# forck.live — 10–16 June 2024 > AI lab announcements, news reports and podcasts. forck.live follows what companies announce about their own products. Press may corroborate an official story; it never leads and never appears as an item of its own. A missing date stays missing. A title is the publisher's words — where a company does not publish in English, an English rendering is listed separately as `Title in English` and named as this site's translation, never as theirs. This file is one issue of the weekly archive: every confirmed first-party announcement of ISO week 2024-W24, 10–16 June 2024, newest first. The window never moves, so this URL always names the same seven days — but a post published inside them and detected later is added when it arrives, so a closed week can still gain entries. Every entry is one first-party announcement, and carries the same eight fields in this order before any others: Company · Announced · Category · Coverage · Announcement · Group · Source · Record `Source` is the company's own page the announcement was traced to and `Record` is this site's page for it. Both are safe to cite. `Coverage` is a number of outlets, or `not counted` — which means this record reached no count of its own, never that nobody wrote about it. `Category` is this tracker's own analysis, and is authoritative wherever it is stated. `not stated` means either that nothing has analysed the post yet or that the analysis placed it outside this taxonomy; neither is a verdict about the announcement. `Announcement` is whether the post is news of something the company did, rather than a how-to, a customer story, a changelog listing or a staffing note. `Group` is which of this record's groups the entry falls in, carried on the entry itself so it holds where the groups are headings and where they are not. One of: `models`, `covered`, `announcements`, `routine`. Anything after `Record` is held detail that only some announcements carry. - This week's page: https://forck.live/week/2024-W24 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Replicate — Run Stable Diffusion 3 on your own machine with ComfyUI - Company: Replicate (replicate.com) - Announced: 2024-06-14 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/run-sd3-on-comfyui - Record: https://forck.live/items/2532-run-stable-diffusion-3-on-your-own-machine-with-comfyui - Subject: Platform Replicate published a guide on how to run Stable Diffusion 3 locally using ComfyUI on a machine with a GPU. ### Replicate — Replicate Intelligence #4 - Company: Replicate (replicate.com) - Announced: 2024-06-14 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/replicate-intelligence-2024-06-14 - Record: https://forck.live/items/2531-replicate-intelligence-4 - Subject: Platform This is a weekly bulletin from Replicate covering open-source AI models, tools, and research. It highlights the release of Stable Diffusion 3 Medium, a new image generation model with a 2B parameter size, available under a non-commercial license. It also mentions tools like finding concepts in GPT models using dictionary learning, real-time speech-to-text in the browser with Whisper, and research on a new way to tokenize images. Additionally, it announces upcoming support for NVIDIA H100 GPUs. ### Replicate — Push a custom version of Stable Diffusion 3 - Company: Replicate (replicate.com) - Announced: 2024-06-14 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replicate.com/blog/push-a-custom-sd3 - Record: https://forck.live/items/2530-push-a-custom-version-of-stable-diffusion-3 - Subject: Platform Replicate published a guide showing how to push a custom version of Stable Diffusion 3 to their platform using Cog, an open-source tool for packaging ML models in Docker containers. ### OpenAI — OpenAI appoints Retired U.S. Army General Paul M. Nakasone to Board of Directors - Company: OpenAI (openai.com) - Announced: 2024-06-13T14:00:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/openai-appoints-retired-us-army-general - Record: https://forck.live/items/746-openai-appoints-retired-u-s-army-general-paul-m-nakasone-to-board-of-directors - Subject: GPT / ChatGPT / API OpenAI appoints retired U.S. Army General Paul M. Nakasone to its Board of Directors, bringing cybersecurity experience and joining the Board's Safety and Security Committee. ### Vercel — Getting started with AI: Advice from the experts at Vercel Ship - Company: Vercel (vercel.com) - Announced: 2024-06-13T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/getting-started-with-ai-advice-from-the-experts-at-vercel-ship - Record: https://forck.live/items/20972-getting-started-with-ai-advice-from-the-experts-at-vercel-ship - Subject: v0 / AI SDK / AI Gateway Vercel hosted an AI Enterprise Panel at its Vercel Ship 2024 conference featuring experts from Google, Groq, and OpenAI. The panel discussed how enterprises can leverage AI to improve efficiency, enhance user experiences, and unlock new possibilities, emphasizing the importance of starting with data, developing strong evaluation frameworks, and embracing transparency. ### Hugging Face — From DeepSpeed to FSDP and Back Again with Hugging Face Accelerate - Company: Hugging Face (huggingface.co) - Announced: 2024-06-13 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/deepspeed-to-fsdp-and-back - Record: https://forck.live/items/1871-from-deepspeed-to-fsdp-and-back-again-with-hugging-face-accelerate - Subject: Platform Hugging Face Accelerate introduces a precision-related change and a concept guide to harmonize FSDP and DeepSpeed backends, enabling users to seamlessly switch between them. The change, included in the 0.30.0 release, allows FSDP to operate in a mixed-precision mode similar to DeepSpeed or a low-precision mode for memory-constrained scenarios. ### Sakana AI — Can LLMs invent better ways to train LLMs? - Company: Sakana AI (sakana.ai) - Announced: 2024-06-12T15:00:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://sakana.ai/llm-squared - Record: https://forck.live/items/4730-can-llms-invent-better-ways-to-train-llms - Subject: Sakana models Sakana AI releases a research report on using LLMs to discover preference optimization algorithms, including the state-of-the-art DiscoPOP algorithm, and open-sources the codebase and model checkpoints. ### Vercel — Demystifying INP: New tools and actionable insights - Company: Vercel (vercel.com) - Announced: 2024-06-12T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/demystifying-inp-new-tools-and-actionable-insights - Record: https://forck.live/items/20973-demystifying-inp-new-tools-and-actionable-insights - Subject: v0 / AI SDK / AI Gateway Vercel's guide explains how to optimize Interaction to Next Paint (INP) on web pages, covering the metric's phases and techniques to shorten event handler and browser render durations. ### Suno — Audio Inputs - Company: Suno (suno.com) - Announced: 2024-06-12 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://suno.com/blog/audio-inputs - Record: https://forck.live/items/10349-audio-inputs - Subject: Music / Studio Suno released an Audio Input feature that lets Pro and Premier users upload or record audio clips of 6–60 seconds to create songs. Copyrighted works are blocked from upload, and inputs with vocals remain private and unsearchable. ### Replicate — Run Stable Diffusion 3 with an API - Company: Replicate (replicate.com) - Announced: 2024-06-12 - Category: api-release - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/run-stable-diffusion-3-with-an-api - Record: https://forck.live/items/2534-run-stable-diffusion-3-with-an-api - Subject: Platform Replicate announces that Stable Diffusion 3 is now available via their API, with one-line code examples in JavaScript, Python, and cURL. ### Replicate — H100s are coming to Replicate - Company: Replicate (replicate.com) - Announced: 2024-06-12 - Category: infrastructure-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replicate.com/blog/h100s-are-coming - Record: https://forck.live/items/2533-h100s-are-coming-to-replicate - Subject: Platform Replicate announces upcoming support for NVIDIA H100 GPUs on its platform, with early access available by email. ### Hugging Face — Putting RL back in RLHF - Company: Hugging Face (huggingface.co) - Announced: 2024-06-12 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/putting_rl_back_in_rlhf_with_rloo - Record: https://forck.live/items/1873-putting-rl-back-in-rlhf - Subject: Platform Hugging Face introduces the RLOO (REINFORCE Leave One-Out) Trainer in TRL, a new online RLHF training algorithm that is more memory-efficient and faster than PPO. ### Hugging Face — Diffusers welcomes Stable Diffusion 3 - Company: Hugging Face (huggingface.co) - Announced: 2024-06-12 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/sd3 - Record: https://forck.live/items/1872-diffusers-welcomes-stable-diffusion-3 - Subject: Platform Stable Diffusion 3 Medium (2B parameters) is now available on the Hugging Face Hub and integrated with the Diffusers library, including memory optimizations and training scripts. ### Vercel — Never drop the illusion: How Frame.io builds fluid user experiences - Company: Vercel (vercel.com) - Announced: 2024-06-11T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/frameio-never-drop-the-illusion - Record: https://forck.live/items/20974-never-drop-the-illusion-how-frame-io-builds-fluid-user-experiences - Subject: v0 / AI SDK / AI Gateway Frame.io, an Adobe company, describes its 'Fluid UI' approach to building performant web applications for video professionals, based on principles of instant, smooth, and coordinated interactions. The company uses Vercel's automatic Preview Deployments, which reduced preview build times from up to 40 minutes to about 6 minutes, enabling faster iteration and performance regression detection. ### OpenAI — OpenAI and Apple announce partnership - Company: OpenAI (openai.com) - Announced: 2024-06-10T11:55:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/openai-and-apple-announce-partnership - Record: https://forck.live/items/747-openai-and-apple-announce-partnership - Subject: GPT / ChatGPT / API OpenAI and Apple announce a partnership to integrate ChatGPT into Apple experiences. ### OpenAI — OpenAI welcomes Sarah Friar (CFO) and Kevin Weil (CPO) - Company: OpenAI (openai.com) - Announced: 2024-06-10T10:30:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/openai-welcomes-cfo-cpo - Record: https://forck.live/items/748-openai-welcomes-sarah-friar-cfo-and-kevin-weil-cpo - Subject: GPT / ChatGPT / API OpenAI announces the appointment of Sarah Friar as CFO and Kevin Weil as CPO. ### LG AI Research — [ICLR 2024] Compression of Neural Networks Research Trends - Company: LG AI Research (lgresearch.ai) - Announced: 2024-06-10 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=436 - Record: https://forck.live/items/4851-iclr-2024-compression-of-neural-networks-research-trends - Subject: EXAONE The blog post from LG AI Research presents two research papers on neural network compression presented at ICLR 2024: LESS, which combines filter pruning and low-rank decomposition for vision tasks, and SHEARED LLAMA, which applies structured pruning to LLMs.