One of my favorite quotes about artificial intelligence (AI) isn’t from a
data scientist or tech industry analyst. It’s from a doctor.
When asked if AI would eventually replace radiologists, Dr. Curtis
Langoltz of Stanford University pithily replied “No, but radiologists
who use AI will replace those who do not use it.”
This great
insight is highly applicable to IT. Yes, there are many ways we can use
AI methods to fully automate a broad range of DevOps tasks. Companies,
including CA Technologies, are actively involved in the development of
solutions that enable our customers to do exactly that.
Algorithmic
machine learning, however, doesn’t just empower systems to perform
tasks and solve problems autonomously. It also makes them great active
partners with human beings. In fact, much of what machines learn they
also wind up teaching.
The synergy between AI and human
radiologists, for example, stems in part from the fact that digital
systems can differentiate about 200 levels of gray in a diagnostic
image—compared to only about 16-20 that are discernable by the human
eye. Train an AI system with enough images, and that precise cognitive
power can more effectively detect that something is going on.
But
for the most effective diagnostic process, you don’t just depend on
that detection alone. You use that detection to empower a human
diagnostician who can apply a broad understanding of pathologies and
deep experience with the complexities of individual patients to deliver
the highest quality care.
In DevOps, we can do the same. We can
use AI to capture insights that teach us how to continuously optimize
our workflows and processes. We can also use our AI learnings to push
our work up higher on the value chain.
More specifically, the synergy between AI and human intellect can:
Make development smarter.
The speed, quality, and efficiency of development pipelines can be
affected by all kinds of subtle factors. A less-than-optimally designed
API, for example, can be a small but chronic stumbling block to everyone
who has to use it. Scrum outcomes can be undermined by anything from a
particular type of technical challenge to a nascent personality
conflict.
By capturing a rich set of DevOps metrics and applying
machine learning to those metrics, development leaders can discover
process bottlenecks and skilling shortfalls. They can better coach
individuals and promote team collaboration. The result: a better working
environment that facilitates digital agility for the enterprise and
higher satisfaction/retention for valuable employees.
Make ops smarter.
Enterprises are running increasingly volatile and complex workloads on
increasingly hybridized infrastructure. At the same time, the tolerance
of internal and external users for latency and outages continues to
approach zero. There are also real costs associated with performance
problems.
The elastic capacity of public and private cloud does
much to help with workload volatility. But adding cloud capacity also
has its costs—and end-to-end application performance often depends on
back-end systems that are not cloud-based. So not every performance
issue can be solved by simply throwing more capacity at it. Nor should
it be, if a rearchitecting can fix a bottleneck less expensively.
Here
again, AI can teach us a lot. We can uncover opaque interdependencies
in processing load and data throughput. We can spot conditions when it
may make business sense to throttle cloud costs that aren’t
cost-justified. We can even better understand the real-world
conditions—whether patterns in customer behaviors or our own marketing
programs—that are driving our demand spikes and troughs. All of this
helps us deliver consistently responsive digital experiences at a cost
that makes good business sense.
Make security smarter.
AI is already being broadly implemented in security solutions such as
endpoint protection and threat response to automate the detection and
neutralization of anomalous activities in the enterprise environment.
But effective multi-layer security isn’t just about finding and stopping
exploits. It’s also about building applications that are themselves
inherently less vulnerable to hacking. This is the essence of DevSecOps.
AI
has huge potential value here. We are writing a rapidly growing volume
of increasingly sophisticated code. It is very easy for subtle
vulnerabilities to hide in that code. As our development practices
become more complex—often including multiple contractors—it becomes more
difficult to understand exactly where and why these vulnerabilities
were introduced into our code. Machine learning can teach us the answers
to these questions, so we can more proactively secure our data and our
businesses.
It’s especially interesting to consider what may
happen as we start to apply AI to DevSecOps across our organizations, as
well as within them. More diverse inputs enable machine learning to
discover more factors that impact code pipeline performance. By
aggregating our knowledge about how we build, deliver and secure our
code, we are all likely to benefit with better practices and stronger
guardrails.
Source: https://devops.com/when-ai-met-devops-machine-teaching/
Friday, August 17, 2018
When AI Met DevOps: Machine Teaching
Reviewed by A
on
2:52 AM
Rating: 5
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment