Jonathan Kemper

Jonathan writes for THE DECODER about how AI tools can improve both work and creative projects.

Jonathan Kemper

Aug 13, 2024

AI research

This AI understands video games, and it could make cheating easier than ever

AI illustration of a rabbit sitting in front of a TV playing video games.

Jonathan Kemper

Aug 9, 2024

ByteDance's Jimeng AI lets users create short AI-generated videos tailor-made for TikTok

ByteDance, the parent company of TikTok, has launched the text-to-video AI Jimeng AI in China. Like CapCut, the app was developed by ByteDance subsidiary Faceu Technology and is available exclusively for Chinese users. Jimeng AI creates videos that are perfect for sharing on TikTok, even if there is no direct connection between the apps. However, the model's capabilities appear inconsistent, with sometimes faulty or inconsistent results. Compared to OpenAI's Sora, Jimeng AI performs worse. ByteDance wants to hit the ground running in AI and significantly expand its workforce in this area. With TikTok and CapCut, the company has ideal platforms for generative AI. However, legal problems in the USA could limit ByteDance's global reach.

Jonathan Kemper

Aug 6, 2024

AI research

Stable Fast 3D brings better lighting, textures, and materials to AI-generated 3D models

Jonathan Kemper

Aug 2, 2024

AI research

Nvidia uses Apple Vision Pro to record humans controlling robots for more realistic training data

The scaling of training data has already led to great success with language models. The same tactics should also lead to great progress in robotics.

Jonathan Kemper

Aug 2, 2024

ChatGPT's new voice mode can bark like a dog and laugh like a human

OpenAI has started testing the "Advanced Voice Mode" for a small group of ChatGPT Plus users. This mode enables more natural real-time conversations where the user can interrupt at any time and the system responds to emotions. To improve the security and quality of voice conversations, OpenAI tested the capabilities of GPT-4o, the model behind the "Advanced Voice Mode", with over 100 external Red Teams in 45 languages. In addition, safeguards were implemented to block output that deviates from the four predefined voices or contains violent or copyrighted content.

Jonathan Kemper

Jul 31, 2024

AI research

Google DeepMind's new Gemma 2 model outperforms larger LLMs with fewer parameters

Jonathan Kemper

Jul 31, 2024

Early users report real-world performance of Apple Intelligence in iOS 18.1

Because of deepfake "disgust factor": Apple Intelligence does not create photorealistic images

Jonathan Kemper

Jul 31, 2024

Midjourney releases V6.1 model, which is faster and "generally more beautiful"

Photorealistic AI illustration, close-up of a model's face with orange make-up and blue eyes.

Jonathan Kemper

Jul 30, 2024

AI research

AI data isn't destroying AI models after all, researchers say

A recent study published in Nature appears to demonstrate model collapse in several AI architectures, but other researchers say the study makes unrealistic assumptions that are inconsistent with common practice, such as discarding all previous data after each iteration. The critical researchers disagree with the theory that training AI models on synthetic data inevitably leads to "model collapse," where performance gradually declines until the model becomes inefficient. This effect is not a real threat under realistic conditions, they say. With LLaMA 3.1, Meta shows that synthetic data generation processes that iteratively correct errors can improve performance without causing model collapse.

Jonathan Kemper

Jul 28, 2024

AI research

Study reveals major weaknesses in AI's ability to understand diagrams and abstract visuals

Colorful AI illustration of a clock that is impossible to read.