And surely a risk that has to be part of any national biodiversity security review.

Media class have forgotten how to celebrate success.

That's even worse for Robbins.

"Why don't you blockade the Strait of Baltic"

- General Keith Kellogg

I don't think he knows what a strait is.

#bbcR4

People think Mamdani just does it naturally. He works at it, he's translating to normal people speak.

Not that I'm pro Zionism but his redefining is a stretch.

I hope the new Mummy movie is good. Personally I was disappointed that Universal gave up on it's most recent mossters universe. I enjoyed the the Tom Cruise version.

The best.

Which is why we isolate the workloads in their own network.

When you configure an analysis, you're saying this is my "Bed pan workflow". So that's you context for the analysis. You choose the attributes and steps of a workflow to capture. If you just care about cost of bedpan service, you're not capturing anything about the patient.

So software captures specified attributes from workflow data and converts to vectors, it doesn't store the original attribute (because that's not needed to produce analytics). This all has to be configured, out of the box it can't do anything.

It's just vectors though. A "1" in one data set doesn't mean the same as a "1" in another. The vectors are contextual to the analysis being conducted.

"Metaverse" was a waste of time. NFT is interesting as an idea, but wouldn't be missed if it didn't exist.

LLMs, that's a technology, like it or loathe it, that is here to stay. Will it be abused, of course, like any technology. Does it have its uses, yes.

Remember also, these system aren't storing data like a name or an address, they're storing just vectors.

That's why it's important to not have attributes that can be combined. And also to not have the datasets in the same database, and not use the same keys.

Isn't it interesting though that no one seems concerned by the systems used in UKSV. 🙂

No, there was no steering committee in that process.

No, because each analysis is separately configured and generates its own dataset. And as I say, in each analysis, you choose on the attributes you want to capture, so those attributes wouldn't be capture anyway and so aren't present to derive from.

They can't though, because there are very strict checks and balances that have to be passed for projects to go live, steering committee and stakeholders have to sign off.

From that specific workflow's perspective, that data isn't relevant, it's just bed number X.

Not in my experience. Technical and solutions architects highlight the risks. They take pride in their work. Vendor's specialists too tend to take the side of the security folk because they don't want reputational risk.

It isn't even loosely correlated. It's like this, say a nurse had to take a bed pan to a bed. All that is capture is that a nurse had to take a bed pan to a bed, where the pan had to be fetched from, when it was recovered, cleaned, etc. Not interested in the nurse or the patient.

Yeah, but ministers are so far removed from the operational end of these implementations. There are really good experienced technical folk in the NHS and they know best practice too.

Re side note, thankfully an MP doesn't do the actual design and architecture. 😄

For example, we might develop a model that is good for analyzing procurement workflows to look at the differences in different NHS trusts.

That's something that's often overlooked, we use these tools to create our own models too.

Also, we don't always use their models, the platform is extensible, so we can plug-in models of our choosing. It depends on the use cases.

We're analyzing it, it doesn't leave our data centers. We don't give them the keys, and they don't have access to the network.

No, there's not patient data. It's just about which workflows are used, i.e. how often is a cancer patient is seen, how often, length of consultation, etc. The specific patient isn't important for the workflow analysis. It's more about are processes and procedures consistent or is there variance.

It's standard encryption algorithms, we generate the keys, they don't have those. So they've never be able to decrypt anyway without the keys.

Doesn't work like that. Workflows are turned into vectorized data for the models to analyze. When an analysis model is configured you can name it, like "Update patient notes flow", but it's the workflow actions that are captured not the data of the actions (not the notes).