Good morning! This is your daily ☕️ Techpresso.
On this day, 50 years ago, Atari introduced Gran Trak 10, the first arcade game to use solid-state ROM for storing sprites.
In today's Techpresso:
💥 Apple in talks with Google to use their AI models
🕶️ Leak reveals specs of Meta's more affordable Quest 3 headset
🍎 Apple introduces new 'MM1' AI model
🤖 xAI open sources base model of Grok, but without any training code
🔍 Reddit IPO reveals company's aspirations and concerns
📹 YouTube mandates AI content disclosure by creators
👀 Google researchers unveil ‘VLOGGER’, an AI that can bring still photos to life
🎁 + 8 other news you might like
🔮 + 4 handpicked research papers and tools
💥 Apple in talks with Google to use their AI modelsLINK
Apple and Google are discussing a partnership to integrate Google's Gemini AI into Apple's iPhone software features.
The collaboration could enhance their existing search partnership, which currently involves Google paying Apple approximately $20 billion annually to be the default search engine on iOS devices.
Despite ongoing negotiations and potential antitrust concerns, the deal, aimed at introducing powerful AI capabilities to iPhones, may not be announced until Apple's developer conference in June.
🕶️ Leak reveals specs of Meta's more affordable Quest 3 headsetLINK
Meta is reportedly developing the Quest 3s, a more affordable and compact alternative to the Quest 3, with speculated specifications revealed through a leaked Meta User Research meeting.
The purported Quest 3s features a lower resolution of 1920x1832 pixels and half the internal storage capacity at 256GB, but it might include enhanced tracking capabilities with six sensors.
Despite being more budget-friendly due to reduced resolution and storage, the Quest 3s' final specifications and price remain speculative until confirmed by Meta.
🍎 Apple introduces new 'MM1' AI modelLINK
Apple researchers have unveiled the 'MM1' AI model, which is capable of training on both text and visual inputs, aiming to create more intelligent and flexible AI systems.
The MM1 model utilizes a diverse dataset that includes image-caption pairs and text-data, improving its performance on tasks like image captioning and visual question answering.
The research highlights the MM1 model's advanced in-context learning abilities, especially in its largest configuration, enabling multi-step reasoning over images with minimal examples.
🤖 xAI open sources base model of Grok, but without any training codeLINK
Elon Musk's xAI has released the base code of its Grok AI model, described as a "314 billion parameter Mixture-of-Expert model" on GitHub, without including its training code.
The Grok model, under Apache License 2.0 allowing commercial use, was not designed for specific applications and lacks the training code and connections to the X social network.
While xAI's Grok joins the ranks of open-sourced AI models from companies like Meta and Google, its release sparks interest among AI tool developers for potential implementation in their solutions.
🔍 Reddit IPO reveals company's aspirations and concernsLINK
Reddit is preparing for its stock market debut, having revised its IPO pitch 10 times, reflecting changes in company priorities and market strategies. CEO Huffman's messaging evolved, emphasizing Reddit's unique, user-driven community and downplaying previous concerns. Reddit confronts user growth stagnation, limited expansion in non-English markets, increased reliance on Google, and new revenue models like Reddit Pro.
📹 YouTube mandates AI content disclosure by creatorsLINK
YouTube now mandates creators to inform viewers when AI was used to make content appear realistically, through a new tool in Creator Studio for disclosing altered or synthetically generated media.
The policy aims to reduce deception among viewers by distinguishing synthetic content from real, especially amid concerns about AI and deepfakes influencing U.S. presidential election perceptions.
Exemptions to the disclosure requirement include clearly fantastical content and use of AI in production assistance, focusing instead on realistic depictions of people, places, events, and voices.
👀 Google researchers unveil ‘VLOGGER’, an AI that can bring still photos to lifeLINK
Google researchers have developed VLOGGER, an AI technology that transforms single still photos into realistic videos incorporating speech and movement.
VLOGGER, using diffusion models and a vast dataset called MENTOR, can animate diverse individuals speaking and gesturing without requiring person-specific training.
While VLOGGER showcases potential uses in dubbing, virtual reality, and presentations, it also raises concerns about deepfakes and misinformation.
Other news you might like
Microsoft promises Copilot will be a 'moneymaker' in the long term.LINK
Apple may launch two new AirPods 4 models this year.LINK
Cargo ship with futuristic sails saves thousands of tonnes of fuel in first test.LINK
Hertz CEO steps down following Tesla EV purchase debacle.LINK
Sony suspends PlayStation VR2 headset manufacturing amid declining sales.LINK
Qualcomm's latest chip brings on-device generative AI to mid-tier smartphones.LINK
LinkedIn is developing in-app games to further distract you from your job hunt.LINK
Nvidia AI developer conference kicks off with new chips in focus.LINK
Latest research and tools
Grok-1 repository: a collection of JAX example code for operating the Grok-1 model with open weights, requiring download of a large checkpoint and a high-GPU memory machine for testing due to its 314 billion parameters, with an emphasis on straightforward implementation over efficiency, and available for download under the Apache 2.0 license.LINK
Vector: a high-performance, open source observability data pipeline that collects, transforms, and routes logs and metrics across various vendors, offering cost savings, data enrichment, and data security.LINK
Planka: an open-source project tracking tool that offers a self-hosted alternative to Trello.LINK
3DGS.cpp: a cross-platform implementation of Gaussian Splatting using the Vulkan API to facilitate high-performance point-based radiance fields on a wide range of GPUs.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}
