Mistral has already attracted attention with freely available AI language models. Mistral Large is now the most powerful offering with which the French company is targeting OpenAI’s GPT-4. A special feature is its multilingualism. Mistral Large should not only work well in English, but also in French, Spanish, German and Italian. Unfortunately, Mistral Large is … Read more
Text has always been a big problem for image generators. In the beginning it was illegible gibberish, later it looked better, but was still full of errors and therefore mostly unusable. Ideogram promises an improvement with version 1.0. Our tests show: It can work well, but unfortunately not always. You can test Ideogram for free.
Anthropic has released Claude 3, the latest version of its AI language model. It comes in three sizes: Haiku, Sonnet, and Opus. The smallest offering, Haiku, will be released later. Sonnet and Opus are already available, but not yet in German-speaking countries. This is a pity, as both Claude 3 variants delivered very good results … Read more
Stability AI is a provider of numerous AI tools that focuses more than others on open availability. There is now a large family of image generators that can be used on your own computer under the name Stable Diffusion. The recently announced Stable Diffusion 3 promises better quality in the future. This is especially true … Read more
Local AI is an interesting concept: AI assistants similar to ChatGPT that do not run in the cloud, but on your own PC or on a server you operate yourself. One challenge is the speed of the answers. Nvidia has now presented “Chat with RTX”, which uses the computing power of its own graphics cards. … Read more
OpenAI has caused quite a stir with a preview of the video AI “Sora”. The examples on the official website are indeed impressive. However, without our own tests, it is not yet possible to make a meaningful assessment of how well Sora works in everyday life and what the video clips can be used for. … Read more
Google has visibly accelerated AI development. At the same time, the company is unifying its activities under the “Gemini” brand. The ChatGPT competitor Google Bard is now called Google Gemini and the proprietary AI models are also called Gemini. Until now, Google’s chatbot used the “Gemini Pro” AI model, but now “Gemini Ultra” is also … Read more
Temperature in Large Language Models (LLMs) and other generative AI systems is a crucial control parameter that influences the randomness and creativity of the outputs. The value typically ranges from 0 to 1, where 0 represents highly deterministic (predictable) and 1 represents very random responses. With a low temperature setting, the system consistently repeats the … Read more
The Browser Company has introduced Arc Search, an app for iPhones and iPads that searches the web in response to a user’s search query and presents the information it finds on a page. If someone clicks on one of the listed sources, ads are automatically blocked, among other things. In his article on Engadget, Pranav … Read more
Nebojša Vujinović Vujo buys abandoned news websites and fills them with AI-generated content to generate advertising revenue. Vujo’s most profitable acquisition was The Frisky, which generated over $500,000 in revenue in its first year. However, traffic to its sites has declined over time, possibly because search engines are getting better at identifying spam content. More … Read more