Good morning! This is your daily ☕️ Techpresso.
In today's Techpresso:
🖥️ Anthropic’s new AI can use computers like a human
🚀 Elon Musk's xAI launches API for Grok
🤖 Reddit CEO says the platform is in an 'arms race' for AI training
⚖️ Major publishers sue Perplexity AI for scraping without paying
📸 Meta is testing facial recognition to fight celebrity scams
🧠 Lab-grown human brain cells drive virtual butterfly in simulation
🎁 + 10 other news you might like
🔮 + 9 handpicked research papers and tools
🖥️ Anthropic’s new AI can use computers like a humanLINK
Anthropic has released an enhanced Claude 3.5 Sonnet model capable of controlling desktop applications using a "Computer Use" API, allowing it to replicate human-like interactions such as keystrokes and mouse movements.
This upgrade aims to position Anthropic in the competitive landscape of AI-powered virtual assistants, with the 3.5 Sonnet model demonstrating improved performance in certain coding tasks compared to OpenAI’s flagship, despite facing challenges in basic actions like scrolling and zooming.
Anthropic acknowledges the potential risks associated with providing desktop access to AI models, but it has implemented several security measures to mitigate misuse and has partnered with government agencies to evaluate the model's safety before deployment.
🚀 Elon Musk's xAI launches API for GrokLINK
Elon Musk's AI venture, xAI, has launched an API featuring its flagship generative AI model, Grok, but currently, it only includes the basic "grok-beta" version for use.
The pricing for xAI's API is set at $5 per million input tokens and $15 per million output tokens, with each token representing a small data segment like a syllable.
xAI is racing to compete with AI giants such as OpenAI, utilizing X's data for training and aiming to integrate Musk's different companies' data to enhance technological advancements.
🤖 Reddit CEO says the platform is in an 'arms race' for AI trainingLINK
Reddit CEO Steve Huffman stated that the platform is a vital player in the AI "arms race," emphasizing its role in providing high-value training data for artificial intelligence development.
The platform's extensive user-generated content has become crucial in shaping AI models, leading Reddit to explore its strategic position within the artificial intelligence sector.
In response to large corporations utilizing Reddit data without proper agreements, Huffman revealed ongoing efforts to secure deals and safeguard the platform's valuable information against exploitation.
⚖️ Major publishers sue Perplexity AI for scraping without payingLINK
Major publishers Dow Jones & Co and NYP Holdings have filed a lawsuit against AI search engine startup Perplexity for copying their content without compensation, alleging copyright infringement and trademark violations.
News Corporation, representing The Wall Street Journal and New York Post, accuses Perplexity of presenting the scraped material as a substitute for original sources, consequently harming the brands and sometimes providing inaccurate information.
News Corp seeks $150,000 for each infringement instance, a sum that could financially devastate Perplexity, highlighting the importance of protecting intellectual property while also showing a willingness to license content for appropriate fees, as demonstrated by their agreement with OpenAI.
📸 Meta is testing facial recognition to fight celebrity scamsLINK
Meta is testing facial recognition technology to combat 'celeb-bait' scam ads by comparing ad images against celebrities' profile pictures on Facebook and Instagram.
Facial recognition is also being explored as a faster method for users to regain account access through video selfies, providing an alternative to traditional ID verification methods.
While the tests show promising results, they are not yet being conducted in the U.K. or the EU, due to stringent data protection regulations in these regions.
🧠 Lab-grown human brain cells drive virtual butterfly in simulationLINK
Researchers at FinalSpark have created a 3D simulation where a virtual butterfly is guided by lab-grown human brain cells, marking a significant advancement in biocomputing and cognitive technologies.
The brain organoids, which are miniature brains grown from stem cells, respond to human input in a virtual setting, allowing the butterfly model to move in response to stimuli through a Python software framework.
These biological neural networks promise advantages like lower energy consumption and advanced cognitive functions, though they currently require traditional computing infrastructure support, with potential ethical questions regarding consciousness and usage implications.
Other news you might like
Meta Joins Twitter/X in Suspending Celebrity Jet Tracker Accounts From the Infamous Jack Sweeney.LINK
Blade Runner 2049 producers sue Elon Musk over AI art at robotaxi event.LINK
Qualcomm brings laptop chip technology to mobile devices for AI enhancement.LINK
Meta’s smart glasses outsell traditional Ray-Bans in some stores, even before AI features roll out.LINK
MIT engineers create solar-powered desalination system producing 5,000 liters of water daily.LINK
Asana launches a no-code tool for designing AI agents - aka your new 'teammates'.LINK
Google’s taking the extra search box out of your search results.LINK
Former Nvidia engineer discovers 41-million-digit prime — largest prime number known to man was uncovered and verified with the help of GPUs.LINK
Marc Andreessen says AI model makers are in ‘race to the bottom’ and it’s not good for business.LINK
Basecamp-maker 37Signals says its “cloud exit” will save it $10M over 5 years.LINK
Latest research and tools
LTESniffer: an open-source tool that captures and analyzes LTE wireless communications between cell towers and smartphones, supporting security research by enabling the study of unencrypted messages and data traffic.LINK
Data Formulator: an AI tool from Microsoft Research that simplifies the creation of data visualizations by blending user interface interactions and natural language inputs.LINK
Scalene: a Python profiler that provides detailed information on CPU, GPU, and memory usage, and suggests AI-powered optimizations to improve code efficiency.LINK
Guide to Fine-Tuning LLMs: the document discusses methods and best practices for fine-tuning large language models to improve their performance and adaptability.LINK
Rust Web Framework Rwf: a framework for creating web applications in Rust, offering a classic MVC architecture to simplify the development of fast and secure web apps.LINK
Firefox-Passwords-Decryptor: extracts and decrypts passwords saved in Firefox, helping users to manage and secure their online credentials effectively.LINK
Vimium: a browser extension that lets users navigate and control the web using keyboard shortcuts, inspired by the Vim text editor.LINK
Transformers Utilization in Chart Understanding: the paper reviews advancements in applying Transformer models for interpreting charts and identifies potential areas for future research.LINK
Machine Learning to Computational Plasma Physics Reduced-Order Plasma Modeling: the paper discusses how machine learning can be applied to simplify plasma physics models, making the research accessible to a broader audience.LINK
Want to get the latest news differently? Find us on:
See you tomorrow for a new dose of ☕️ Techpresso!
-min.png){kind=small}
.png){kind=small}
