L16 - AI @ Edge

Question 1

What applications dont need AI at the edge?

Answer 1

• Slower applications
• these are applications where we do not need real-time predictions, like big data analysis and modelling
• in this application the only solution is to leverage to the power of cloud/physical machines
• This implies that the closer we move to the production line the more we need edge computing
  
  • most industrial applications require a deterministic real-time response to make decisions
  • Distributing computation power is key

Question 2

Challenges of edge computing?

Answer 2

• many application could be achieved with both edge or cloud computing
• edge computing very often means:
  
  • low computing power (CPU/Memory bounds)
  • lower power consumption –> aka no GPU acceleration
  • unreliable internet connect
  • large number of devices to handle

Question 3

To achieve an intelligent system what does our system need?

Answer 3

• Sensing - acquiring data from all the sensors (physical or virtual)
• Planning - choose the optional decision giving data and experience (model)
• Active - commanding the Cobot and monitoring that the actuation has completed successfully

Question 4

What is data in relationship to AI?

Answer 4

Data is the asset - dont need big data need quality data

data + application + human = AI

Question 5

In order to create an intelligence ready company we need?

Answer 5

• a strategic, and coherent, data acquisition
• data warehouse or data lake
• mapping of processes and elements
• a new work organisation with data scientists in the loop