11 ms·

Agreed. If the resource usage can be optimized further, it'll be feasible to train specialist models both on-prem and from scratch. That would sidestep the liab

Agreed. If the resource usage can be optimized further, it'll be feasible to train specialist models both on-prem and from scratch. That would sidestep the liability and privacy issues of current cloud-based offerings. Google's search appliances did the same thing before they were retired. Hospitals were especially keen on them. They eliminated HIPAA risks because data never left the hospital intranet, but then Google eliminated the product line in favor of cloud offerings.

jebarker 3y ago
My belief, which may be false and not informed by direct experience, is that the search appliances didn't really work that well since the hyperlink based search ranking method of the day doesn't really work on relatively small data holdings of an individual organization. Can anyone confirm if that was the case?
- pushfoo 3y ago
  TL;DR: It seemed to work for organizations which fit the appliance's assumptions > hyperlink based search ranking method of the day doesn't really work on relatively small data holdings of an individual organization In my experience, it depends on what you mean by relatively small. Cloud editing and permissions aside, there seemed to be the 3 key factors: 1. Structure: hub-and-spoke sites interlinking crucial documents and pages 2. Culture: interdepartmental and personal rivalries which demand links as credit 3. Size: lots of employees creating and maintaining sites and content I was too young to understand it then, but all of these help an intranet's content resemble the web of the 90s and 2000s. That's the era Google's algorithm was designed for. The results weren't always perfect, but they were nearly always better than the alternatives. In-house or pre-existing alternatives tended to be awful, especially not having any search at all. Keep in mind, all of this was ~10 years ago. My memory isn't perfect and a lot of things have changed since then. I even made sure some of them did. It was my job at a few, but the relationships and boundaries between these were complicated. For convenience, let's say n_observed roughly equals 4.
  - jebarker 3y ago
    Good point on structuring internal knowledge holdings to enable PageRank. The organizations I've worked in seem to put zero thought into this and instead just leave it up to individuals.
    - pushfoo 3y ago
      TL;DR: There was no intentional structuring, just Conway's law[1] in action I think you misunderstood: the organizations where PageRank worked best also put zero thought into it. I think it's more like holdovers from 90s and 2000s management trends emergently created an intranet structure which fit the assumptions underlying PageRank. Even interdepartmental tensions may have helped due to imitating the company behaviors of those eras. 1. https://en.wikipedia.org/wiki/Conway%27s_law https://en.wikipedia.org/wiki/Conway%27s_law
- snarg 3y ago
  I used it in, what, 2004? 2005? It was pretty good, especially compared to today's typical corporate on-prem search options, e.g. Sharepoint or Confluence, both of which are almost certainly inferior to a physical filing cabinet.

jebarker 3y ago

My belief, which may be false and not informed by direct experience, is that the search appliances didn't really work that well since the hyperlink based search ranking method of the day doesn't really work on relatively small data holdings of an individual organization. Can anyone confirm if that was the case?

pushfoo 3y ago
TL;DR: It seemed to work for organizations which fit the appliance's assumptions > hyperlink based search ranking method of the day doesn't really work on relatively small data holdings of an individual organization In my experience, it depends on what you mean by relatively small. Cloud editing and permissions aside, there seemed to be the 3 key factors: 1. Structure: hub-and-spoke sites interlinking crucial documents and pages 2. Culture: interdepartmental and personal rivalries which demand links as credit 3. Size: lots of employees creating and maintaining sites and content I was too young to understand it then, but all of these help an intranet's content resemble the web of the 90s and 2000s. That's the era Google's algorithm was designed for. The results weren't always perfect, but they were nearly always better than the alternatives. In-house or pre-existing alternatives tended to be awful, especially not having any search at all. Keep in mind, all of this was ~10 years ago. My memory isn't perfect and a lot of things have changed since then. I even made sure some of them did. It was my job at a few, but the relationships and boundaries between these were complicated. For convenience, let's say n_observed roughly equals 4.
- jebarker 3y ago
  Good point on structuring internal knowledge holdings to enable PageRank. The organizations I've worked in seem to put zero thought into this and instead just leave it up to individuals.
  - pushfoo 3y ago
    TL;DR: There was no intentional structuring, just Conway's law[1] in action I think you misunderstood: the organizations where PageRank worked best also put zero thought into it. I think it's more like holdovers from 90s and 2000s management trends emergently created an intranet structure which fit the assumptions underlying PageRank. Even interdepartmental tensions may have helped due to imitating the company behaviors of those eras. 1. https://en.wikipedia.org/wiki/Conway%27s_law https://en.wikipedia.org/wiki/Conway%27s_law
snarg 3y ago
I used it in, what, 2004? 2005? It was pretty good, especially compared to today's typical corporate on-prem search options, e.g. Sharepoint or Confluence, both of which are almost certainly inferior to a physical filing cabinet.