# forck.live — 22–28 May 2023 > 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 2023-W21, 22–28 May 2023, 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/2023-W21 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Replicate — Make any large language model a better poet - Company: Replicate (replicate.com) - Announced: 2023-05-26 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://replicate.com/blog/turn-your-llm-into-a-poet - Record: https://forck.live/items/2566-make-any-large-language-model-a-better-poet - Subject: Platform Replicate released Poet Vicuna-13B, a modified version of Vicuna-13B designed to generate poems and lyrics with specific syllabic patterns. ### Replit — Super Colliding Nix Stores: Nix Flakes for Millions of Developers - Company: Replit (replit.com) - Announced: 2023-05-25T17:40:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/super-colliding-nix-stores - Record: https://forck.live/items/18066-super-colliding-nix-stores-nix-flakes-for-millions-of-developers - Subject: Replit Agent Replit partnered with Obsidian Systems and Tweag to develop a method for merging multiple Nix stores, including remote ones. This capability aims to bring Nix Flakes and portable development environments to Replit's user base, allowing developers to write a Nix Flake once and use the same reproducible environment across platforms. ### OpenAI — Democratic inputs to AI - Company: OpenAI (openai.com) - Announced: 2023-05-25T07:00:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/democratic-inputs-to-ai - Record: https://forck.live/items/859-democratic-inputs-to-ai - Subject: GPT / ChatGPT / API OpenAI is launching a grant program that will award ten $100,000 grants to fund experiments on democratic processes for determining AI system rules within legal boundaries. ### Hugging Face — Optimizing Stable Diffusion for Intel CPUs with NNCF and 🤗 Optimum - Company: Hugging Face (huggingface.co) - Announced: 2023-05-25 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/train-optimize-sd-intel - Record: https://forck.live/items/2053-optimizing-stable-diffusion-for-intel-cpus-with-nncf-and-optimum - Subject: Platform A workflow for optimizing Stable Diffusion models for Intel CPUs using OpenVINO NNCF and Optimum, achieving 5.1x inference acceleration and 4x model footprint reduction compared to PyTorch. The method involves quantization-aware training, knowledge distillation, and token merging. ### Hugging Face — Hugging Face Collaborates with Microsoft to launch Hugging Face Model Catalog on Azure - Company: Hugging Face (huggingface.co) - Announced: 2023-05-24 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/hugging-face-endpoints-on-azure - Record: https://forck.live/items/2055-hugging-face-collaborates-with-microsoft-to-launch-hugging-face-model-catalog - Subject: Platform Hugging Face collaborates with Microsoft to launch a Hugging Face Model Catalog within Azure Machine Learning Studio, enabling deployment of open-source models from the Hugging Face Hub on Azure infrastructure. ### Hugging Face — Making LLMs even more accessible with bitsandbytes, 4-bit quantization and QLoRA - Company: Hugging Face (huggingface.co) - Announced: 2023-05-24 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/4bit-transformers-bitsandbytes - Record: https://forck.live/items/2054-making-llms-even-more-accessible-with-bitsandbytes-4-bit-quantization-and-qlora - Subject: Platform Hugging Face announces integration of 4-bit quantization into transformers using bitsandbytes, enabling running models in 4-bit precision and finetuning with QLoRA, as introduced in the QLoRA paper. ### Hugging Face — Instruction-tuning Stable Diffusion with InstructPix2Pix - Company: Hugging Face (huggingface.co) - Announced: 2023-05-23 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/instruction-tuning-sd - Record: https://forck.live/items/2058-instruction-tuning-stable-diffusion-with-instructpix2pix - Subject: Platform Instruction-tuning to teach Stable Diffusion to follow instructions to translate or process input images, extending the InstructPix2Pix training strategy to tasks like cartoonization and image deraining. It covers dataset preparation, training experiments, and results, and provides code, pre-trained models, and datasets. ### Hugging Face — 🐶Safetensors audited as really safe and becoming the default - Company: Hugging Face (huggingface.co) - Announced: 2023-05-23 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/safetensors-security-audit - Record: https://forck.live/items/2057-safetensors-audited-as-really-safe-and-becoming-the-default - Subject: Platform Following a security audit by Trail of Bits, Hugging Face, EleutherAI, and Stability AI plan to make safetensors the default format for saved models, citing its safety and speed advantages. ### Hugging Face — Hugging Face and IBM partner on watsonx.ai, the next-generation enterprise studio for AI builders - Company: Hugging Face (huggingface.co) - Announced: 2023-05-23 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/huggingface-and-ibm - Record: https://forck.live/items/2056-hugging-face-and-ibm-partner-on-watsonx-ai-the-next-generation-enterprise - Subject: Platform Hugging Face and IBM announce a partnership to collaborate on watsonx.ai, IBM's enterprise AI studio, integrating Hugging Face libraries and open-sourcing IBM's LLMs. ### OpenAI — Governance of superintelligence - Company: OpenAI (openai.com) - Announced: 2023-05-22T07:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/governance-of-superintelligence - Record: https://forck.live/items/860-governance-of-superintelligence - Subject: GPT / ChatGPT / API The text suggests that now is a good time to start thinking about the governance of superintelligence, which refers to future AI systems much more capable than AGI.