Author HubMaximilian Schreiner

Max is managing editor at THE DECODER. As a trained philosopher, he deals with consciousness, AI, and the question of whether machines can really think or just pretend to.

AI research

Nov 19, 2024Nov 19, 2024

Study reveals AI models have hidden capabilities they can't access through normal prompts

AI in practice

Oct 11, 2024Oct 11, 2024

Tesla unveils Cybercab robot taxi, but robot Optimus is the bigger deal

AI and society

Sep 27, 2024Sep 27, 2024

Update

OpenAI CEO Sam Altman denies rumors of giant equity stake as company considers restructuring

AI research

Dec 13, 2023Dec 13, 2023

Nvidia's Adaptive Shells offer a glimpse into the near future of rendering

AI in practice

Dec 12, 2023Dec 12, 2023

Publish or perish: 'extremely productive' researchers quadruple

AI research

Dec 12, 2023Dec 12, 2023

NeurIPS 2023: These are the top papers and award winners at the largest AI conference

AI in practice

Dec 12, 2023Dec 12, 2023

Short

Nvidia, the world's most valuable chipmaker, has become one of the leading investors in AI start-ups this year, having invested in over two dozen companies. According to Dealroom, Nvidia participated in 35 deals in 2023, nearly six times more than last year, making it the most active large-scale investor in AI. Nvidia's investments total $872 million, over ten times its investment in the same period last year, the Financial Times reports. The companies Nvidia invests in all use its technology, with its portfolio now including Inflection AI, Cohere, Hugging Face, and CoreWeave. Nvidia's latest investment was in Paris-based AI start-up Mistral.

Financial Times

AI research

Dec 10, 2023Dec 10, 2023

StripedHyena: A new architecture for next-generation generative AI?

AI research

Dec 9, 2023Dec 9, 2023

Short

French startup Mistral AI has released its new language model Mixtral 8x7B via a torrent link. Mixtral is a mixture-of-experts model, following an architecture that OpenAI is rumored to be using for GPT-4, but on a much larger scale.

There are no benchmarks, blog posts, or articles about the model yet, but Mistral-7B — the first version of Mistral AI — generally performed very well and was quickly adopted by the open-source community. Mistral is thought to have used the MegaBlocks MoE library for training. The Paris-based company was recently valued at nearly $2 billion.

AI research

Dec 8, 2023Dec 8, 2023

VideoRF paves the way for real-time 3D streaming on smartphones and VR headsets

AI research

Dec 8, 2023Dec 8, 2023

Short

With Generative Power of Tens, researchers demonstrate a method enabling "extreme semantic zooms" from wide-angle views to macro shots. Unlike traditional super-resolution methods, the team from the University of Washington, Google Research, and UC Berkeley uses text prompts for each scale, enabling deeper zoom levels. Compared to traditional outpainting techniques, the approach produces a consistent zoom in which the content of the coarser and finer zoom levels are consistent.