We've seen a lot of speculation in the last couple of weeks around OpenAI's roadmap, and it seems everyone will finally find out what is in the works on Monday next week! We also see some interesting headlines from Anthropic about the evolution of prompts and learn about some limitations in fine-tuning! Enjoy.

— Sasha

OpenAI Schedules Live Demo for Updates

First impacted: AI researchers, technology enthusiasts
Time to impact: Short

OpenAI has scheduled a live demonstration for Monday, May 13, at 10 AM PT to showcase updates to ChatGPT and GPT-4, as announced by Sam Altman on Twitter. Altman emphasized that the event will not feature GPT-5 or a new search engine, but will focus on recent enhancements to existing models. [OpenAI] Explore more of our coverage of: OpenAI Updates, GPT-4 Enhancements, AI Demonstration. Share this story by email

Anthropic Enhances AI with Automated Prompt Engineering

First impacted: AI developers, product managers
Time to impact: Short

Anthropic says it has enhanced its console to allow users to generate production-ready prompts using Claude, an AI that employs chain-of-thought reasoning to refine prompt engineering. This tool, according to the company, aims to streamline the prompt creation process by making it more precise and reliable, thus reducing the need for specialized skills. [via @AnthropicAI] Explore more of our coverage of: AI Prompt Engineering, Claude AI, Production-Ready Prompts. Share this story by email

Study Reveals LLMs Increase Hallucinations When Learning New Facts During Fine-Tuning

First impacted: AI researchers, AI model developers
Time to impact: Medium

Researchers at the University of Michigan have demonstrated that LLMs may struggle to integrate new factual knowledge during supervised fine-tuning, leading to increased errors and a tendency to generate factually incorrect responses. Their study indicates that LLMs are more effective when fine-tuning leverages their pre-existing knowledge rather than attempting to incorporate new information. [Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?] Explore more of our coverage of: LLM Fine-Tuning, Knowledge Integration, AI Errors. Share this story by email

Salesforce Shares BLIP3 Model's Results

[Salesforce/blip3-phi3-mini-instruct-r-v1 · Hugging Face] Share this story by email

Lumina-T2X AI Model Launched That Transforms Text into Various Modalities

[GitHub - Alpha-VLLM/Lumina-T2X: Lumina-T2X is a unified framework for Text to Any Modality Generation] Share this story by email

Read more of today's top AI news stories like this at [https://aitimetoimpact.com/].