Ad
Skip to content

Matthias Bastian

Matthias is the co-founder and publisher of THE DECODER, exploring how AI is fundamentally changing the relationship between humans and computers.
Read full article about: CentML aims to tackle high cost of AI models

AI startup CentML has raised $27 million in an extended seed round to develop tools that reduce the cost of deploying machine learning models and improve performance. With participation from Gradient Ventures, TR Ventures, Nvidia, and Microsoft Azure AI VP Misha Bilenko, the funding will be used to bolster product development, research, and staff expansion. CentML's software identifies bottlenecks in model training and predicts the total time and cost of deployment, with a compiler that optimizes workloads for best performance on target hardware.

CentML was founded in 2022 by CEO Gennady Pekhimenko, an expert in ML systems, and a team from Amazon, Google, NVIDIA, and IBM. The startup provides a software platform for optimizing AI models for performance and cost-effectiveness, which is much needed in generative AI.

Read full article about: AI Safety gets a push from Chris Meserole and a $10 million fund

Chris Meserole is the first executive director of the Frontier Model Forum, an industry body focused on global AI safety and responsible development. Meserole brings expertise in technology policy and governance, having previously served as director of the Artificial Intelligence and Emerging Technology Initiative at the Brookings Institution.

In addition, Forum members Anthropic, Google, Microsoft, and OpenAI, along with philanthropic partners, are committing more than $10 million to a new AI Safety Fund. The fund will advance AI safety research and support independent researchers around the world. The Frontier Model Forum also releases its first technical working group update on red teaming to expand the conversation on responsible AI governance approaches.

Read full article about: EU's race against time to finalize AI rules before year-end

European lawmakers are struggling to agree on new rules for artificial intelligence, with an agreement unlikely before December. The draft AI rules, which include regulations for foundation models and high-risk AI systems, must be approved by the European Parliament and EU member states. Spain, which holds the EU presidency, has proposed compromises to accelerate the process, such as a tiered approach to regulating foundation models with more than 45 million users. But opponents warn that smaller platforms could be just as risky. However, a final agreement is not expected in the upcoming fourth trilogue. A fifth trilogue is scheduled for early December, and failure to reach an agreement could push negotiations into next year, according to Reuters.

Read full article about: Qualcomm focuses on generative AI capabilities with Snapdragon 8 Gen 3

Qualcomm is set to release its latest flagship mobile platform, the Snapdragon 8 Gen 3, designed specifically for generative AI, according to a leaked internal document obtained by mspoweruser.com. The new chip can run AI models with more than 10 billion parameters, according to the document. Meta's Llama 2 is specifically mentioned, running at 20+ tokens/sec. Qualcomm is using Stable Diffusion, which it has previously demonstrated, for generative AI backgrounds. The Snapdragon 8 Gen 3 might power Samsung's Galaxy S24, due out in February 2024, according to mspoweruser.

Image: Qualcomm via mspoweruser.com