The AI ​​revolution: Intel’s challenge

For professionals, students, businesses and anyone who is passionate about technology and wants to immerse themselves in the near future, thus lifting a corner of that great veil that separates the present from the future of the technological spring that awaits us, all that remains is to take part in the next appointment : The International Innovation and Digital Fair . A 3 day event where the protagonists will be: over 100 events, training, B2B meetings, networking, culture, concerts, shows and entertainment. So, write down this date and this place: 13-14-15 June 2024, Bologna Fiere .


Returning to the " Artificial Intelligence Festival ", an important point of view was that expressed by Walter Riviera (Tech Lead EMEA Intel corporation). Riviera, at Intel since 2017 as AI TSS (Technical Solution Specialist), deals with AI and in particular "artificial intelligence projects in the Data Center sector" in the EMEA area (Europe, Middle East and Africa). Twenty minutes in which the Intel engineer discussed the revolutionary impact that AI will have and how Intel intends to address this challenge.

Regarding AI, attention immediately shifted to the macro-world ( Llm ). Acronym that stands for large language model (Large Language Models, based on machine learning). A machine learning model specialized in understanding and generating natural text. This language model underlies generative artificial intelligence like OpenAI 's ChatGPT .

But how did all this come about? Were wide language models ( Llms ), machine learning, generative AI created just for the purpose of making us ask ChatGPT questions? Suggest us recipes? Or help us study?

Source: AI*Festiva

In the beginning it was the command line

These applications represent the tip of the iceberg of the revolution that artificial intelligence will bring. But let's take a step back. In the beginning it was the command line; in "a few" years we have gone from: PCs managed via command line with glittering cursors flashing on green phosphor monitors and old cars with dashboards with analogue instruments to computers and cars managed by panels with a graphic interface ( GUI ). You might ask: why? The answer seems obvious but it isn't! The main objective was to facilitate visual human-machine interaction so that these tools reached as many consumers as possible. To achieve this objective it was necessary to allow even people without specialist skills to easily use a PC. Allowing him to exploit all the power of the machine. Even those who did not have particular IT knowledge had an additional means of interacting with a PC, making it do a job. Therefore an evolution dictated by the need to broaden the target to which electronic devices are intended.

From the command line to artificial intelligence

And today? Considering that we have moved from graphical interfaces to voice commands, you might wonder what the keyboard and mouse are for? They will probably remain but not as before, mutatis mutandis, these innovations aim to broaden the consumer base even further. Just think of older people who are not familiar with input devices. In similar cases, these innovations could allow sending an email using only voice commands. But the field of application could be further broadened, such as assistance. Support for non-self-sufficient people, monitoring them remotely with the use of a whole series of sensors managed by AI and capable of sending a whole series of alert messages at the right time.

Next frontier

Today, thanks to ( Llm ), we are able to make machines read, interpret and write natural text ( ChatGPT ), but the next frontier concerns multimodality . With multimodality it will be possible to accommodate new types of data as input. Data such as: images, videos and audio. This opens up a boundless universe of new possibilities. It suggests that AI is much more than what it is used for today. AI is a tool that can enrich every aspect of our lives.

Source: AI*Festival

References according to Intel

For Intel, AI is not an answer to the “What” but the “How”. What does all this mean? Intel breaks it down into four key points:

  • Accelerate innovation : an ever-increasing drive towards innovation and greater value maximisation.
  • Maximize value : What does Intel mean by maximize value? Data is turning the concept of value on its head. Given that it must be inviolable, information is so important that infrastructures are built around data. This perspective introduces a new concept, a new paradigm.
  • AI everywhere
  • Responsible AI
Source: AI*Festival

Federated learning

These algorithms create a collaboration between different actors giving life to a Federated model. To learn autonomously, artificial intelligence needs to train using enormous amounts of data. This enormous amount of data can be distributed and therefore taken from different sources. Moving and using information from existing databases poses privacy and security issues.

For example: the use of sensitive medical information. The Federated learning model technique tries to remedy this type of problem. A collaborative paradigm. With this method, it is no longer the data that travels and is exchanged, but the correlations, the parameters learned by the models trained on the data possessed by each Entity. Since the data is no longer exchanged and does not leave the place where it belongs, it enjoys greater protection and security. But even this technique is not without problems.

Intel, AI and PC

In response to privacy and security needs, given the need to train AI with enormous quantities of data, perhaps even having to move it remotely, Intel launched AI PC last December. A Personal Computer that contains three computational units ( CPU , integrated GPU , NPU ). Some software will reside in the NPU (Neural Processing Unit). Among the various tasks that this software will have is that of checking whether sensitive data is inadvertently being shared during normal activities. These functions are called: " Protection Guard ". Here, the direction traced, AI everywhere . AI at the service of cybersecurity.

Source: AI*Festival

Not only that, in this architecture, some of the processes executed in the background during ordering operations will no longer use the computational resources of the CPU or GPU but will exploit those of the NPU , continuing to be executed in the background but managed more efficiently. Intel has embarked on a real path that it defines as Responsible AI , in which the microprocessor manufacturer is committed to advancing AI technology in a responsible manner by limiting its potentially harmful uses. For this purpose, the creation of FakeCatcher can be traced back, a technology that allows you to verify the authenticity of videos in which people appear. Or the C2PA consortium of which Intel is a part which has developed a protocol for the authenticity of online content.

The article The AI ​​revolution: Intel's challenge was written on: Tech CuE | Close-up Engineering .