Matthias Bastian

Matthias is the co-founder and publisher of THE DECODER, exploring how AI is fundamentally changing the relationship between humans and computers.

Matthias Bastian

Jul 20, 2024

Update

OpenAI's GPT-4o mini is built from the ground up to resist the most common LLM attack

Editorial illustration of malicious and dangerous red data files flying towards a striped data blocking IT wall.

Matthias Bastian

Jul 19, 2024

AI research

AI models might need to scale down to scale up again

Ex-OpenAI researcher explains the backwards race in AI models

Matthias Bastian

Jul 18, 2024

Update

OpenAI's new GPT-4o mini offers higher performance at 60 percent lower price

A small robot illustration

Matthias Bastian

Jul 18, 2024

AI research

LLMs hit a wall when processing complex information from lengthy texts

Researchers have discovered that special AI models can solve complicated tasks just as well if they only use points such as "......" instead of whole sentences. This could make it more difficult to control what is going on in these models.

Matthias Bastian

Jul 18, 2024

Meta's future open-source AI models won't launch in the EU for now

Following the release of the Llama 3 AI language model, Meta plans to invest heavily in even more advanced AI models and services in the coming years in order to develop the world's largest and most powerful AI services. As with Metaverse, it remains to be seen when the company intends to make money from this.

Matthias Bastian

Jul 17, 2024

China orders AI companies to align their models with socialist values

Chinese authorities are testing the responses of AI models to ensure they conform to the party line. Companies need to adapt their systems to pass censorship.

Matthias Bastian

Jul 17, 2024

AI research

OpenAI pits AI models against each other to produce clearer, more verifiable text

OpenAI has created a new method to enhance the clarity and verifiability of AI-generated text. This approach could boost confidence in AI systems and expand their use in critical fields.

Matthias Bastian

Jul 17, 2024

DeepL launches new AI model for translation and text editing that supposedly beats GPT-4

German AI company DeepL has unveiled a new specialized large language model (LLM) for translating and editing text. The company claims this update sets a new industry standard for translation quality.

Matthias Bastian

Jul 17, 2024

Former Tesla AI chief aims to transform education with new "AI-native" school

Matthias Bastian

Jul 16, 2024

Tech giants allegedly used thousands of YouTube videos for AI training without creators' consent

According to an investigation by the news platform Proof News, several leading technology and AI companies, including Apple, Nvidia, Anthropic and Salesforce, have apparently used thousands of YouTube videos to train their AI models without the knowledge of the creators.