ODSC’s AI Weekly Recap: Week of March 8th, 2023

Open Data Science Blog Recap

Bench IQ, a Toronto-based startup, has unveiled an AI platform that promises to change how lawyers prepare for court. (Source)

In a blog post released today, OpenAI fired back at Elon Musk’s lawsuit and moved to dismiss his claims about the company’s motives. (Source)

According to a report, Apple is hoping to push forward its efforts in generative AI in a bid to catch up with competitor Microsoft. (Source)

In a lawsuit filed last Thursday night in San Francisco Superior Court, Elon Musk has accused OpenAI, along with its CEO Sam Altman and Microsoft, of reneging on the foundational ethos of the artificial intelligence research company. (Source)

AI News Highlights

During an economic forum at Stanford University, Nvidia’s CEO stated that he thinks artificial intelligence (AI) that thinks like humans could arrive within five years

A new online master’s in AI from Purdue is designed for people who build AI systems and for people who use them

Artificial intelligence is key to the evolution of Florida’s $160 billion agriculture industry

Why are more and more people turning to AI when they’re in need of companionship?

Do AI video generators dream of San Pedro? Madonna among early adopters of AI’s next wave

AMD’s custom Instinct MI309 GPU for China fails export license test from U.S. government: Report

Air Force asks industry for artificial intelligence and machine learning in command and control

Microsoft engineer warns company’s AI tool creates violent, sexual images, ignores copyrights

Ex-Google engineer charged with stealing AI trade secrets while working with Chinese companies.

Convai aims to put an end to basic ‘one-line’ NPCs with AI but says it ‘needs more high-quality writers and artists, not less’ to get it done

New launch alert: FreewayAI provides a lightweight standard for managing prompt templates, database query templates, user input points, and system diagrams. \

Brazil’s electoral authority warns against employing artificial intelligence in electoral campaigning.

A remake done using AI of the cult classic, Silent Hill, is proving that AI still has a long way to go.

Trending New Open Source Projects

• Large World Model (LWM) is a general-purpose large-context multimodal autoregressive model. It is trained on a large dataset of diverse long videos and books using RingAttention and can perform language, image, and video understanding and generation.

• You Only Look Once (YOLO) is one of the most well-known computer vision model architectures. This repository is an implementation of the paper – YOLOv9.

• This repository is the official PyTorch implementation of Google’s Gemma models. Gemma is a family of lightweight, state-of-the-art open models built from research and technology used to create Google Gemini models. They are text-to-text, decoder-only large language models, available in English, with open weights, pre-trained variants, and instruction-tuned variants.

• gemma.cpp is a lightweight, standalone C++ inference engine for the Gemma foundation models from Google. gemma.cpp provides a minimalist implementation of Gemma 2B and 7B models, focusing on simplicity and directness rather than full generality.

• MetaVoice is a text-to-speech foundational model for human-like expression. MetaVoice-1B is a 1.2B parameter base model trained on 100K hours of speech for TTS (text-to-speech). It has been built with the following priorities including – Emotional speech rhythm and tone in English, Zero-shot cloning for American & British voices, and support for (cross-lingual) voice cloning with finetuning. Playground available at https://ttsdemo.themetavoice.xyz

• Official PyTorch codebase for the video joint-embedding predictive architecture, V-JEPA, a method for self-supervised learning of visual representations from video.

• The Open-Sora Plan project‘s aim is to reproduce OpenAI’s Sora. Sora is a generative AI model that creates realistic videos based on prompt descriptions. This project aims to create a simple and scalable repo, to reproduce Sora.

• OpenCodeInterpreter is a suite of open-source code generation systems aimed at bridging the gap between large language models and sophisticated proprietary systems like the GPT-4 Code Interpreter. It significantly enhances code generation capabilities by integrating execution and iterative refinement functionalities.

• A framework for rapid development and deployment of production-ready RAG systems.R2R was conceived to bridge the gap between experimental RAG models and robust, production-ready systems.

New Trending Research From Around The Globe

• A new paper aims to prove that any linear transformer maintains an implicit linear model and can be interpreted as performing a variant of preconditioned gradient descent while also investigating the use of linear transformers in a challenging scenario where the training data is corrupted with different levels of noise. – Paper

• Generative artificial intelligence: synthetic datasets in dentistry – Paper

• CausalGym adapts and expands the SyntaxGym suite of tasks to benchmark the ability of interpretability methods to causally affect model behavior – Paper

• Scaling Up LLM Reviews for Google Ads Content Moderation- Paper

• Sequoia is a scalable, robust, and hardware-aware algorithm for speculative decoding that aims at better scalability by introducing a dynamic programming algorithm to find the optimal tree structure for the speculated tokens. – Paper

• OmniPred: Language Models as Universal Regressors – Paper

• PALO offers visual reasoning capabilities in 10 major languages, including English, Chinese, Hindi, Spanish, French, Arabic, Bengali, Russian, Urdu, and Japanese, that span a total of ~5B people (65% of the world population) – Paper

• Watermarking Makes Language Models Radioactive- Paper

• AgentScope, a developer-centric multi-agent platform with message exchange as its core communication mechanism. – Paper

• Vision-Flan: Scaling Human-Labeled Tasks in Visual Instruction Tuning – Paper

• This paper proposes a recipe for tailoring LLMs to multiple tasks present in translation workflows – Paper

• StructLM: Towards Building Generalist Models for Structured Knowledge Grounding – Paper

Start-Up Funding News

China-based MiniMax, developer of AI technology for social connections and interaction, converting text into visual, audio, and text-to-text modes. raised $600M from private investors.

Massachusetts-based Overjet, an AI-powered dental technology company that provides algorithms to analyze dental images and deliver diagnoses raised over $130M from private investors.

Toronto-based Taalas is a creator of a platform for quickly turning any AI model into custom silicon, raising $50M from private investors.

San Francisco-based Baseten an AI infrastructure company that integrates machine learning into business operations, production, and processes, raised $60M in the Series B round

New York-based Healthee, an AI-driven benefits administration technology that provides customized responses to employee health benefits questions, raised over $55M from private investors.

ODSC News

ODSC Schedule Now Live! See the Full Schedule