Comment by spwa4

Yeah, I bet this is involved in the problem. This is describing the NHS as essentially nonfunctional, hospitals have their own IT and attempts to migrate to NHS infrastructure have largely failed. No doubt that makes sure people having long-term illnesses will spend most of the time being long-term ill waiting in NHS lines.

Of course, the complaint from economists is long-term sick people being occupied 12+ months getting better is worsening at about 7% per year, over 40% in the 2020-2026 period ( ... and in the supposed big success cases Palantir Foundry brings 28% improvement.

Also: collecting the parts and posting here.

https://archive.ph/Hg7xN https://www.computerweekly.com/opinion/Inside-FDP-part-1-Und...

https://archive.ph/IE9oi https://www.computerweekly.com/opinion/Inside-FDP-part-2-Del...

https://archive.ph/zp5cU https://www.computerweekly.com/opinion/Inside-FDP-part-3-The...

The TLDR: because the NHS leadership has systematically refused to upgrade their IT[1], does not actually make the data they get available, their IT is considered unhelpful by hospitals and clinicians which leads to "gaps in the data" [2], and because NHS management will not either hire people to do analyses and will not learn/understand things like SQL to do it themselves, they cannot act on the data.

SO ... what Palantir brings is:

1) a quick AI-driven way to make apps to implement particular workflows. AI writes apps instead of "shadow IT". Just AI written applications shared by frontline staff on a "solution exchange".

2) a data ingestion process that makes up whatever data is missing (apparently this is called "ontology": the AI model fills missing data in with what it thinks is happening. No mention of hallucination and it's effects).

Somehow Palantir does this without any kind of change in IT at the edge. Ie. hospitals, mental health wards, ... keep all their existing IT (he mentions this means no investment is needed in it at all)

3) a way for NHS management to "cooperate", which appears to mean to query this data that does not involve any kind of actually knowing what they're doing or hiring analysts. Which apparently leads to delays (no doubt because analysts ask for clarification which management doesn't know or care about). "Plain-English analytics". Oh and in case you're wondering: he descirbes it as a POSITIVE that it is fast but inaccurate.

Apparently "somehow" a frustrating part of this is that "data extraction is frustrating" and it clashes with traditional NHS data engineering practices (translation: existing NHS analysts say that the data Palantir produces is bullshit)

It ends on a "positive" note: no worries, it'll be pushed through past these people calling bullshit. Just makes things harder and slower. No explanation as to how this statement vibes with him leaving.

[1] They still use access databases running on single machines NOT even VMs

(He comments many of those systems cost hundreds of thousands of pounds "before counting the staff needed to run them". Note that this is supposedly a source of savings of going with Palantir, but also they won't be replaced, because that costs too much ...)

[2] ie. you go to another hospital, even with a referral, often they don't know about that referral. This has caused staff at hospitals to bypass the NHS organization in many cases, and even the NHS itself acknowledges that bypassing the NHS administration is a necessity.