Must Read
Per a recent Wall Street Journal analysis, Microsoft is struggling to make profits from GitHub's AI-powered code completion tool, Copilot.
The article cited the cost associated with running the service as the primary reason leading to a loss.
More:
Sources told WSJ that Microsoft is losing, on average, $20 a month per user, while heavier users are causing a loss of up to $80 each month.
Microsoft charges $10 per month for GitHub's Copilot service.
Microsoft leverages tens of thousands of Nvidia A100s and H100s, priced at around $30,000 per piece, to train and run its LLMs.
Wall Street Journal expects profitability to increase once computing costs come down.
According to the State of AI 2023 Report, AI hardware is strongly consolidated towards NVIDIA, and the company controls nearly 95% of the market for graphics processors used for machine learning.
Andrew Walbran from the Android Rust Team provided a detailed account of the adoption of the new Rust pVM firm that was released in Android 14.
The decision to move native code in Android from C++ to Rust is motivated by the desire for improved safety and productivity and to reduce the risk of security vulnerabilities.
More:
To support the transition, Google has contributed to the Rust community by using and contributing to existing crates and publishing new ones.
Walbran acknowledged that there are concerns regarding binary size when using Rust but noted that careful optimization and build practices can yield results comparable to C.
Despite its architectural support and binary size challenges, Google considers Rust to be a significant improvement over C and C++ in terms of safety and productivity.
Further, the team behind Trusty is working on adding support for Trusted Applications written in Rust.
Low-code platform Retool announced the general availability of Workflows, its visual automation tool.
Workflows empowers developers to compose APIs and database queries using JavaScript and Python to automate work.
More:
To enhance developer experience, Retool has added building blocks and deeper integrations with other Retool products to support more complex, production-grade task orchestration.
Workflows allows calling a workflow from another to build complex processes and examining past runs, logs, data, and historical code overlays to identify and address bugs.
Workflows is built on top of Temporal, an open-source programming model that provides a reliable execution engine.
Retool claims developer teams from Rippling, Stripe, and Taco Bell have used Workflows since its beta announcement last November.
Gradient, an API platform for AI developers, raised a $10M seed round led by Wing VC.
The investment round saw participation from Mango Capital, Tokyo Black, The New Normal Fund, Secure Octane, and Global Founders Capital.
More:
Gradient offers open-source pre-trained neural networks for AI applications, which developers must set up and manage the infrastructure for running these models.
Its features include full model ownership, rapid LLM creation, the ability to build on Llama 2 models and domain expert models for finance, healthcare, and law, and SOC2 and HIPAA compliance.
The startup claims training one of its LLMs on a custom dataset requires just a few simple lines of code.
The Burlingame-based startup intends to use the funds to develop its enterprise-focused development platform.
