# forck.live — 11–17 December 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-W50, 11–17 December 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-W50 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Replit — Updates to Social Features on Replit - Company: Replit (replit.com) - Announced: 2023-12-15T22:16:19+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/social-updates - Record: https://forck.live/items/18017-updates-to-social-features-on-replit - Subject: Replit Agent Replit announced upcoming changes to its social features, including cover pages that will primarily display a project's README or code instead of running the program, removal of comments on cover pages, and removal of the embedded Repl component on profile pages. The changes aim to focus cover pages on code consumption rather than running and viewing program outputs. ### Replit — Skip the README, let us install for you - Company: Replit (replit.com) - Announced: 2023-12-15T17:00:00+00:00 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://replit.com/blog/system-dependencies-python - Record: https://forck.live/items/18018-skip-the-readme-let-us-install-for-you - Subject: Replit Agent Replit now automatically installs system dependencies for hundreds of Python packages when users add them via the Packages tool or through automatic package guessing. The system also handles dependencies for packages already listed in pyproject.toml, adding them to the replit.nix file. This eliminates the need to manually install system-level dependencies like pkg-config and cairo. ### LG AI Research — LG AI Research expands its AI Tech Leadership at NeurIPS 2023 - Company: LG AI Research (lgresearch.ai) - Announced: 2023-12-15 - Category: not stated - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/news/view?seq=371 - Record: https://forck.live/items/11038-lg-ai-research-expands-its-ai-tech-leadership-at-neurips-2023 - Subject: EXAONE LG AI Research continues to showcase its leadership in AI technology by presenting 19 papers at NeurIPS (Neural Information Processing Systems) 2023, the world’s largest machine learning conference. ### LG AI Research — LG AI Research expands its AI Tech Leadership at NeurIPS 2023 - Company: LG AI Research (lgresearch.ai) - Announced: 2023-12-15 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=371 - Record: https://forck.live/items/4882-lg-ai-research-expands-its-ai-tech-leadership-at-neurips-2023 - Subject: EXAONE LG AI Research presented 19 papers at NeurIPS 2023 and demonstrated EXAONE Universe, a generative AI service for AI researchers. ### OpenAI — Superalignment Fast Grants - Company: OpenAI (openai.com) - Announced: 2023-12-14T08:00:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/superalignment-fast-grants - Record: https://forck.live/items/821-superalignment-fast-grants - Subject: GPT / ChatGPT / API OpenAI is launching $10M in grants for technical research on AI alignment and safety, including weak-to-strong generalization, interpretability, and scalable oversight. ### OpenAI — Increasing accuracy of pediatric visit notes - Company: OpenAI (openai.com) - Announced: 2023-12-14T08:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/summer-health - Record: https://forck.live/items/819-increasing-accuracy-of-pediatric-visit-notes - Subject: GPT / ChatGPT / API Summer Health uses OpenAI to reimagine pediatric doctor's visits. ### Upstage — Upstage's Solar 10.7B Emerges as World's Top Pre-trained LLM - Company: Upstage (upstage.ai) - Announced: 2023-12-14 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.upstage.ai/news/solar-10-7b-emerges-as-worlds-top-pre-trained-llm - Record: https://forck.live/items/18450-upstage-s-solar-10-7b-emerges-as-world-s-top-pre-trained-llm - Subject: Solar Pro / Solar Mini Upstage announced that its Solar 10.7B large language model achieved the top position on the Hugging Face Open LLM Leaderboard with an average score of 74.2, surpassing models including GPT-3.5 Turbo, Llama 2, Qwen 72B, and Mixtral 8x7B. The model is pre-trained on over three trillion tokens using a depth up-scaling technique based on open-source 7B models. Solar 10.7B is available as an open-source LLM under the Apache 2.0 license. ### Cartesia — Mamba‑3B-SlimPJ: State-space models rivaling the best Transformer architecture - Company: Cartesia (cartesia.ai) - Announced: 2023-12-14 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://www.cartesia.ai/blog/mamba-3b-slimpj - Record: https://forck.live/items/8033-mamba-3b-slimpj-state-space-models-rivaling-the-best-transformer-architecture - Subject: Sonic / Ink Cartesia and Together released Mamba-3B-SlimPJ, a 2.8B parameter state-space model trained on 600B tokens of the SlimPajama dataset, under an Apache 2.0 license. The model matches the performance of the strongest comparable 3B Transformer models with 17% fewer training FLOPs. ### OpenAI — Weak-to-strong generalization - Company: OpenAI (openai.com) - Announced: 2023-12-14 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/weak-to-strong-generalization - Record: https://forck.live/items/822-weak-to-strong-generalization - Subject: GPT / ChatGPT / API OpenAI announces a new research direction called weak-to-strong generalization, which explores using the generalization properties of deep learning to enable weak supervisors to control strong models, with initial results. ### OpenAI — Partnership with Axel Springer to deepen beneficial use of AI in journalism - Company: OpenAI (openai.com) - Announced: 2023-12-13T08:00:00+00:00 - Category: partnership-acquisition - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://openai.com/index/axel-springer-partnership - Record: https://forck.live/items/823-partnership-with-axel-springer-to-deepen-beneficial-use-of-ai-in-journalism - Subject: GPT / ChatGPT / API OpenAI announces a partnership with Axel Springer, the first publishing house to integrate journalism more deeply with AI technologies. ### Runway — Introducing General World Models - Company: Runway (runway.com) - Announced: 2023-12-11T10:10:10.110000+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://runway.com/research/introducing-general-world-models - Record: https://forck.live/items/7930-introducing-general-world-models - Subject: Gen / Aleph Runway describes the concept of general world models and notes that its video generative system Gen-2 is a very early and limited form of such models. The company is building a team to address open research challenges in this area. ### Mistral AI — La Plateforme - Company: Mistral AI (mistral.ai) - Announced: 2023-12-11T07:00:00+00:00 - Category: api-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://mistral.ai/news/la-plateforme/ - Record: https://forck.live/items/7378-la-plateforme - Subject: Mistral / Le Chat Mistral AI opened beta access to its platform, La Plateforme, offering three chat endpoints (mistral-tiny, mistral-small, mistral-medium) and one embedding endpoint (mistral-embed). The endpoints serve different models, including Mistral 7B Instruct v0.2 and Mixtral 8x7B, with varying performance and pricing. Anyone can register for the beta. ### Mistral AI — Mixtral of experts - Company: Mistral AI (mistral.ai) - Announced: 2023-12-11T07:00:00+00:00 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://mistral.ai/news/mixtral-of-experts/ - Record: https://forck.live/items/7377-mixtral-of-experts - Subject: Mistral / Le Chat Mistral AI released Mixtral 8x7B, an open-weight sparse mixture of experts model under Apache 2.0 license. The model outperforms Llama 2 70B on most benchmarks and matches GPT3.5, with a 32k token context and multilingual support. An instruction-tuned variant, Mixtral 8x7B Instruct, was also released. ### Hugging Face — Mixture of Experts Explained - Company: Hugging Face (huggingface.co) - Announced: 2023-12-11 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/moe - Record: https://forck.live/items/1967-mixture-of-experts-explained - Subject: Platform Mixture of Experts (MoE) models, their history, training, and tradeoffs. ### Hugging Face — Welcome Mixtral - a SOTA Mixture of Experts on Hugging Face - Company: Hugging Face (huggingface.co) - Announced: 2023-12-11 - Category: new-model - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/mixtral - Record: https://forck.live/items/1966-welcome-mixtral-a-sota-mixture-of-experts-on-hugging-face - Subject: Platform Mistral released Mixtral 8x7b, a Mixture-of-Experts open-access model that outperforms GPT-3.5 and Llama 2 70B on many benchmarks, with Apache 2.0 license and 32k context length. Hugging Face announced comprehensive integration including models on the Hub, Transformers, Inference Endpoints, and Text Generation Inference.