# forck.live — 3–9 April 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-W14, 3–9 April 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-W14 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Replit — Hackers, Pros, and Teams users can now code for hours without restarts - Company: Replit (replit.com) - Announced: 2023-04-06 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/regular-vms - Record: https://forck.live/items/18075-hackers-pros-and-teams-users-can-now-code-for-hours-without-restarts - Subject: Replit Agent Replit has upgraded all Hacker, Pro, and Teams user VMs from spot rate to regular provisioning when coding in the Workspace, resulting in a 10x reduction in container restarts. Users previously experienced at least one restart per hour but can now code for multiple hours straight without restarts, with only 1-2 restarts per day expected. ### Hugging Face — Snorkel AI x Hugging Face: unlock foundation models for enterprises - Company: Hugging Face (huggingface.co) - Announced: 2023-04-06 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/snorkel-case-study - Record: https://forck.live/items/2077-snorkel-ai-x-hugging-face-unlock-foundation-models-for-enterprises - Subject: Platform Snorkel AI and Hugging Face announce a partnership to integrate Hugging Face's Inference Endpoint service into Snorkel Flow, enabling enterprises to easily access and use a wide range of open-source foundation models for their AI applications. ### Replit — Replit + Chroma: AI for the next billion software creators - Company: Replit (replit.com) - Announced: 2023-04-05T12:21:40+00:00 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/chroma - Record: https://forck.live/items/18077-replit-chroma-ai-for-the-next-billion-software-creators - Subject: Replit Agent Replit and Chroma announced a Chroma template for Replit that lets developers build AI applications with state and memory using embeddings. The template includes a demo for question-answering using OpenAI's ChatGPT API, with an embedded textbook for conversational queries. ### OpenAI — Our approach to AI safety - Company: OpenAI (openai.com) - Announced: 2023-04-05T07:00:00+00:00 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/our-approach-to-ai-safety - Record: https://forck.live/items/865-our-approach-to-ai-safety - Subject: GPT / ChatGPT / API OpenAI states that ensuring the safe building, deployment, and use of AI systems is critical to its mission. ### Replit — Replit Bounties: The Best Place to Build and Launch MVPs - Company: Replit (replit.com) - Announced: 2023-04-05 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replit.com/blog/bounties-christian-ulstrup-case-study - Record: https://forck.live/items/18129-replit-bounties-the-best-place-to-build-and-launch-mvps - Subject: Replit Agent Christian Ulstrup, founder and CEO of GSD @ Work, used Replit Bounties to build an automated tool that processes meeting transcripts and summarizes them for social media using GPT-3. The project was completed in two bounties totaling $50, with the first delivered in 5 hours. Ulstrup reported that the tool made his workflow "10x faster" and allowed him to delegate note-taking to an AI pipeline. ### Replicate — Language models are on Replicate - Company: Replicate (replicate.com) - Announced: 2023-04-05 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replicate.com/blog/language-models - Record: https://forck.live/items/2570-language-models-are-on-replicate - Subject: Platform Replicate announces that users can now deploy, run, and fine-tune large language models on its platform, with official versions of FLAN-T5, GPT-J, and LLaMA, and a preview of fine-tuning. ### Hugging Face — StackLLaMA: A hands-on guide to train LLaMA with RLHF - Company: Hugging Face (huggingface.co) - Announced: 2023-04-05 - Category: new-model - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/stackllama - Record: https://forck.live/items/2078-stackllama-a-hands-on-guide-to-train-llama-with-rlhf - Subject: Platform Hugging Face releases the StackLLaMA model, a version of LLaMA fine-tuned with RLHF on Stack Exchange data, and provides a detailed guide on the training process including supervised fine-tuning, reward modeling, and reinforcement learning. ### Replit — April 2 Potential GitHub Credentials Exposure - Company: Replit (replit.com) - Announced: 2023-04-04T02:00:00+00:00 - Category: safety-policy-update - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/april-02-potential-github-credentials-exposure - Record: https://forck.live/items/18120-april-2-potential-github-credentials-exposure - Subject: Replit Agent Replit disclosed a vulnerability that may have exposed GitHub authentication tokens for less than 0.01% of its users, stemming from the GitHub import feature. The tokens were written to the git reflog in public or HTTP-served Repls, potentially allowing unauthorized read/write access to repositories. Replit fixed the vulnerability, revoked all existing tokens, and notified affected users. ### LG AI Research — Diving into the journey of the 2nd LG Aimers and the AI Hackathon - Company: LG AI Research (lgresearch.ai) - Announced: 2023-04-03 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/news/view?seq=318 - Record: https://forck.live/items/11051-diving-into-the-journey-of-the-2nd-lg-aimers-and-the-ai-hackathon - Subject: EXAONE LG AI Research concluded the 2nd LG Aimers program, which included an AI Hackathon focused on developing a product quality classification AI model for smart factory control systems. The competition involved 495 teams and 898 participants in online preliminaries, with top teams advancing to offline finals held at LG Academy. The winning team, 'Noah's Ark,' proposed a two-part model combining Regressor and Classification to improve data detection performance. ### LG AI Research — Diving into the journey of the 2nd LG Aimers and the AI Hackathon - Company: LG AI Research (lgresearch.ai) - Announced: 2023-04-03 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/blog/view?seq=318 - Record: https://forck.live/items/4901-diving-into-the-journey-of-the-2nd-lg-aimers-and-the-ai-hackathon - Subject: EXAONE LG AI Research reports on the conclusion of the 2nd LG Aimers program, which combined theoretical AI education with a practical AI Hackathon using real business data. The hackathon focused on developing a product quality classification AI model for smart factory control systems. The top three teams were awarded prizes.