Why Not Have One or Two Smart Software Platforms? Great Idea!

September 8, 2017

I have been writing about online for decades. In my Eagleton Lecture in the 1980s (I forget the year I received an ASIS Award and had to give a talk as part of the deal), I pointed out that online concentrates. It is not just economy of scale, a single online service operates like a magnet. Once critical magnetism is achieved, other services clump around or two that central gizmo. There are fancy economic explanations; for example, economies of scale, convenience, utility, and so on.

Now couple the concentration with one of the properties of digital information flows and we have some exciting things to think about. In that Eagleton Lecture, I used the example of the telegraph to illustrate how even inefficient forms of information movement can rework social landscapes.

The point is that concentration and flows are powerful forces. When data flows through an institution, that institution comes apart. When online power is concentrated, the nature of “facts” changes. Even the unconscious decisions of a widely used online service alter how individuals perceive “facts” and set their priorities. (Are you checking your email now, gentle reader?)

When I read “Facebook and Microsoft collaborate to simplify conversions from PyTorch to Caffe2,” I thought about that decades old Eagleton Lecture of mine. The write up describes what seems to be a ho-hum integration play. Here’s a passage I highlighted:

The collaborative work Facebook and Microsoft are announcing helps folks easily convert models built in PyTorch into Caffe2 models. By reducing the barriers to moving between these two frameworks, the two companies can actually improve the diffusion of research and help speed up the entire commercialization process.

Just a tiny step, right?

A few observations:

  • Clumping is now evident between Google and Walmart
  • Amazon and Microsoft are partnering to make digital gizmos play well together
  • Fusion (both data and services) are the go-to idea for next generation information access and analysis services.

The questions which seem interesting this morning are:

  1. Why do we need to have just multiple artificial intelligence platforms? Won’t one be more efficient and “better”?
  2. How will users understand reality when smart software seamlessly operates across what appear to be separate functions?
  3. What will regulators do to control clumping which will command the lion’s share of revenue, resources, and influence?

Yep, just a minor step with this PyTorch and Caffee2 deal.

Online can be exciting and transformative too.

Stephen E Arnold, September 8, 2017


Comments are closed.

  • Archives

  • Recent Posts

  • Meta