Research center imec and KU Leuven present energy-efficient AI chips for robots
Research center imec and KU Leuven have developed a new type of chips for robots that work with artificial intelligence (AI). The chips combine a digital and analog coprocessor: in this way, the calculations that the AI performs can be accelerated, and energy is saved at the same time.
Research center imec and KU Leuven have developed a new type of chips for robots that work with artificial intelligence (AI). The chips combine a digital and analog coprocessor: in this way, the calculations that the AI performs can be accelerated, and energy is saved at the same time.
Research center imec and KU Leuven have developed a new type of chips for robots that work with artificial intelligence (AI). The chips combine a digital and analog coprocessor: in this way, the calculations that the AI performs can be accelerated, and energy is saved at the same time.
Two years ago, imec developed a new type of chip with analog technology. The calculations that that type of chip performs are performed directly in computer memory. That makes it possible to perform most of the operations ten to a hundred times more energy-efficient than in a digital accelerator. For another part of the operations, in turn, the computational precision and programmability of a digital accel erator is better suited.
To combine these two advantages, imec has developed a processor that merges analog and digital technology. This led to the development of a new chip: the DIgital and ANalog Accelerator (DIANA).
The hybrid chip automatically performs the operations on the coprocessor best suited for each specific task. For some applications, such as pattern recognition, an analog processor is best suited. Other applications, such as reasoning about those observations, also require a digital coprocessor. The chip from KU Leuven and imec combines the two processors so that computations are always performed in the most energy-efficient way.
The chips combine a digital and analog coprocessor: in this way, the calculations that the AI performs can be accelerated, and energy is saved at the same time.
Latest insights & stories
Recticel lanceert de RE-engineered IP PIR isolatie
Recticel streeft naar hoogperformante isolatieoplossingen die de koolstofvoetafdruk voortdurend verkleinen, zonder concessies te doen aan de strengste technische vereisten.
MOBILIDATA
In Vlaanderen werken we vandaag al aan de mobiliteit van morgen, onder andere met het Mobilidata-programma. Hiermee ontwikkelen overheden, bedrijven en onderzoekers samen innovatieve, technologische verkeersoplossingen voor de weggebruiker. Denk aan beter routeadvies, verkeersmeldingen op maat en intelligente verkeerslichten. Maar deze nieuwe, geconnecteerde mobiliteit en knowhow stoppen natuurlijk niet aan onze landsgrenzen. Er is dus nood aan internationale samenwerking om kennis uit te wisselen en gezamenlijke projecten op te zetten om dit ook te implementeren.
Digitale soevereiniteit garandeert veiligheid in de publieke cloud
Wanneer bedrijven overwegen om naar public cloud te migreren, worden ze soms tegengehouden door beveiligingsrisico’s of compliance- en governancebeperkingen. Daarom is er interesse in digitale soevereiniteit, legt Gwénaëlle Hervé uit, Public & Sovereign Cloud Lead bij Proximus NXT.