Launches Interrupted? OpenAI and Google Clash Again with New Models | Weekly AI Highlights

5 Major AI Developments You Can't Miss This Week

AI models are developing fast. It is important for professionals to keep learning about new things in AI.

AI Disruption wants to give weekly updates about AI to readers. It will review the AI industry and share key insights.

Let's look back at the major events for AI models from last week.

Key Highlights

This week, many new AI models were released. Big companies like OpenAI, Google, Baidu, and Tencent launched new AI.

OpenAI's GPT-4o and Google's Gemini models stood out the most. These new AI can understand multi-modes, and long texts, and are faster.

The new AI brings new possibilities for using AI in different ways and improving user experience. This shows AI will play an even more important role in the future.

Ongoing AI Model Updates

• May 12: Stanford made a tool called ThunderKittens. It helps write fast code for GPUs. ThunderKittens was very fast on the RTX 4090 GPU. It beat FlashAttention-2 by 30% on the H100 GPU.

• May 14: OpenAI released a new AI model called GPT-4o. This model can understand text, speech, and images. It can generate text, audio, and images in real time. GPT-4o is much better at English, coding, other languages, vision, and audio.

• May 15: Google made new Gemini models - Gemini 1.5 Flash and Gemini 1.5 Pro. Gemini 1.5 Flash is small and very fast. It can process 1 million pieces of data. Gemini 1.5 Pro can handle huge amounts of information like long videos, audio, code, or text.

• May 15: Baidu launched a self-driving AI model called Apollo ADFM. It is the first Level 4 autonomous driving model. Baidu says it is over 10 times safer than human drivers. It can drive in complex city areas.

• May 17: ByteDance released the Doubao AI model family. It used to be called Lark models. There are 9 Doubao models for different uses. ByteDance also released AI apps called Button and Doubao App.