|
Current State
|
|
Applications today consists of large amount of software
code. The application software consists of many
API.
These API are integrated with application features,
functionality and use cases to perform complex
operations in a seamless manner. The applications
code (API)
consists of database calls, computations,
extract/output data, display
revalent data in UI and so on. They are customized to
perform in cloud, mobile devices, web applications and
mainframe.
|
| |
|
Basic Intelligence
|
|
The applications are embedded with functional, rule-based
validation, dependency management and exception handling
such as catch zero divide
error, wrong data type input (numeric/string), missing data
and so on, to name a few. These typically use
|
| |
|
IF-THEN-ELSE, boolean, CASE syntax and exception constants
|
| |
|
that are
available in programming languages along with custom
code/API.
These results in a fixed set of failure
prevention -
zeroth level of intelligence.
|
| |
|
Intelligent Automation of Exception Handling
|
|
Some applications initially tried to autonomously
solve issues that were encountered on a day-to-day
basis. Typically the exception handling was enhanced
to autonomously perform corrections and restart or
continue forward from the point of failure in
mission critical/driver processes -
daily sales, inventory level, stock-trade data,
lookup data, etc.
This is a good starting point to embed
intelligence.
|
| |
|
The big issue with this approach is, it may work well
in several cases, but can also result in hung
process or application. If the key/driver job
execution excountered an exception, it would start
a new execution job. This causes resource
contention - too many executing jobs overloading the
system by
starting and failing in an infinite
loop - locking the system. This scenario has
been improved by having a counter or logic to skip
starting new execution and perform additional
intelligent analysis. Such jobs were enhanced to
have autonomous monitoring and dependencies checks,
that will stop
(abend)
the job.
|
| |
|
In the world of robotics, when a hardware encounters an
infinite loop, it is considered as
Singularity.
The vector matrix computation gets stuck in a infinite
state and the motion of a robot gets stuck or locked
state, requiring a restart or fixing of the issue.
|
| |
|
Embedding Intelligence
|
|
To start, the applications can be enhanced with steps shown below.
|
| |
|
1. Use statistical functions and features to monitor
the behavior on a constant basis.
|
|
2. Training data to be setup to enable
machine learning and use
relavent algorithms for unsupervised (autonomous) learning.
|
|
3. Analyze every new scenario or issue encountered so that
a possible solution can be achieved or treat as a new
scenario.
|
| |
|
It will react intelligently to see
if the scenario can be resolved without human
intervention.
|
| |
|
Autonomous Applications
|
|
In many applications, when a primary server fails
and application automatically switches over to a
stand-by server, some human interaction may be
necessary. Many
application-to-applications communication may need
to be reconfigured by human intervention. This will
slow down the response and could result in loss and
interruption of business.
|
| |
|
Use of AI to Achieve Humanless Autonomy
|
|
Here is a simple case of two
applications (A,B), with both systems having some
AI engine, that can inter-communicate and make
continuous adjustments. This can result in systems
operating in normal mode constantly.
|
| |
- System A goes to standy-by mode (server/hardware)
- AI engine informs System B to switch all System-A communication to stand-by setup
- System-B updates configurations and all dependencies
- AI system constantly monitors operations on System A and B
- AI engine automatically analyzes the reason for failure of the System-A primary
- AI engine determines that it is critical hardware failure - to be procured
- AI engine does complete analysis and creates a request for replacement hardware
- AI engine checks availability of replacement hardware within the organization
- AI engine creates a purchase order for replacement hardware due to unavailability within the organization
- AI engine analyzes hardware vendors and their stock levels and lead-time using past data
- AI engine makes a decision to pay a premium for immediate/urgent procurement to avoid System-A downtimes
- AI engine decides if human effort is needed for repairing hardware after procurement
- AI engine creates a timeline of hardware procurement to installation and risk analysis
- AI engine starts the procurement process
- AI engine selects the best vendor for hardware procurement
- AI engine upon arrival of hardware performs thorough inspection
- AI engine starts the installation process
- AI engine monitors human and machine installation process
- AI engine performs end-to-end test on the newly installed hardware on System-A
- AI engine certifies that System-A is functioning normally with fully functional primary and standy-by
|
| |
|
Currently each activity explained above, has significant human
activity. Most activities are done by humans and decisions are
taken by humans. Budgets, finance, approvals and many more are
very totally performed by humans. In future human could be
removed most activities using AI enabled systems. To achieve
this level of automation, some thresholds need to be set for
AI engine to adhere to and at what point human may enter in
and amend the flow of action.
|
| |
|
This is a totally autonomous AI based system with least amount
of human interaction. Many current management and
workforce will be in shock if a system were to perform at
this level of autonomy. The AI system is continuously
making management decisions and taking actions at free will,
with the goal to keep the systems functioning without
downtimes.
|
| |
|
Factory of The Future
|
|
In the mid 1980s, one of the course project in Automation
and Robotics - was computer simulation of
"Factory of The Future".
All aspects of our studies and
computer simulation, has
now become Factory of The Present,
Industrial robots welding, AGV transferring materials,
gantry robots lifting heavy assembly etc. Car
manufacturing companies are aggressively researching
to have humanoid robots work with humans with high degree
of efficiency.
|