![](https://decodingdatascience.com/wp-content/uploads/2024/07/Copy-of-Copy-of-Copy-of-Copy-of-Copy-of-Copy-of-Copy-of-Copy-of-Copy-of-newsletter-4-1024x512.png)
AI News Roundup GPT-4o Mini AGI Developments and Industry Updates
GPT-4o mini: advancing cost-efficient intelligence
OpenAI has announced the release of GPT-4o mini, their most cost-efficient small model aimed at making intelligence more accessible. This model offers improved performance over previous versions, scoring 82% on MMLU and outperforming GPT-3.5 Turbo. It is priced at 15 cents per million input tokens and 60 cents per million output tokens, making it more affordable for a wider range of applications.
GPT-4o mini supports text and vision in the API, with plans for additional input and output options in the future, such as image, video, and audio. The model has a context window of 128K tokens and shows strong performance in tasks involving both text and vision, as well as in math and coding proficiency
AGI Developments OpenAI on the Path to AGI
OpenAI has mapped out five levels for achieving Artificial General Intelligence (AGI). They are:
Level 1: Conversational AI, like chatbots (achieved)
Level 2: Human-level problem-solving AI (almost achieved)
Level 3: Agents that can take actions on our behalf (like booking flights)
Level 4: Innovative AI that can aid in invention
Level 5: Organizational AI that can run an organization
OpenAI is working on a new reasoning technology called Strawberry. It is aimed at level two on the path to AGI.
Some speculate that Strawberry is a rebranded version of Qar, a project previously mentioned in a leaked document.
The goal of Strawberry is to not only answer questions but also to autonomously navigate the web to conduct research.
AI News Roundup GPT-4o Mini AGI Developments
Meta and Apple Withhold AI Models from EU Users
Meta has joined Apple in deciding to withhold certain AI models from users in the European Union. This move is driven by concerns over compliance with the EU’s stringent data privacy regulations and the Digital Services Act. By not deploying these AI models in the EU, both companies aim to avoid potential legal and financial repercussions associated with non-compliance.
TSMC’s AI Boom
Taiwan Semiconductor Manufacturing Company (TSMC) is forecasting record growth due to the booming demand for AI chips, despite rejecting a joint venture proposal from the United States. This growth is fueled by the increasing need for advanced semiconductors in AI applications, showcasing TSMC’s pivotal role in the global tech supply chain.
Mistral AI and NVIDIA’s 12B NeMo Model
Mistral AI and NVIDIA have launched the Mistral NeMo, a state-of-the-art 12-billion-parameter AI model designed for enterprise use. The model excels in various tasks including multilingual processing, coding, and summarization. With a context window of 128,000 tokens, Mistral NeMo offers improved performance and efficiency, making it ideal for complex applications. It uses the new Tekken tokenizer for better compression and is available under the Apache 2.0 license, promoting broader adoption in the AI community.
New TelecomGPT Model by Abu Dhabi Researchers
Researchers at Khalifa University’s 6G Centre and the Technology Innovation Institute (TII) in Abu Dhabi have developed TelecomGPT, the first large language model (LLM) specifically designed for the telecom industry. This model is built and trained on a specialized dataset tailored for telecom applications, outperforming several of the world’s most powerful LLMs such as GPT-4, Llama-3, and Mistral in telecom-related benchmarks.
SoftBank Acquires Graphcore
SoftBank has acquired British AI chipmaker Graphcore, doubling down on its investment in AI hardware. This acquisition aims to strengthen SoftBank’s position in the AI hardware market, enabling it to better support the growing computational demands of AI applications.
Samsung’s AI-Powered Wearables and Foldables
Samsung has unveiled a new line of AI-powered wearables and foldable devices. These innovations integrate advanced AI capabilities to offer enhanced user experiences, such as personalized health monitoring and improved multitasking features. The new products mark a significant step forward in the evolution of smart technology.
Join us in celebrating an exciting milestone! Decoding Data Science is now a proud partner of the Dubai AI & Web3 Festival 2024. By joining forces with this prestigious event, we are dedicated to fostering diversity, driving innovation, and advancing our ecosystem.
This collaboration with the Dubai AI & Web3 Festival will significantly enhance our mission to build and empower the AI community.
Discover more about exclusive discounts for our community by registering here: https://lu.ma/6f1mrtxf.
Join our two amazing free workshops happening in the community this week