Atlantic puffin in full breeding plumage stands outside its burrow looking soulful

Bliss

I haz Teams - I’m gonna be mostly offline for a week now tho, I’ll give ya a shout when I get back

The UK, a different flavour of shit show

Thank you, I will! Always good to connect with fellow geeks

No, AI, I Do NOT Think Nuclear War is a Good Thing No matter what Google would have you believe.

No, AI, I Do NOT Think Nuclear War is a Good Thing

thecosmicshamblesnet...

Do you know what it's like to have AI misunderstand your work to the extent it could wipe out civilisation? Dean Burnett does.

#AIslop #Misinformation #Nuclear #War

I had had far too much experience with official corporate statements to ever trust a word of one!

The best one can hope for is to cross one’s fingers that they have competent people on the job whose recommendations are actually listened to

What worries me most is how many people accept such assurances at face value

One of my favourites. Epic track. That and ‘Stripsearch’.

Low numbers with big impact does sound more like a data & records management issue tbh

Why most GDPR audits fail. And how to fix your privacy audit approach. YouTube video by GDPR Register

Beyond box-ticking - ICYMI: my hot takes on how to make data protection auditing a useful exercise. I had great fun discussing this with Krete Paal from GDPR Register on her webinar last week.

youtu.be/50eAaYS0kvk

Your regular reminder that “no evidence of compromise” could, in theory, mean any of the following:

-no log data exists to look at
-evidence was not wanted, therefore not looked for
-evidence was cleverly removed/concealed
-semantic chicanery around what “compromise” means

That’s an effect of how NT society treats autistics, not with autism itself.

There are valid reasons as to why autistic young men are vulnerable to radicalisation - a lifetime of being bullied, rejected, held to toxic standards, taken advantage of and laughed at makes the experience of being welcomed into an elite group of ‘truth-seers’ a compelling lure.

This, all day long.

Also, detoxify the corporate culture - if you’re pissing *that* many people off, something is rotten beneath the surface

…and the UK gov’t - which is always on a mission to replace public policy with commercial automation in education, social care, health, justice; to deny support to disabled people; to suppress dissent, to scapegoat and ‘Other’ particular marginalised communities, to exploit the housing crisis…

Blocking targeted ads doesn’t prevent the data surveillance that enables targeting, it just hides it from your view, giving you a false sense of assurance.

This is a corporate diligence and compliance issue, not a matter of individual choice.

It’s too big for any one person to tackle but together, we can reduce the harms by:
-saying “no” to tracking as end users
-dumping developers who build in unlawful functionality and dataflows
-exercising our data rights, especially objection and consent withdrawal
-paying attention and caring

This massive network of unfair, unlawful activity causes real-world harms. It systemically disadvantages and discriminates against vulnerable, marginalised and minoritised people; incentivises division; fuels radicalisation and conflict; contributes enormously to ecocide; wastes precious resources

And that’s without even getting into the ePrivacy aspects of how these data-slurping features are implemented, and lack of valid consent for those mechanisms.

If unlawful processing were a criminal offence, the surveillance advertising industry would be the world’s biggest organised crime syndicate

>>
4. Unambiguous - failed where ‘consent’ is inferred from another action (eg installing or using a function/feature)
5. Revocable - fail; unless the withdrawal is applied to each onward recipient and use (LOL)
6. Evidenced - fail; a chain of “they said yes, trust me bro” is not ‘evidence’.

1. Informed - fail; lack of specific info, use of deceptive language and no mention of how far the data may travel
2. Freely-given - failed where ‘consent’ is a condition of basic functionality
3. Specific - fail; unless each recipient Controller is individually listed
>>

Note the deceptive language - it’s not ‘sharing’ when you don’t know where it’s going, to whom or why. That is unilateral mass *surveillance*, datafication and monetisation.

It also fails the legal threshold for valid consent to processing….

Yes, this is what I'm always on about! Nothing beyond computer correlation of old data is happening. Veracity & accuracy tasks are being farmed out to human sweat shops in the global south, and they are only addressing small fraction of non-expert corrections. Models have no discernment or insight.

you genuinely love to see it.

Only if you’re not one of those liberal types who subscribes to the fancy notion that feeeemales are actually real people.

….which is why the Datafication Age is such a damned horror show

Relevant, suitably accurate and precise, reliable, timely data requires enormous cognitive and administrative labour to maintain.

Generalisations, approximations, assumptions, and misappropriations however, are super easy/fast/cheap to spew out, recycle, build upon and mistake for useful info.

The big data challenge is that data is super easy to generate, obtain, re-use; cheap to store and transmit; and widely mistaken for ‘objective fact’, or static’ truth’. Faith in data is misguided and dangerous.

Data is merely an artefact produced by human design and intention.

For ‘AI’ to reliably function as advertised, it would need such a huge amount of human effort in data management, quality assurance and change control that it could never be cost-effective.

The true cost of fit-for-purpose data is starting to emerge, and it is way higher than anyone anticipated.

Which is correct in data protection terms. The minimum data to confirm someone’s identity does not include either sex or gender.