Model Ops 101
In the Industrial Revolution era, public libraries democratized knowledge by making it freely accessible to everyone, not just a few privileged elites. In the digital era, Google and Wikipedia democratized knowledge shared on the internet. In the social media era, Facebook, Instagram, and TikTok democratized thin social connections.
In the AI era, Model Ops will democratize AI. Model Ops is like a public library, Google, and Amazon.com for AI, all rolled into one, like this:
Creating AI (Write a book)
Data scientists create algorithms like writers write books. But instead of sentences, paragraphs, and chapters, they use Python, R, and tools like Datarobot or Dataiku.
Sharing AI (Put the book in the library)
AI must be reviewed and approved like acquiring books for a library. Model Ops has librarian functions, too, to approve models before they can be used.
Finding AI (Searching the card catalog)
Once approved, the general public must be able to find AI models like books in a library. Model Ops is like Google for AI models in the enterprise, including ratings, like on Amazon.com.
Deploying AI (Check-out and read!)
Deploying AI models is more complicated than checking out a book from the library. AI must be deployed carefully and have myriad performance, security, bias, and management considerations. Model Ops makes it easy to deploy AI models by anyone, anywhere.
Sharing AI (Comment, review, debate)
The internet makes it easy to review and share book recommendations. Similarly, Model Ops helps the community review, discuss, and share AI-fueled insights.
Moderation (Wikipedia moderation)
Like Wikipedia entries, AI requires moderation to adjudicate accuracy, impact, and bias. Model Ops provides tools to monitor, test, and tune AI. Model Ops helps remove models. Each algorithm’s value, effectiveness, provenance, performance, quality, and safety is mediated. Bias is more easily identified and mitigated.
Regulation (Wikipedia edit history)
Model Ops tools document decisions about AI. Like the edit history of a Wikipedia article, this informs AI forensics. In heavily regulated industries like Pharmaceuticals, AI documentation is regulated.
* * *
The Dawn of a New AI Culture
With books and websites, creators, librarians, tools, moderators, patrons, and the community must work together to safely scale its use. Model Ops provides this process for AI. It helps data engineers, data scientists, business users, IT operations, and app developers work together.
At the dawn of the industrial era, knowledge was controlled by a few intellectual elites. Data science is the same. But Model Ops tools promise to democratize AI and unlock creativity, innovation, and agility.
