Good morning! This is your daily ☕️ Techpresso.
In today's Techpresso:
🤖 Instagram now lets you create an AI chatbot of yourself
💸 Perplexity's new revenue sharing plan
🦾 Nvidia announces new support for humanoid robots
🖼️ Meta's new open-source model could be the 'GPT-4 moment' for computer vision
🍎 Apple says its AI models were trained on Google's custom chips
🫠 Ghostery’s CEO says regulation won’t save us from ad trackers
🤔 Microsoft app tracks employee activity
🎁 + 5 other news you might like
🔮 + 1 handpicked research papers and tools
🤖 Instagram now lets you create an AI chatbot of yourselfLINK
Meta has released a new tool called AI Studio, enabling users in the US to create AI characters on Instagram or the web to interact with followers on their behalf.
These AI profiles can engage in direct chat threads, respond to comments, and are customizable based on the creator’s Instagram content and specified interaction guidelines.
In addition to creating personalized AI, users can also design entirely new characters to use across Meta’s platforms, with Meta ensuring these AI profiles are clearly labeled to avoid confusion.
💸 Perplexity's new revenue sharing planLINK
Perplexity has started a program to share advertising revenue with publishers after facing plagiarism accusations from several media outlets.
The "Publishers' Program" includes partners like Time, Der Spiegel, and Automattic, who will receive a portion of ad revenue for their content used by Perplexity.
This initiative follows investigations by Forbes and Wired, which reported Perplexity’s AI misusing and paraphrasing their articles without proper attribution.
🦾 Nvidia announces new support for humanoid robotsLINK
Nvidia has introduced a new suite of services, including the NIM microservices platform and the OSMO orchestration service, to aid in the development, simulation, and training of humanoid robots.
CEO Jensen Huang emphasized that Nvidia is advancing its robotics stack to support global humanoid developers, offering platforms, acceleration libraries, and AI models tailored for their needs.
At the SIGGRAPH conference, Nvidia showcased an AI-enabled teleoperation workflow and detailed three robotics development platforms: Nvidia AI supercomputers, Nvidia Isaac Sim, and Nvidia Jetson Thor humanoid robot computers.
🖼️ Meta's new open-source model could be the 'GPT-4 moment' for computer visionLINK
Meta has introduced SAM 2, a cutting-edge open-source model for segmenting both images and videos, marking a significant advancement in computer vision similar to OpenAI's GPT-4 in natural language processing.
While the original SAM focused solely on images, SAM 2 excels in video segmentation, effectively handling lower-quality footage and partially obscured objects, thanks to training on a vast new video dataset.
SAM 2's improved accuracy, enhanced memory module for better object tracking, and faster processing speed positions it as a groundbreaking tool in the fields of video editing, robotics, and generative AI, despite some limitations.
🍎 Apple says its AI models were trained on Google's custom chipsLINK
Apple used Google's tensor processing units (TPUs) to train two artificial intelligence models, according to a recent research paper.
To train its AI models, the company employed 2,048 TPUv5p chips for devices like iPhones and 8,192 TPUv4 processors for server-based models.
Unlike Nvidia’s GPUs, Google’s TPUs are accessible only via Google Cloud Platform, requiring customers to build software through this platform to utilize the chips.
🫠 Ghostery’s CEO says regulation won’t save us from ad trackersLINK
According to Ghostery CEO Jean-Paul Schmetz, protecting online privacy cannot solely depend on regulations like GDPR because companies are adept at circumventing such rules.
Schmetz highlights that Big Tech and regulatory bodies are unlikely to provide sustained privacy protection, emphasizing the need for users to adopt tools like Ghostery for comprehensive security against ad trackers.
He observes that despite occasional regulatory impacts, tracking and ad technologies have continued to evolve and proliferate, necessitating personal action to manage online privacy effectively.
🤔 Microsoft app tracks employee activityLINK
Microsoft’s Dynamics 365 Field Service software uses AI to monitor workers' performance, raising concerns about employee surveillance according to a study by Cracked Labs.
AI predictions within the software can impact task duration estimates for individual workers, potentially affecting employment decisions despite Microsoft claiming it is not intended for such use.
The integration of Microsoft’s Copilot AI aims to streamline work orders and performance metrics, but its accuracy and reliability have been questioned regarding its impact on worker well-being and job security.
Other news you might like
UK antitrust body probes Google’s ties with AI rival Anthropic.LINK
Canva adds a new generative AI platform to its growing creative empire.LINK
New U.S. Commerce Department report endorses ‘open’ AI models.LINK
Meta's AI safety model vulnerable to simple space bar trick.LINK
Meta’s Quest 3 ‘Layout’ app lets you take measurements in mixed reality.LINK
Latest research and tools
Diffusion Training from Scratch on a Micro-Budget: the paper discusses how to train diffusion models effectively with limited financial resources.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}
