Farhan Hammad 一 21 Jan 2024
Pilot Edition #5
Note: This newsletter is published under the pilot edition until all segments are integrated.
New Sections: Highlights & Under The Radar (one project worthy of recognition)
Today’s Edition at a glance:
Resend Forward: Shifts Focus Towards Email Marketing
Apple Vision Pro Pre-Orders Fly Off the Shelves
Meta’s Strategic Moves Towards an Open-Source AGI
Read Time: 6 Minutes
1. Resend Holds Annual Launch Week
Jan 15th 2024 - Resend Forward's recent Launch Week showcased a plethora of features, introduced with a focus on sophistication and user-centric capabilities. Announced this week, these enhancements mark a significant evolution in email marketing.
The standout Resend Broadcasts feature empowers both enterprises and individual builders with a cutting-edge WYSIWYG editor for seamless content creation across various communication needs. Additionally, subscriber management is now natively available on the platform with Resend Audiences, respecting user preferences and streamlining content delivery.
Resend also optimized email delivery times for users in Asia and introduced a new batch emails feature to facilitate sending high volumes of emails with a single click.
The new pivotal features underscore Resend’s commitment to refining the email marketing experience, promising ongoing innovations and insights for developers and users.
2. Apple Vision Pro Sells Out Pre-Orders
Jan 21, 2024 - The highly anticipated Apple Vision Pro, priced at $3,500 rapidly sold out in notable cities, igniting significant market interest. Cities like Seattle, New York, Austin, and Miami have reported a depletion of available units, emphasizing the exceptional demand for this cutting-edge device. Even in Dallas, Texas, where a brief availability window existed, supplies were swiftly exhausted, underscoring the unprecedented demand.
Scheduled deliveries on February 2 align with a strategic approach, focusing on enhancing the in-store experience over immediate stock replenishment. The Apple Vision Pro introduces an extensive orientation process, involving a facial scan and inquiries into the user's eyewear preferences. Those securing a pre-order can expect a meticulously curated package, including the Apple Vision Pro headset, a dual loop band, a protective cover, an external battery, a light seal cushion, a polishing cloth, and a USB-C charging cord with a power adapter.
While the limited availability is apparent, anticipation grows for broader market access in the coming weeks.
3. Meta is Racing towards Open Sourcing AGI
Jan 18, 2024 - Meta CEO Mark Zuckerberg is making waves in the race for Artificial General Intelligence (AGI), even without a defined timeline or precise definition for AGI. To bolster Meta's AI capabilities, Zuckerberg is reshuffling Meta's AI research group, FAIR, integrating it with the team developing generative AI products across Meta's apps. The strategic move aims to ensure Meta's AI breakthroughs directly benefit its vast user base.
In an exclusive interview, Zuckerberg acknowledges the intense competition for AI talent and highlights Meta's commitment to building for general intelligence. The company is investing heavily in computing power, boasting ownership of over 340,000 Nvidia H100 GPUs, a key chip for generative AI, and aiming for nearly 600,000 GPUs by the end of 2024. This substantial capacity positions Meta uniquely in the AI landscape.
The AGI focus stems from Meta's recent release of Llama 2, a large language model. Zuckerberg emphasizes that AGI's arrival will likely be a gradual process, and Meta's expanded AGI focus is influenced by the diverse capabilities of Llama models. While training Llama 3, Meta aims to achieve state-of-the-art models with advanced reasoning and planning abilities.
Addressing the debate on open versus closed AI development, Zuckerberg leans towards open sourcing for accessibility and opportunity. He contrasts Meta's approach with OpenAI, emphasizing the value of sharing and transparency. The ultimate control over AGI, if achieved, remains in Zuckerberg's hands, and while leaning towards open sourcing, he remains non-committal, emphasizing the importance of safety and responsibility in decision-making.
Stability AI unveils the Stable LM 2 1.6B, a cutting-edge small language model trained on multilingual data. The model outperforms rival models under 2B on most tasks, and even some larger ones, while offering compact size and speed when tested with few-shot performance. The release includes both pre-trained and instruction-tuned versions. Available for commercial and non-commercial use through the Stability AI Membership, the model is available for testing on Hugging Face.
Embedding models are being used to minimize training costs. Read more in VentureBeat's new article on the untapped potential of embedding models beyond simple retrieval.
Apple continues to make strides in open-source Language Models with the release of Autoregressive Image Models (AIM) this week, marking its second release in the last 3 weeks. Taking inspiration from large language models (LLMs), AIM introduces a scalable method for pre-training vision models without supervision. The 7 billion parameter model achieved an impressive 84.0% accuracy on ImageNet-1k, showcasing scalability without saturation in performance.
Curated List of Projects Released This Week
GPT-SoVITS: TTS Model providing an instant text-to-speech conversion with a short 5-second vocal sample.
SST AI App: A proof-of-concept for an advanced movie search application showcasing the capabilities of AI through integration with data and infrastructure. The project was built to demonstrate AI’s uses in real-life applications beyond chatbots.
WhisperSpeech: An open-source text-to-speech system built by inverting Whisper. The system consists of a complex pipeline of Whisper, (transcription) EnCodec (acoustic modeling), and Vocos (vocoder)
We trust you found this week's edition engaging and informative. As we navigate the early days of our newsletter, your feedback becomes invaluable. We welcome your thoughts and suggestions for future publications, so please feel free to connect with us on Twitter.
We appreciate your readership.