By Editorial Team Fine-tune Your GPT 40 in Your Own Way—The Advance Feature
Fine-tuning for GPT-4o is now available which allows developers to enhance their model performance and accuracy with custom datasets. This feature enables tailored responses and domain-specific instructions with strong results from just a few dozen examples.
Through September 23, organizations receive 1M free training tokens daily. The cost of fine-tuning is $25 per million tokens including the inference fees of $3.75 per million input tokens.
The successful use cases of GPT 4o fine-tuning include Cosine’s AI assistant, Genie, achieving state-of-the-art results on the SWE-bench, and Distyl’s top performance on BIRD-SQL. In addition to this, data privacy and safety are prioritized which will ensure control over all business data.
Official Source: https://openai.com/index/gpt-4o-fine-tuning/
AI Advantage: Which Job Roles Gain Most Benefits from Copilot?
Microsoft’s research into AI adoption at work reveals that some job functions are gaining early advantages from AI that are particularly useful for time-saving features.
A survey of 1,300 Copilot users found that professionals that are in cybersecurity, product development, sales, etc reported that it is the most time-saving platform while procurement, legal, and supply chain roles saw the least.
Sales professionals benefit from AI in lead prioritization and email drafting while other fields like the law face challenges due to their confidentiality needs. An understanding of these variations is essential for guiding AI integration across various businesses. Most users see AI as valuable with just 11 minutes of daily time savings.
Official Source: https://www.microsoft.com/en-us/worklab/ai-data-drop-which-jobs-have-an-ai-advantage?ocid=FY25_soc_omc_br_x_
Project Astra: The Future of Multimodal AI Assistants
Project Astra, the Model that has been designed based on Gemini and carries the aim of creating advanced AI assistants. The model can process multimodal information and can understand the context of natural conversation.
Project Astra is the model that has been tested in real-time using a Google Pixel phone and prototype glasses. Astra has the capability of integrating perception, reasoning, and conversational skills to enhance user interaction.
This AI agent is designed to be proactive, teachable, and personal, with improved response times and varied intonations. The future developments of Astra will bring more advanced capabilities to Google products, like the Gemini app and web experience, making expert AI assistants accessible through phones and glasses.
Official Source: https://deepmind.google/technologies/gemini/project-astra/
AI Meets Journalism: OpenAI and Condé Nast Join Forces
OpenAI is partnering with Condé Nast to enhance AI-driven news discovery. This collaboration will integrate content from brands like Vogue, The New Yorker, and GQ into ChatGPT and the SearchGPT prototype.
Whereas, SearchGPT combines conversational models with web information to offer fast and intuitive search results that have direct links to news stories. The feedback from Condé Nast and other news partners will guide the improvements to ChatGPT.
This partnership, along with others like the Associated Press and Financial Times, is forwarding with the intention of ensuring AI supports accurate and high-quality journalism.
Official Source: https://openai.com/index/conde-nast/
Related Posts
