Artificial intelligence technologies, which have started to be used more widely in daily life recently, continue to develop rapidly. Almost every day there is a remarkable development in this field, while important thresholds are vaccinated one by one. In our “AI Diaries” series, we continue to regularly record the development of this technological revolution, which has the potential to change the life in the world in the coming period.

What happened this week in the world of artificial intelligence?

“AI Diaries” series, this week, humanoid robots are about to become reality as the world fully realizes the week we can fall. Because the developments in the world of artificial intelligence this week, humanoid service robots will enter our lives in the very near future. Nvidia, Boston Dynamics, Mercedes, such as companies such as overlapping shares, now the sector has started to turn the focus here.

On the other hand, as in every week, new models and new features have been introduced in both large language models and visual creation tools this week. This is the prominent developments in the world of artificial intelligence this week:

Nvidia: Humanoid Robot Revolution can take place earlier than expected

This week, Nvidia introduced software tools that will help humanoid robots more easily navigated, while the company’s CEO Jensen Huang made remarkable statements about the developments in this field. Huang says that humanoid robots will enter our lives before expected and thinks that these robots will first be adopted in the production sector. Huang says that the current cost of a humanoid robot rent is probably around $ 100,000, and he believes that this will be a financial logic investment.

As a matter of fact, Nvidia announced this week’s new basic model Isaac Gr00T N1 and gave the message, “The age of the general purpose robots has begun”. The company said that this model will accelerate the development of humanoid robots with its pre -trained but customizable structure.

The new Atlas Robot showed that Nvidia might be right

The new Atlas robot introduced by Boston Dynamics this week, showed that Jesen Huang may not be very unfair in the prediction of Jesen Huang with its advanced mobility. Trained with artificial intelligence, reinforced learning and movement capture techniques, the new Atlas can now imitate human movements with incredible natural, beyond walking and jumping.

According to Boston Dynamics, this new generation Atlas is not just physically more agile; It can also react better to unpredictable situations with artificial intelligence -based learning skills. These features aim to make it an ideal robot that can help people in industrial production.

Humanoid Robots will be tested in homes this year

When it comes to humanoid robots, new initiatives emerge, even though we make a name for companies such as Figure, Boston Dynamics and Tesla. One of them, the Norwegian robotic initiative 1X announced that they will start to test their robots in homes before the end of this year. Early tests of the service robot, named Neo Gamma, will begin this year. In the first stage, the robot will be tested several hundred or several thousand houses. Neo Gamma is one of the robots using Nvidia’s GR00T N1.

Intuicell introduced the robot dog that can learn self -learning like a human being

Intuicell, a Swedish attempt, introduced a robot dog called Luna, equipped with a digital nervous system that mimics human education. In this new approach, in which Intuicel claims to revolutionize the robotic field, the robot dog will learn by a method similar to neurons, rather than using a model based on traditional large data sets and productive artificial intelligence. This will allow Luna to decide, adapt and set independent goals, just like a real dog. At least Intuicell’s claim is in this direction. We will wait and see if the company can fill this claim. But at least in the concept phase, Luna seems quite ambitious.

Nvidia introduced new artificial intelligence -oriented products and services in GTC 2025

This week, Nvidia organized a new artificial intelligence -oriented event and introduced its new products and services to be available in this field.

One of the new products introduced at the GTC 2025 event is the company’s new “Superchip” Blackwell Ultra GB300, Vera Rubin and Rubin Ultra. Blackwell Ultra, which will be released this year, will be an improved version of the Blackwell architecture announced last year by Nvidia. The new Blackwell Ultra chip offers 20 Petaflop artificial intelligence performance, while the memory capacity is upgraded from 192GB HBM3E to 288GB HBM3E. Vera Rubin and Rubin Ultra, which will offer even more superior performance, will be released in 2026 and 2027, respectively. You can find our news about these super chips here.

Personal Artificial Intelligence Computers from Nvidia: DGX Spark and DGX Station

Nvidia has announced its new super chips as well as its new AI -supported desktop computers. The world’s smallest artificial intelligence super computer, which was previously announced under the name “Digits”, will be released under the name DGX Spark. It will also be accompanied by a larger and stronger DGX station model.

Mac mini -size DGX Spark gets power from Nvidia’s GB10 Blackwell Superchip processor. Providing NVME SSD storage up to 128GB of combined memory and 4TB, this device provides 1,000 trillion processing/seconds (Tops) AI performance thanks to its fifth generation tensor nuclei with FP4 -supported. So, it has enough power to educate and operate complex artificial intelligence models locally. You can find our news about DGX models here.

Gok’s visual capabilities are developing

Developed by Elon Musk’s XAI company, Gok develops its capabilities on its visual side. Elon Musk, this week, shared by sharing Gok’un now made specific changes on the visuals announced. For example, when you give Gok your own picture and add a sunglasses or hat to add it.

On the other hand, the XAI bought Hotshot working on artificial intelligence -supported video creation vehicles this week. After this purchase, Gok is thought to be added to Grok. Thus, Gok will have similar features to Penai’s Sora.

Claude now has web access

Developed by Anthropic, Claude was one of the most advanced language models on the market today, but lag behind his competitors. Because Anthropic is a little late, while adding the search feature on the web to competitors such as Google, OpenAI, Perplexity. Fortunately, they made this expected update at the end. Claude, who has previously responded only to his own educational data, will be able to access the data on the web on current issues. In addition, Claude users will be able to search directly on the internet via chatbot.

OpenAI Introduced new AI models with sound -oriented

The features of artificial intelligence to transform the voice speeches and conversations into text are increasingly developing. We saw the last example of this from OpenAI this week. OpenAI introduced the new sound-oriented artificial intelligence models GTP-4O Transcrib, GTP-4O Mini Transcribe and GTP-4O Mini TTS. The GTP-4O Mini TTS can do it so that you can look natural as if you were reading the texts loudly. On the other hand, the GTP-4O Transcrib and GTP-4O mini transcribes the audio files and decipher their speeches and removes their written text.

New features added to Google Gemini: Audio Overview and Canvas

Google added Artificial Intelligence Tools Audio Overview and Canvas to Gemini, an artificial intelligence assistant. Audio Overview can convert any document to a voice podcast in just a few minutes. Canvas, which is defined as an interactive field of study, makes it easy to create and edit content by providing the opportunity to change the tone, length and format of the texts.

Audio Overview, which has previously been specially for Google’s notebooklm service, can now be used by all Gemini users and does not require subscriber to the paid plan. However, for now it only supports English language.

New AI video tools introduced this week:

Stability.ai introduced a new video tool that allows you to activate a fixed image using the camera angles and movements you want.

Topaz Labs, which attracts attention with the AI ​​vehicle that makes it high resolution, is now doing the same for visuals. Gigapixel V8.3.0 introduced by Topaz is the fastest model to uptake the images of the visuals.

Pika, one of the most impressive tools for editing video with AI, has been added to a new feature this week. Users will now be able to select a certain element in the video and only change their movements.

Roblox introduced the new AI tool that can create 3D objects based on the text. This AI tool can be quite useful, especially for game developers who want to create in -game objects.

Shorter than the world of artificial intelligence: