Embedded AI

6 Trends for 2023

29. November 2022, 16:00 Uhr | Tobias Schlichtmeier
Embedded KI

In the digital age, innovation cycles are short and reaction times are correspondingly short. In 2023, companies from all industries will continue to face new challenges in order to remain competitive and innovative. Viacheslav Gromov explains how they can meet these with AI.

Local artificial intelligence (AI) technologies continue to advance and will reach a new peak record in 2023. Viacheslav Gromov, Founder and CEO of AITAD, reveals the top six embedded AI trends that will play a key role in shaping 2023.

1. dealing with the energy crisis remains an important item on the corporate agenda.

The energy crisis is already affecting companies in 2022 and will continue to cause many changes in 2023. Computing resources on servers and cloud costs are increasing significantly as these systems are very energy intensive. Even more product manufacturers will turn to decentralized systems with embedded AI. These run on inexpensive small semiconductors due to space constraints, are resource-efficient, and act autonomously in the field. Pure cloud or intensive edge applications will get a bump.

2. supply shortages drive Predictive Maintenance.

Due to the current acute and persistent supply problems from electric drives to tubes in the mechanical engineering industry, agile and flexible manufacturers are deploying new products from new suppliers in shorter cycles. Predictive maintenance is becoming more important as the lack of experience of the newly deployed components needs to be addressed in terms of their quality. In this way, the end customer can be prevented from feeling the disadvantages of the compromise solutions, since failure cases can be remedied at an early stage without downtime or even damage to the image. In addition, predictive maintenance is becoming increasingly in-depth and effective due to the further increasing computing power of smaller semiconductors - such as those with the latest NPUs (Neural Processing Units) - which requires immediate processing on site in real time (embedded AI) due to high data volumes.

3. increasing computing power leads to deeper and more complex interaction with the user or operator.

The trend towards increasing computing power in the smallest and cheapest semiconductor computing units, accompanied by the advancement of algorithms in artificial neural networks (ANNs), leads to possibilities of deeper and more complex interaction with the user or operator. User interaction such as person-space recognition, occupant detection, or keyword spotting will begin to add value by responding to human behavior and emotions.

4. embedded AI research and development is booming.

What was previously only possible on larger systems will become feasible with inexpensive systems in the two- to three-digit euro range. Concrete consequences include technology leaps in incremental learning and speech separation. The former allows local machine learning models to be adapted to usage (e.g.: right-handed or left-handed?) or environment (e.g.: is the process plant in dry heat or in damp cold?), while the latter leads to better, local speech recognition of the operator's voice in noisy environments.

5. The shortage of skilled workers continues to worsen.

The need to make household and medical devices in particular smarter, and thus more time-saving while reducing the need for supervision, will grow exponentially. This could be, for example, a surgical laser device that listens directly to the surgeon's instructions without an assistant nurse operator thanks to voice recognition, or a toothbrush that monitors dental status at the ultrasound level, saving preventive visits to the doctor. Vacuum robots or room ventilators are also becoming more reliable thanks to local object and dust analyses, so that household chores take up less time.

6. trend hardware-as-a-service (HaaS) becomes normality

Global systems competition will intensify in the coming year. Western industry will no longer be able to prevail against Asia through classic product performance, such as sheet metal bending and process functions. Instead, the already emerging trend of hardware-as-a-service (HaaS), or the marketing of equipment and machine usage in place of unit sales, will become more commonplace. By using embedded AI to detect wear and tear and analyze user behavior, manufacturers will achieve optimal lifetimes and machine availability while decreasing maintenance costs. Service and customer orientation, but also sustainability, are consequently coming to the fore in such leasing, rental, subscription and service models.

Anbieter zum Thema

zu Matchmaker+

Das könnte Sie auch interessieren

Verwandte Artikel