# forck.live — 29 January – 4 February 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-W05, 29 January – 4 February 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-W05 - The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md - Catalogue: https://forck.live/llms.txt ## Announcements ### Alibaba — Introducing Qwen1.5 - Company: Alibaba (alibaba.com) - Announced: 2024-02-04T05:33:00+00:00 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://qwenlm.github.io/blog/qwen1.5/ - Record: https://forck.live/items/2319-introducing-qwen1-5 - Subject: Qwen Alibaba open-sourced Qwen1.5 base and chat models in multiple sizes (0.5B to 110B) and an MoE model. ### Hugging Face — SegMoE: Segmind Mixture of Diffusion Experts - Company: Hugging Face (huggingface.co) - Announced: 2024-02-03 - Category: open-weight-release - Coverage: not counted - Announcement: yes - Group: models - Source: https://huggingface.co/blog/segmoe - Record: https://forck.live/items/1944-segmoe-segmind-mixture-of-diffusion-experts - Subject: Platform SegMoE is a framework for creating Mixture-of-Experts Diffusion models. The blog announces the release of three SegMoE models (SegMoE-4x2, SegMoE-2x1, SegMoE-SD4x2) and the segmoe package for creating custom MoE models, all integrated with Hugging Face diffusers. ### LG AI Research — [NeurIPS 2023] Multimodal Generation for Audio and Music - Company: LG AI Research (lgresearch.ai) - Announced: 2024-02-02 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://www.lgresearch.ai/blog/view?seq=398 - Record: https://forck.live/items/4872-neurips-2023-multimodal-generation-for-audio-and-music - Subject: EXAONE Multimodal generation, highlighting LG AI Research's paper on music loop generation and mentioning Google DeepMind's Lyria and Meta's Audiobox. ### Hugging Face — NPHardEval Leaderboard: Unveiling the Reasoning Abilities of Large Language Models through Complexity Classes and Dynamic Updates - Company: Hugging Face (huggingface.co) - Announced: 2024-02-02 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/leaderboard-nphardeval - Record: https://forck.live/items/1945-nphardeval-leaderboard-unveiling-the-reasoning-abilities-of-large-language - Subject: Platform Hugging Face introduces the NPHardEval leaderboard, a dynamic benchmark for evaluating LLM reasoning abilities using complexity classes, with 900 algorithmic questions updated monthly. ### Sakana AI — We received a supercomputing grant from the Japanese government - Company: Sakana AI (sakana.ai) - Announced: 2024-02-01T15:00:00+00:00 - Category: funding-company-news - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://sakana.ai/nedo-grant - Record: https://forck.live/items/4731-we-received-a-supercomputing-grant-from-the-japanese-government - Subject: Sakana models Sakana AI announces it is one of seven institutions in Japan selected by the Japanese government to receive a supercomputing grant, organized by NEDO under METI, to support development of foundation AI models. ### LG AI Research — LG AI Research's Perspective on CES 2024 and the Future of AI Technology - Company: LG AI Research (lgresearch.ai) - Announced: 2024-02-01 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/news/view?seq=397 - Record: https://forck.live/items/11036-lg-ai-research-s-perspective-on-ces-2024-and-the-future-of-ai-technology - Subject: EXAONE LG AI Research published a perspective on CES 2024, analyzing AI technology trends showcased at the event. The analysis highlights three key takeaways: on-device AI and user experience transformation, activities of Korean companies at the startup exhibition, and the combination of generative AI with existing services. The document discusses keynote speeches from industry leaders including Intel, Qualcomm, L'Oréal, Walmart, and Microsoft, and describes demonstrations of on-device AI capabilities and new hardware innovations. ### LG AI Research — LG AI Research's Perspective on CES 2024 and the Future of AI Technology - Company: LG AI Research (lgresearch.ai) - Announced: 2024-02-01 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://www.lgresearch.ai/blog/view?seq=397 - Record: https://forck.live/items/4873-lg-ai-research-s-perspective-on-ces-2024-and-the-future-of-ai-technology - Subject: EXAONE LG AI Research summarizes key AI technology trends observed at CES 2024, focusing on on-device AI, Korean companies' activities, and the integration of generative AI into existing services. The post highlights Qualcomm's on-device AI demos and Microsoft's introduction of a Copilot key for keyboards, but does not announce any new AI product or model from LG AI Research itself. ### Hugging Face — Patch Time Series Transformer in Hugging Face - Company: Hugging Face (huggingface.co) - Announced: 2024-02-01 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/patchtst - Record: https://forck.live/items/1948-patch-time-series-transformer-in-hugging-face - Subject: Platform Blog post providing a tutorial on getting started with the PatchTST model for time series forecasting, including installation, training on the Electricity dataset, and zero-shot transfer learning on the ETTh1 dataset. ### Hugging Face — Hugging Face Text Generation Inference available for AWS Inferentia2 - Company: Hugging Face (huggingface.co) - Announced: 2024-02-01 - Category: availability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/text-generation-inference-on-inferentia2 - Record: https://forck.live/items/1947-hugging-face-text-generation-inference-available-for-aws-inferentia2 - Subject: Platform Hugging Face announces general availability of Text Generation Inference (TGI) on AWS Inferentia2 and Amazon SageMaker, enabling deployment of open LLMs like Zephyr 7B on AWS Inferentia2. ### Hugging Face — Constitutional AI with Open LLMs - Company: Hugging Face (huggingface.co) - Announced: 2024-02-01 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/constitutional_ai - Record: https://forck.live/items/1946-constitutional-ai-with-open-llms - Subject: Platform Hugging Face presents an end-to-end recipe for Constitutional AI using open models, releasing a new tool called llm-swarm for scalable synthetic data generation on GPU Slurm clusters, along with datasets and fine-tuned Mistral-7B models based on Anthropic's and Grok's constitutions. ### Vercel — How streaming helps build faster web applications - Company: Vercel (vercel.com) - Announced: 2024-01-31T13:00:00+00:00 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://vercel.com/blog/how-streaming-helps-build-faster-web-applications - Record: https://forck.live/items/21014-how-streaming-helps-build-faster-web-applications - Subject: v0 / AI SDK / AI Gateway Vercel explains how streaming in the Next.js App Router, using React Suspense, allows progressive UI delivery from server to client without waiting for all data to load, improving performance and user experience. The post also covers how streaming benefits AI applications by displaying parts of LLM responses as they become available, reducing perceived wait times. ### OpenAI — Building an early warning system for LLM-aided biological threat creation - Company: OpenAI (openai.com) - Announced: 2024-01-31T08:00:00+00:00 - Category: research-paper - Coverage: not counted - Announcement: no - Group: routine - Source: https://openai.com/index/building-an-early-warning-system-for-llm-aided-biological-threat-creation - Record: https://forck.live/items/810-building-an-early-warning-system-for-llm-aided-biological-threat-creation - Subject: GPT / ChatGPT / API OpenAI is developing a blueprint for evaluating the risk that an LLM could aid in creating a biological threat. An evaluation with GPT-4 found at most a mild uplift in biological threat creation accuracy, which is not conclusive but serves as a starting point for further research. ### Supabase — NoSQL Postgres: Add MongoDB compatibility to your Supabase projects with FerretDB - Company: Supabase (supabase.com) - Announced: 2024-01-31 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://supabase.com/blog/nosql-mongodb-compatibility-with-ferretdb-and-flydotio - Record: https://forck.live/items/21448-nosql-postgres-add-mongodb-compatibility-to-your-supabase-projects-with-ferretdb - Subject: Supabase Supabase published a guide showing how to add MongoDB compatibility to Supabase projects using FerretDB, an open-source document database that translates MongoDB queries to Postgres. The post walks through running FerretDB locally via Docker, testing with mongosh, and deploying to Fly.io for production use. ### Hugging Face — Introducing the Enterprise Scenarios Leaderboard: a Leaderboard for Real World Use Cases - Company: Hugging Face (huggingface.co) - Announced: 2024-01-31 - Category: developer-tool-release - Coverage: not counted - Announcement: no - Group: routine - Source: https://huggingface.co/blog/leaderboard-patronus - Record: https://forck.live/items/1949-introducing-the-enterprise-scenarios-leaderboard-a-leaderboard-for-real-world - Subject: Platform Patronus announces the Enterprise Scenarios Leaderboard, built on Hugging Face's template, to evaluate language models on real-world enterprise tasks. ### Supabase — pgvector 0.6.0: 30x faster with parallel index builds - Company: Supabase (supabase.com) - Announced: 2024-01-30 - Category: developer-tool-release - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://supabase.com/blog/pgvector-fast-builds - Record: https://forck.live/items/21449-pgvector-0-6-0-30x-faster-with-parallel-index-builds - Subject: Supabase Supabase announced pgvector 0.6.0, which adds parallel builds for HNSW indexes. Building an HNSW index is now up to 30x faster for unlogged tables and 7–9 times faster for logged tables on a 4XL instance. The release also allows tuning of HNSW build parameters without affecting search accuracy or queries per second. ### Replicate — Run Code Llama 70B with an API - Company: Replicate (replicate.com) - Announced: 2024-01-30 - Category: api-release - Coverage: not counted - Announcement: no - Group: routine - Source: https://replicate.com/blog/run-codellama-with-an-api - Record: https://forck.live/items/2541-run-code-llama-70b-with-an-api - Subject: Platform Replicate announces that Code Llama 70B is now available via its API, with code examples for JavaScript, Python, and cURL, and mentions three variants: Base, Python, and Instruct. ### Hugging Face — Accelerate StarCoder with 🤗 Optimum Intel on Xeon: Q8/Q4 and Speculative Decoding - Company: Hugging Face (huggingface.co) - Announced: 2024-01-30 - Category: capability-change - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/intel-starcoder-quantization - Record: https://forck.live/items/1950-accelerate-starcoder-with-optimum-intel-on-xeon-q8-q4-and-speculative-decoding - Subject: Platform Hugging Face and Intel present optimizations for StarCoder-15B on 4th gen Xeon, achieving over 7x inference acceleration through 8-bit and 4-bit quantization combined with assisted generation (speculative decoding). The work uses SmoothQuant for INT8 quantization and demonstrates significant speedup without accuracy loss on the HumanEval benchmark. ### Supabase — Getting started with Ruby on Rails and Postgres on Supabase - Company: Supabase (supabase.com) - Announced: 2024-01-29 - Category: not stated - Coverage: not counted - Announcement: no - Group: routine - Source: https://supabase.com/blog/ruby-on-rails-postgres - Record: https://forck.live/items/21450-getting-started-with-ruby-on-rails-and-postgres-on-supabase - Subject: Supabase Every Supabase project comes with a full Postgres database, a free and open source database which is considered one of the world's most stable and advanced databases. Postgres is an ideal choice for your Ruby on Rails applications as Rails ships with a built-in Postgres adapter! ### Hugging Face — The Hallucinations Leaderboard, an Open Effort to Measure Hallucinations in Large Language Models - Company: Hugging Face (huggingface.co) - Announced: 2024-01-29 - Category: research-paper - Coverage: not counted - Announcement: yes - Group: announcements - Source: https://huggingface.co/blog/leaderboard-hallucinations - Record: https://forck.live/items/1951-the-hallucinations-leaderboard-an-open-effort-to-measure-hallucinations-in - Subject: Platform Hugging Face announces the Hallucinations Leaderboard, a comprehensive platform evaluating LLMs on hallucination-related benchmarks using in-context learning, with tasks including closed-book QA, summarisation, reading comprehension, instruction following, fact-checking, hallucination detection, and self-consistency.