[pbourke]

> t isn’t just illegal, the US infrastructure is intentionally designed to make it extremely difficult to do at a technical implementation level, with the idea that it would hinder potential abuses

You think that the country which went from never leaving the earth to landing on the moon in 8 years, from 0 to hundreds of planes and dozens of ships per month in a few short months (WW2) couldn’t stand up a national infrastructure for contact tracing? Don’t confuse lack of will with lack of capability.

[jandrewrogers]

This challenge isn’t just legal and operational, the required computer science and data infrastructure software literally doesn’t exist. I am a subject matter expert on this and was consulted by a few national governments to see if it was feasible. Large-scale technical experiments were run in the US several years ago.

If you use off-the-shelf computer science, you have an O(n^2) problem on a data model that is growing by several petabytes per day. Existing systems can’t even ingest data at that rate in a useful way, never mind analyze it. Much more scalable algorithms exist; this was literally the topic of my supercomputing research back in the day. If you can severely constrain the data model in certain ways that are not possible in many countries, or you have a small enough population, you can kinda sorta brute force it. In practice no one is AFAIK, because off-the-shelf software simply isn’t designed for it, even at small scales. The US has an unconstrained data model combined with extremely high scale and topological complexity. You can’t magick state-of-the-art exotic data infrastructure into existence.

I did “what would it take” studies earlier this year for a couple governments. We are talking about 100k lines of advanced C++ for a system design that has never been built before, only theorized. It isn’t something you do over the weekend. Even in the most optimistic scenario, building the software would take at least 18 months. For COVID, there is no point (but they are still interested for the next pandemic).

We know from prior experiments that if a disease is endemic in a population and extremely draconian lockdown measures are not realistically possible — both true in the US and some parts of Europe — then contact tracing is basically an exercise in futility unless you have some exceptionally advanced data infrastructure (we don’t) and the legal authority to use it (the US doesn’t nor many countries in Europe).

Don’t confuse the lack of capability with a lack of will. The fact is that I was approached by multiple national governments on this subject very early on because they know I design this type of data infrastructure.

[kaba0]

Not who you are responding to, but really interesting comment, thanks!

I am not too knowledgeable on the topic, may I ask you what sort of confidence is required for a well working contact tracing application? A naiv those-who-live-together is so utterly useless that it was rejected? And do I get it right, you meant to real-time track everyone and if someone turns out to be infected, everyone gets quarantined who they met in the preceding days? What percentage of contacts can “slip through” so that it is still effective? Isn’t there a working trade-off like counting people who work together and family or is it still subject to unimplementable exponential explosion?

Or am I asking the wrong questions as these are epidemiologic ones rather than technical? If so, excuse my lack of knowledge on the topic.

[pbourke]

I’m not sure what you’re talking about. China did this in a few weeks. I’m not even talking about automated tracing of cell phone signals - just databases and people making phone calls.