Read full article about: Nvidia shows interest in smaller AI models that run on local hardware
In February, Nvidia acquired OmniML, a two-year-old AI startup known for developing software that compresses machine learning models to run seamlessly on devices instead of in the cloud.
The acquisition could signal Nvidia's intent to improve its AI chips for cars, industrial robots, and drones, and potentially shrink the size of AI software to allow chatbots to run on devices instead of data centers. OmniML has demonstrated significant improvements in model performance and cost savings across multiple industries, running machine learning tasks 10 times faster on a variety of hardware devices. The acquisition is confirmed by the startup's LinkedIn profile and The Information.
Comment
Source: LinkedIn | The Information
Jun 30, 2023
Read full article about: More than 150 European companies say EU AI Act is a bad idea
Over 150 executives from European companies, including Airbus and Siemens, have criticized the proposed EU artificial intelligence regulation in an open letter, arguing that the rules could harm competitiveness and technological sovereignty without adequately addressing challenges.
The concerns focus on the regulations' heavy emphasis on foundation models, which underpin chatbots, and the potential for disproportionate compliance costs and liability risks. The companies have called for a regulatory body of industry experts to monitor the implementation of the law instead of focusing on rigid compliance.
In our assessment, the draft legislation would jeopardise Europe’s competitiveness and technological sovereignty without effectively tackling the challenges we are and will be facing. [...] Europe cannot afford to stay on the sidelines.
Excerpt from the open letter sent to the EU Commission, the parliament and member states
Comment
Source: Financial Times
Read full article about: Terence Tao thinks we are largely unprepared for the advent of LLMs
Generative AI tools like GPT-4 will disrupt traditional expectations and workflows in several fields, including research mathematics, according to famous mathematician Terence Tao.
He suggests that the robustness and adaptability of these models could lead to AI tools that integrate with traditional software or serve as compassionate conversationalists, translators, teachers, and more. Tao predicts that by 2026, AI could be a trusted co-author in mathematical research and various other fields. However, the introduction of AI assistance will also present challenges for adapting existing institutions and practices. You can read his full essay here.