# forck.live — 13–19 March 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-W11, 13–19 March 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-W11 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Replicate — Week 3 of LLaMA 🦙 - Company: Replicate (replicate.com) - Announced: 2023-03-18 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/llama-roundup - Record: https://forck.live/items/2572-week-3-of-llama - Subject: Platform A roundup of community projects and tools built around Meta's LLaMA model in its third week since release, highlighting the rapid ecosystem growth. ### Replicate — Fine-tune LLaMA to speak like Homer Simpson - Company: Replicate (replicate.com) - Announced: 2023-03-17 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/fine-tune-llama-to-speak-like-homer-simpson - Record: https://forck.live/items/2573-fine-tune-llama-to-speak-like-homer-simpson - Subject: Platform Fine-tuning LLaMA on a TV show script can make it generate text in the voice of the characters, specifically a Homer Simpson bot, using ~60K lines of dialogue and 90 minutes of training. ### Replit — Worldwide Repls, part 3: Firing Up The Engines - Company: Replit (replit.com) - Announced: 2023-03-16T16:00:00+00:00 - Category: infrastructure-release - Coverage: not counted - Announcement: no - Group: routine - Source: https://replit.com/blog/geo-part-3-deploy - Record: https://forck.live/items/18111-worldwide-repls-part-3-firing-up-the-engines - Subject: Replit Agent Replit describes improvements to its cloud infrastructure, including dividing into failure domains with clusters, migrating global state management from Redis to an SQL-backed Control Plane, and building a custom load balancer for assigning Repls to machines. The post also covers challenges with geographic distribution, such as load balancer spillover causing suboptimal connections for users. ### Replicate — Train and run Stanford Alpaca on your own machine - Company: Replicate (replicate.com) - Announced: 2023-03-16 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/replicate-alpaca - Record: https://forck.live/items/2574-train-and-run-stanford-alpaca-on-your-own-machine - Subject: Platform A blog post from Replicate providing a step-by-step guide to training and running Stanford Alpaca, a fine-tuned version of LLaMA, on one's own machine, using LLaMA weights and the Alpaca training script. ### LG AI Research — Can Large Language Models Improve Few-shot Retrieval for Complex Question Answering? - Company: LG AI Research (lgresearch.ai) - Announced: 2023-03-15 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=314 - Record: https://forck.live/items/4903-can-large-language-models-improve-few-shot-retrieval-for-complex-question - Subject: EXAONE LG AI Research proposes PromptRank, a data-efficient retrieval method for multi-hop question answering that uses a large language model to rerank document paths based on the conditional likelihood of generating the question given the path. ### OpenAI — GPT-4 - Company: OpenAI (openai.com) - Announced: 2023-03-14T07:00:00+00:00 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://openai.com/index/gpt-4 - Record: https://forck.live/items/874-gpt-4 - Subject: GPT / ChatGPT / API OpenAI announced GPT-4, its most advanced language model, which is 82% less likely to respond to requests for disallowed content and 40% more likely to produce factual responses than GPT-3.5. The model was made available on ChatGPT Plus and as an API for developers. ### OpenAI — Preserving languages for the future - Company: OpenAI (openai.com) - Announced: 2023-03-14T07:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/government-of-iceland - Record: https://forck.live/items/873-preserving-languages-for-the-future - Subject: GPT / ChatGPT / API Iceland is using GPT-4 to preserve its language. ### OpenAI — Transforming visual accessibility - Company: OpenAI (openai.com) - Announced: 2023-03-14T07:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/be-my-eyes - Record: https://forck.live/items/872-transforming-visual-accessibility - Subject: GPT / ChatGPT / API OpenAI announces that Be My Eyes uses GPT-4 to transform visual accessibility. ### OpenAI — Stripe - Company: OpenAI (openai.com) - Announced: 2023-03-14T07:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/stripe - Record: https://forck.live/items/871-stripe - Subject: GPT / ChatGPT / API Stripe is using GPT-4 to improve user experience and fight fraud. ### OpenAI — Filling crucial language learning gaps - Company: OpenAI (openai.com) - Announced: 2023-03-14T07:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/duolingo - Record: https://forck.live/items/870-filling-crucial-language-learning-gaps - Subject: GPT / ChatGPT / API GPT-4 is used to deepen conversations on Duolingo. ### OpenAI — Powering virtual education for the classroom - Company: OpenAI (openai.com) - Announced: 2023-03-14T07:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/khan-academy - Record: https://forck.live/items/869-powering-virtual-education-for-the-classroom - Subject: GPT / ChatGPT / API Khan Academy is launching a limited pilot program to explore the potential of GPT-4. ### LG AI Research — Research Trends in Causal Representation Learning - Company: LG AI Research (lgresearch.ai) - Announced: 2023-03-14 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=312 - Record: https://forck.live/items/4904-research-trends-in-causal-representation-learning - Subject: EXAONE The paper 'Towards Causal Representation Learning' and discusses how causal inference can address limitations in machine learning, such as robustness, learning reusable mechanisms, and causality perspective.