Mistral NeMo 12B, GPT-4o Mini, Llama 3

July 26-28, 20246th International Conference on Artificial Intelligence and Computer Science (AICS 2024) Wuhan, China. Register Now

July 30-31, 2024 Fortune Brainstorm AI Singapore Register Now

GPT-4o Mini: The Budget-Friendly AI Model

Cost Efficiency: Priced at 15 cents per million input tokens and 60 cents per million output tokens, GPT-4o mini is much cheaper than GPT-4, making advanced AI more accessible to businesses and developers.

Performance: Despite its small size, GPT-4o mini outperforms GPT-3.5 Turbo and other models in benchmarks, excelling in textual intelligence, multimodal reasoning, math, and coding.

Safety: Built with robust safety measures, GPT-4o mini uses reinforcement learning with human feedback and new techniques to enhance reliability and resistance to attacks.

Accessibility: Available through various APIs, GPT-4o mini is easy to integrate into projects for solo developers and large enterprises alike.

Future Vision: OpenAI aims to make AI affordable and integrated into daily digital life, with GPT-4o mini leading the way.

Read the full version Here!

Mistral AI and NVIDIA Unveil Mistral NeMo 12B

Mistral AI and NVIDIA have launched the Mistral NeMo 12B, a powerful language model designed for chatbots, multilingual tasks, coding, and summarization.

Key Features

• Training: Leveraged NVIDIA DGX Cloud AI platform with 3,072 H100 80GB GPUs, using NVIDIA’s AI architecture and TensorRT-LLM for enhanced performance.

• Performance: Excels in conversations, math, reasoning, and coding with 128K context length and 12 billion parameters, ensuring high accuracy and efficiency.

• Deployment: Packaged as an NVIDIA NIM inference microservice for quick and secure enterprise-grade deployment.

• Licensing: Open-source under Apache 2.0, promoting innovation and easy integration into commercial applications.

• Availability: Usable across various platforms including cloud, data centers, and RTX workstations, available now via ai.nvidia.com.

Mistral NeMo 12B combines Mistral AI’s expertise and NVIDIA’s technology, offering a robust and efficient AI solution for enterprises.

Read the full version Here!

Chinese Firms Access Banned Nvidia AI Chips via Cloud Services

Despite the U.S. ban on exporting Nvidia AI chips to China, Chinese companies access these chips through Google and Microsoft cloud servers outside mainland China. Both tech giants claim compliance with U.S. export rules. These cloud services use Nvidia’s A100 and H100 chips, banned for shipment to China since October 2022.

OpenAI has restricted ChatGPT access in China, leading Chinese firms to develop local AI chatbots from Alibaba, Tencent, and Baidu. Reports indicate significant purchases of Nvidia's H20 chips tailored for the Chinese market.

The U.S. Commerce Department is considering further restrictions, including "know your customer" regulations, to prevent Chinese firms from using advanced AI chips to enhance military capabilities or suppress minorities. However, the loophole allowing access via cloud services remains a contentious issue.

New Tools for ChatGPT Enterprise to Enhance Compliance, Security, and User Management

OpenAI has expanded its ChatGPT Enterprise offerings, enabling global organizations like Boston Consulting Group, PwC, Los Alamos National Laboratory, and Moderna to securely incorporate AI into their workflows. The new tools include the Enterprise Compliance API, which aids in auditing and managing workspace data, and integrates with leading eDiscovery and Data Loss Prevention (DLP) companies. These integrations support compliance with regulations such as FINRA, HIPAA, and GDPR, and assist with archiving, audit trails, data redaction, and policy enforcement, ensuring robust security and compliance for enterprises.

In addition, OpenAI introduces automated user management through the SCIM (System for Cross-domain Identity Management) integration, allowing admins to sync internal employee directories with ChatGPT Enterprise. This feature ensures accurate and up-to-date user provisioning and deprovisioning, supporting directories like Okta Workforce, Microsoft Entra ID, Google Workspace, and Ping. The SCIM integration, currently in beta, will be broadly available next week, further simplifying user management for enterprise customers.

ChatGPT Enterprise also sees enhancements in GPT control settings, providing admins with granular management over custom GPT versions and third-party GPTs. These settings enable precise control over GPT actions, user groups, and global GPT capabilities, ensuring safe and effective use of AI within enterprises. Alongside these new tools, ChatGPT Enterprise maintains stringent data privacy and security standards, with no customer data used for training, data encryption, and compliance with CCPA, CSA STAR, and SOC 2 Type 2. These updates highlight OpenAI’s commitment to supporting secure and scalable AI deployments for enterprises and educational institutions alike.

Meta not launching multimodal AI model in EU due to regulatory pressure
Meta has decided not to launch its new multimodal AI model in the European Union due to regulatory concerns. This decision is influenced by the complex and stringent data protection regulations in the EU, particularly the GDPR. Privacy advocacy groups, like NOYB, have raised significant objections to Meta's data handling practices, which they argue do not comply with GDPR requirements for explicit user consent. These regulatory pressures have led Meta to delay the deployment of its AI technologies in the EU, highlighting the ongoing tension between technological innovation and regulatory compliance.

The biggest names in AI have teamed up to promote AI security
Microsoft, Google, OpenAI, and Anthropic have created the Frontier Model Forum to promote safe and responsible AI development. The forum aims to advance AI safety research, establish best practices, and collaborate with policymakers and academics. Additionally, Microsoft and OpenAI are tackling AI misuse by state-affiliated threat actors and cybercriminals, implementing measures to detect and disrupt malicious activities, and enhancing security across the AI ecosystem


The Media Industry’s Race To License Content For AI
Intel's venture capital arm has raised concerns in Washington due to its investments in Chinese tech companies, highlighting tensions over U.S. firms' involvement in China's tech sector. These investments, part of Intel Capital's strategy, are seen as potentially conflicting with U.S. national security interests, especially amid increasing scrutiny of tech ties with China. The situation underscores the delicate balance between fostering business growth and adhering to geopolitical and security considerations.


Cohere and Fujitsu Announce Strategic Partnership To Provide Japanese Enterprise AI Services
Cohere and Fujitsu have announced a strategic partnership to develop and provide AI services for Japanese enterprises. This collaboration will focus on creating AI models tailored to the Japanese market, with secure and private deployment options for global enterprises. The partnership aims to enhance the capabilities of Japanese businesses through advanced AI solutions.


Groq’s open-source Llama AI model tops leaderboard, outperforming GPT-4o and Claude in function calling
Groq has introduced open-source LLaMA 3 models that outperform GPT-4o and Claude in function calling, topping the Berkeley Function Calling leaderboard. Groq's Learning Processing Unit (LPU) generates up to 500 tokens per second for a 7B model, compared to ChatGPT's 30-60 tokens per second on Nvidia GPUs, thanks to a software-first design ensuring fast and accurate AI inference

👥 Connect & Feedback!

👉 Join Us:

📧 Advertise In Weekly AI News:

📧 Contact directly at [email protected]

😍 Share with your friends!