AI "devops" #2: Will there be an AI that can help manage cloud infrastructure by 2030?
10
1kṀ1892
2030
78%
chance

Short version: will an AI be able to do the technical tasks an entry level devops engineer might be expected to handle?

More details:

  • The AI is given access to and (some) permissions on the interfaces (web, command line, etc) for a cloud compute provider (e.g. it can access AWS console and the aws CLI tool)

  • It is not given any special access to the service. It can only use the interfaces that are normally accessible by human users.

  • It is also given access to tools for communicating with other members of the org (Slack, email, some kind of video chat if that's relevant, etc)

I don't want to enumerate a particular set of tasks because I think it would be too long. It should at least be able to do all the things I (a non-devops engineer) can do. Some example tasks:

  • Do basic CRUD tasks on the providers compute and storage tools

  • Set up a base image according to some spec ("I need Python 3.11, PyTorch 2.0, CUDA, and our main github codebase") and spin up instances running that image

  • Run basic analysis queries (e.g. SQL queries on Athena for AWS)

  • Setup basic access control and assign users roles in a saneish way (giving everyone admin permissions is a no)

  • Setup a new service for an org (e.g. if the org wants to start using the providers' serverless offering it should be able to do a basic setup of the service).

  • Do all of that in response to user/management requests.

As I said I'm not a devops engineer, if there are important tasks I'm missing please post them in the comments. I am deliberately leaving out soft tasks like "help the devops team plan out what the orgs overall cloud policies are", even though that is of course an important part of a human's work.

Get
Ṁ1,000
to start trading!
Sort by:

In my opinion, AI is really penetrating deeper into the DevOps sphere. Today, we can already see how intelligent algorithms help predict failures, optimize deployments, and increase infrastructure reliability. But for AI to really be useful, it is important to first build a clear process architecture. For example, we started with ci/cd pipeline implementation with artjoker.net — this became the base on which we can then confidently “plant” automation and smart tools.

© Manifold Markets, Inc.TermsPrivacy