,” “Capital,” and “Concept.”
Venture = [Insert Here]
Product or Service = [Insert Here]
Industry = [Insert Here]
Target Audience = [Insert Here]
Budget Size = [Insert Here]
📒FINAL NOTE
FEEDBACK
How would you rate today’s email?It helps us improve the content for you! |
❤️TAIP Review of The Day
“Use my review pleeaaassee!”
REFER & EARN
🎉Your Friends Learn, You Earn!
You currently have 0 referrals, only 1 away from receiving ⚙️Ultimate Prompt Engineering Guide.
Refer 3 friends to learn how to 👷♀️Build Custom Versions of OpenAI’s ChatGPT.
Copy and paste this link to friends: https://theaipulse.beehiiv.com/subscribe?ref=PLACEHOLDER
Reply