Now you have an omnipotent model that reads, sees and listens. Everything at once

Eight years ago, when Nvidia was still a company that made graphics for video games, the company pointed out to something that is starting to enter the conversation: physical robotics. They are the robots with artificial intelligence integrated to behave autonomously. Like a ChatGPT with arms, ears and eyes. It has rained a lot since then and It’s now when we’re starting to enter that future. However, Nvidia has continued to experiment with that way of making the physical and digital worlds converge, and its latest product is the Nemotron 3 Nano Omni. An AI model that sees, hears and reads the physical world. Omni Models. These models are multimodalbut in a much stricter sense. While the models we use every day require separate channels to process and generate audio, text, image and video, an omni model is designed to be inherently multimodal. This implies that they use a unique neural network architecture trained end-to-end so that the interaction between models and stimuli is more natural, faster and capable of recognizing more nuances. An example is an AI that can “see” what a camera captures, analyze the entire situation and give feedback to the user more quickly than one that can do the same, but whose text model has to ask the video model what it has seen and then generate the content. In even fewer words: it better imitates the way humans perceive and respond to the stimuli of the world. Integration. And that’s what Nvidia affirms What Nemotron 3 Nano Omni can do. In the same architecture, it is a model that integrates vision, audio and language capabilities to eliminate the fragmented workflow of current AI agents. According to the company, it is built on a hybrid architecture of mixing experts (AIs trained in various subjects) with 30 billion parameters, of which 3 billion are for inference. It has been designed as a model that is nine times faster than separate models and has three times the performance of other open omni models, consuming 2.75 times less computing power in tasks such as reasoning from a video. Okay, but why?. That is the key question, beyond the numbers and the raw capabilities of this technology. The use cases detailed by the company are the following: Agents: power those agents that navigate graphical user interfaces, reasoning based on the content on the screen and understanding what they are seeing in real time and persistently. The native input resolution is 1920 x 1080 for that HD visual understanding. Documents– Interprets graphs, tables, documents, screenshots, and mixed media inputs. Comprehension audio and video: is able to understand what he sees and hears to maintain consistency in his interpretation instead of reasoning based on disconnected models. For professionals. What is clear is that Nemotron 3 Nano Omni is not something that is launched with the goal of being something for the masses like other AI models that we see every day. Nvidia focuses it on something business, a tool that can be accessed through platforms like Hugging Face and to be deployed on local systems such as DGX Spack or Jetson. That is, it is not something available to everyone. The interesting thing is that it is a technology that is strongly pushing the narrative of agents as omnipotent entities, and it fits with the speech latest from Jensen Huang, CEO of the company, that AI will not come to take away our jobs, but to ‘micromanage’ us. Image | Nvidia In Xataka | There is a company that has grown 3,000% in the stock market, even beating the performance of Nvidia: Sandisk

Mercadona and the rest of the supermarkets spend tons of paper on receipts that no one reads. Now they want to change it

You go to the supermarket, you buy a couple of things (just enough for dinner), you go to the checkout, they give you the ticket, you put it in your pocket and you leave with the bag in the direction of the parking lot. Pure routine. Our daily bread. If the employer’s retail achieves its objective, there is one element of that scene, however, that will change radically. Which? That ticket that you will end up throwing away without even reading it. What has happened? Every year supermarkets print millions and millions of strips of paper in which in many cases only a handful of articles appear, so they end up in the garbage can without anyone having even looked at them. It is a waste, a waste of resources. For chains like Dia, Lidl or Mercadona, but also for the environment. So Asedas (Spanish Association of Distributors, Self-Service and Supermarkets) has had an idea: they want us to start printing receipts only when the customer requests it. What do they want? The news I advanced it on thursday theEconomist. Asedas has proposed to the Government that it slightly tweak the regulations that regulate tickets so that they are no longer printed systematically. That does not mean that they are no longer issued or that the customer no longer has a receipt that clarifies what they have purchased and how much they have been charged. The change would focus on support. The idea, clarifies Ignacio García, head of Asedas, is “that the ticket continues to be generated electronically for control purposes, but that it is printed on paper at the consumer’s request.” That is, the user can request the physical or digital ticket. Right now, remember theEconomistthe regulations provide that supers deliver the receipt in two ways: either in paper or digital format. What’s happening? Since not all clients are in favor of handing over their data (including email) to the chains, in the end they have no choice but to print it. Not only that. The employer’s data They show that many of the times we go to the supermarket we buy only a handful of items, so the receipts show small transactions, for low amounts that we do not even review. Result: those papers end up in the trash as they are printed. It is not even strange for the customer to reject them when the cashier offers them to them. Is it that serious? “Our companies have been confirming for years that, in about a third of operations, the ticket is abandoned at the checkout line,” confirm Garcia. It is not surprising if we take into account the data on the shopping basket managed by Asedas. According to their estimates, 30% of the operations registered in supermarkets respond to almost urgent visits, during which we take home at most four products and spend less than 10 euros. In 60% of cases, purchases involve between five and 25 products with average tickets of between 10 and 50 euros. Only the remaining 10% actually respond to large purchases. In practice, the fact that all operations end up reflected in a receipt means that the supers generate about 5 billion tickets that require the use of almost 4,500 tons of paper and a million-dollar expense. Is it important? Beyond the millions of receipts that are printed each year and the cost that this entails in tons of paper and euros, Asedas’ proposal is interesting for at least two reasons. To start with who throws it. Asedas presume to be “the first food distribution business organization in Spain” and cover 19,200 retail stores and 495 wholesalers. Between your partners Companies such as Mercadona, Lidl, Aldi or Dia appear. Another key is that its idea is in line with what is already done in other European countries. For example, in 2023 France said goodbye to the generation of tickets by default precisely because of the amount of paper it consumed. That doesn’t mean they no longer exist, but they must be requested. In the Netherlands, Switzerland and Sweden there have also been changes related to the generation of receipts. In Spain itself, some large chains they take time moving towards the digital ticket. Images | Xataka Mobile and Wikipedia In Xataka | There was a time not too long ago when the future of supermarkets seemed like Amazon Go. Now Amazon Go is dead

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.