• Front Research
  • Posts
  • Weekly tech roundup: OpenAI Unveils Compact GPT-4o Mini, Expanding AI Chatbot Capabilities

Weekly tech roundup: OpenAI Unveils Compact GPT-4o Mini, Expanding AI Chatbot Capabilities

Biggest tech stories of the week.

In partnership with

TECH BRIEF

OpenAI has launched a mini version of its highly advanced AI model, GPT-4o. Dubbed "GPT-4o mini," the new model aims to bring the power of the larger AI to a more cost-efficient and accessible format. There are plans to integrate multimedia capabilities including image, video, and audio. The original GPT-4o model offers improved performance in audio, video, and text, with support for 50 languages. The smaller model is a part of OpenAI's effort to lead in " multimodality," offering a range of AI-generated media types in one tool, ChatGPT.

đź’ˇ Insight: OpenAI's launch of GPT-4o mini signals a strategic push towards democratizing advanced AI capabilities and enhancing multimodal functionalities in ChatGPT. By providing cost-efficient, powerful tools to a broader audience, OpenAI aims to solidify its leadership while navigating profitability challenges in the rapidly evolving AI landscape.

The Biden administration is considering implementing the foreign direct product rule (FDPR) to tighten controls on foreign-made products using US tech. This move targets companies like Tokyo Electron Ltd. and ASML Holding NV, who provide China with advanced semiconductor technology. The suggested sanctions reportedly led to fallen shares in Tokyo Electron and other companies. While US firms resist giving China access to specific tech services and gear, tech companies argue that these policies have led to loss in revenue and amplify the need for stronger allied cooperation.

OpenAI has launched a new algorithm that enables large language models (LLMs) like GPT-4 to provide clearer justifications for their conclusions. It focuses on helping AI models overcome the problem of being a "black box", and increasing their legibility. The solution, detailed in a paper titled “Prover-Verifier Games Improve Legibility of LLM Outputs”, enhances the trustworthiness of AI applications in critical fields, such as healthcare and defense. The algorithm's Prover-Verifier Game prompts two AI models to contend the correctness of their answers, pushing them to clarify their reasoning.

Alphabet, Google's parent company, is in advanced negotiations to acquire cybersecurity startup Wiz, in a deal potentially worth around $23 billion. This move would significantly boost Alphabet's efforts in expanding its cloud computing business, an area where it has notably been outperformed by competitors. This comes as Wiz's valuation continues to surge after it raised $1 billion with a $12 billion valuation earlier this year. Meanwhile, Google awaits verdicts in several ongoing antitrust lawsuits.

Big tech companies like Apple, Nvidia, Apple, and Salesforce have used transcripts from over 173,000 YouTube videos, including from prominent educational institutions and media brands, to train their AI, despite YouTube’s policies against unauthorized data harvesting. This raises concerns among creators and experts about compensation, consent, and potential misuse of the data. The dataset forms part of the Pile, a compilation released by non-profit EleutherAI and utilized by several tech giants.

Netflix has outperformed Wall Street's 2420’Q2, registering a 17% YoY increase in revenue and a 44% surge in profits after adding 8 million subscriptions within that quarter. Notable growths include a net income of $2.15 billion, earnings per share of $4.88, and revenue of $9.56 billion. However, Netflix projects a fall in paid net additions in Q3 due to the fading impact of paid sharing. In a bid to capture more market share and offer a richer user experience, Netflix is planning to enhance its gaming offerings and revamp its homepage design.

Meta will withhold its future multimodal AI models from the European Union due to regulatory concerns, also impacting the ability of EU businesses to use these models. Meta says the decision will not affect the launch of its text-only Llama 3 model in the region. The move is indicative of increasing tensions between US tech giants and European regulators.

Figma Inc., a design startup, has attracted investment from Coatue Management, Alkeon Capital Management, and General Catalyst Partners, setting its worth at $12.5 billion. The company projects an annual recurring revenue of over $1 billion by 2023.

Meta Platforms is exploring a multi-billion investment in eyewear manufacturer EssilorLuxottica to boost their smart glasses collaboration, pivoting further into the wearable technology market.

Reply

or to participate.