# forck.live — 4–10 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-W49, 4–10 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-W49
- The current issue, and the catalogue of the other weeks: https://forck.live/briefing.md
- Catalogue: https://forck.live/llms.txt

## Announcements

### Replit — New and Improved Console

- Company: Replit (replit.com)
- Announced: 2023-12-07T18:00:40+00:00
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://replit.com/blog/new-and-improved-console
- Record: https://forck.live/items/18019-new-and-improved-console
- Subject: Replit Agent

Replit redesigned its Console pane with a new UI, persisted execution history, metadata per run, and integration with the Package Manager and Debugger. The new Console only shows code execution results, creates separate blocks for each run, and preserves history across runs.

### Sarvam AI — Sarvam AI secures a landmark $41 million in Series A funding

- Company: Sarvam AI (sarvam.ai)
- Announced: 2023-12-07T12:00:00+00:00
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.sarvam.ai/blogs/announcing-series-a/
- Record: https://forck.live/items/18408-sarvam-ai-secures-a-landmark-41-million-in-series-a-funding
- Subject: Sarvam / Bulbul / Saaras

Sarvam AI, an Indian generative AI startup, announced a $41 million Series A funding round led by Lightspeed, with support from Peak XV Partners and Khosla Ventures. The company plans to develop a full-stack generative AI platform, focusing on Indian languages, voice-first interfaces, and domain-specific models for enterprises, as well as public-good applications.

### LG AI Research — EXAONE: Expert AI for Everyone, Bringing Digital Innovation - 2023 Korea Government Expo Report

- Company: LG AI Research (lgresearch.ai)
- Announced: 2023-12-07
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/news/view?seq=363
- Record: https://forck.live/items/11039-exaone-expert-ai-for-everyone-bringing-digital-innovation-2023-korea
- Subject: EXAONE

LG AI Research demonstrated three EXAONE platforms at the 2023 Korea Government Expo in Busan, showcasing AI technology applications across industries and the public sector. EXAONE Universe provides answers via a super-giant language model; EXAONE Atelier, a multimodal model trained on 350 million image-text pairs, generates creative content; and EXAONE Discovery explores scientific data to propose material development methods.

### LG AI Research — EXAONE: Expert AI for Everyone, Bringing Digital Innovation - 2023 Korea Government Expo Report

- Company: LG AI Research (lgresearch.ai)
- Announced: 2023-12-07
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.lgresearch.ai/blog/view?seq=363
- Record: https://forck.live/items/4883-exaone-expert-ai-for-everyone-bringing-digital-innovation-2023-korea
- Subject: EXAONE

LG AI Research demonstrated three EXAONE platforms (EXAONE Universe, EXAONE Discovery, EXAONE Atelier) at the 2023 Korea Government Expo, showcasing their AI technology for public and business use. The event included hands-on demos and seminars on collaborative projects with public institutions and partners.

### Replicate — Clone your voice using open-source models

- Company: Replicate (replicate.com)
- Announced: 2023-12-06
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://replicate.com/blog/how-to-tune-a-realistic-voice-clone
- Record: https://forck.live/items/2543-clone-your-voice-using-open-source-models
- Subject: Platform

To clone a voice using open-source Realistic Voice Cloning (RVC) models on Replicate. It covers creating a training dataset from a YouTube video, training the RVC model, and running inference to generate new audio.

### Replicate — How to create an AI narrator for your life

- Company: Replicate (replicate.com)
- Announced: 2023-12-06
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://replicate.com/blog/how-to-create-an-ai-narrator
- Record: https://forck.live/items/2542-how-to-create-an-ai-narrator-for-your-life
- Subject: Platform

Replicate published a tutorial on creating an AI narrator using vision, language, and text-to-speech models, including Llava 13B, GPT-4-Vision, Mistral 7B, and ElevenLabs.

### Hugging Face — SetFitABSA: Few-Shot Aspect Based Sentiment Analysis using SetFit

- Company: Hugging Face (huggingface.co)
- Announced: 2023-12-06
- Category: research-paper
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/setfit-absa
- Record: https://forck.live/items/1968-setfitabsa-few-shot-aspect-based-sentiment-analysis-using-setfit
- Subject: Platform

Intel Labs and Hugging Face introduce SetFitABSA, a framework for few-shot training of domain-specific Aspect-Based Sentiment Analysis models that is competitive with and even outperforms generative models like Llama2 and T5 in few-shot scenarios.

### Upstage — Understanding document structure with OCR - Document AI technology for LLM

- Company: Upstage (upstage.ai)
- Announced: 2023-12-05
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.upstage.ai/blog/en/document-ai-ocr-for-llm
- Record: https://forck.live/items/18489-understanding-document-structure-with-ocr-document-ai-technology-for-llm
- Subject: Solar Pro / Solar Mini

Upstage plans to launch a document structure analysis API service that accepts any document and returns a markdown-converted result. The post explains the digital assetization process for LLMs, including layout analysis, markdownification, vectorization, and query embedding, and contrasts document structure analysis software with standard OCR or PDF parsing software.

### Replicate — Businesses are building on open-source AI

- Company: Replicate (replicate.com)
- Announced: 2023-12-05
- Category: funding-company-news
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://replicate.com/blog/series-b
- Record: https://forck.live/items/2544-businesses-are-building-on-open-source-ai
- Subject: Platform

Replicate announced a $40 million Series B funding round led by a16z, with participation from NVentures (NVIDIA), Heavybit, Sequoia, and Y Combinator.

### Hugging Face — Goodbye cold boot - how we made LoRA Inference 300% faster

- Company: Hugging Face (huggingface.co)
- Announced: 2023-12-05
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/lora-adapters-dynamic-loading
- Record: https://forck.live/items/1971-goodbye-cold-boot-how-we-made-lora-inference-300-faster
- Subject: Platform

Hugging Face reduced LoRA inference warm-up time from 25s to 3s by dynamically swapping adapters while keeping the base model warm, cutting response time from 35s to 13s and using fewer GPUs.

### Hugging Face — Optimum-NVIDIA Unlocking blazingly fast LLM inference in just 1 line of code

- Company: Hugging Face (huggingface.co)
- Announced: 2023-12-05
- Category: developer-tool-release
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/optimum-nvidia
- Record: https://forck.live/items/1970-optimum-nvidia-unlocking-blazingly-fast-llm-inference-in-just-1-line-of-code
- Subject: Platform

Hugging Face announces Optimum-NVIDIA, an inference library that accelerates LLM inference on NVIDIA GPUs with a simple API change. It claims up to 28x faster inference and 1,200 tokens/second, supports FP8, and provides code examples for LLaMA models.

### Hugging Face — AMD + 🤗: Large Language Models Out-of-the-Box Acceleration with AMD GPU

- Company: Hugging Face (huggingface.co)
- Announced: 2023-12-05
- Category: capability-change
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://huggingface.co/blog/huggingface-and-optimum-amd
- Record: https://forck.live/items/1969-amd-large-language-models-out-of-the-box-acceleration-with-amd-gpu
- Subject: Platform

Hugging Face and AMD announce out-of-the-box support for running Transformers models on AMD Instinct GPUs without code changes, including Flash Attention v2, Paged Attention, DeepSpeed, GPTQ, Optimum-Benchmark, and ONNX Runtime integration. Performance benchmarks show the MI250 GPU delivering over 2.33x more decode throughput and half the prefill latency compared to an A100 card.

### Runway — Runway partners with Getty Images to build enterprise ready AI tools

- Company: Runway (runway.com)
- Announced: 2023-12-04T10:10:10.110000+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://runway.com/news/runway-partners-with-getty-images
- Record: https://forck.live/items/7931-runway-partners-with-getty-images-to-build-enterprise-ready-ai-tools
- Subject: Gen / Aleph

Runway and Getty Images announced a partnership to create the Runway-Getty Images Model (RGM), a baseline model for enterprise video generation that can be fine-tuned with proprietary datasets. The model will be available for commercial use in the coming months.