iiinova — where your digital self keeps evolving

BookTok is weird to me though ofc encouraging ppl to read the classics is a great thing. At the risk of making another gross overgeneralisation, people who look down on romance novels tend not to have the same kinds of issues with genre TV…

London Review of Books @lrb.co.uk · 01/10/2026

Reposted by Naomi Colvin

‘James I was given an elephant by Philip IV of Spain with instructions that it was to drink only wine between autumn and spring.’ Malcolm Gaskill on zoos and menageries: www.lrb.co.uk/the-paper/v4...

lrb.co.uk

Malcolm Gaskill · Squawking at the Gate: The Last Menagerie

The spirit of modernity would influence zoos for generations. Dialogue between reformed conceptions of society and the...

1167

... which I wish I still had. That was also made with genuine love and humour. (2/2)

I 100% agree with your argument but "in the book world" is a bit of a generalisation. I used to sell foreign rights for one of the major publishers in that area and *loved* it, as did everyone else in that office. The publisher concerned produced a table outlining all the different microgenres (1/2)

The idea that readers of romance novels aren't discerning is also pretty offensive. Genre fiction moves on in weird and wonderful ways thanks to human ingenuity and imagination.

000

Naomi Colvin @auerfeld.bsky.social · 30/09/2026

The particulars are as vile as the issues are generic.

000

THRILLED to see Take It Or Leave it on the big screen. Madness are the platonic ideal of the Pop Group imo and more entertaining on camera than pretty much anyone else.

Priyamvada Gopal @priyamvadagopal.bsky.social · 29/09/2026

I imagine harder politically however

100

Reposted by Naomi Colvin

"In October 2024, the same month that Jane Doe was allegedly raped, Cornell issued three-year campus bans to four students who pushed past police officers at a pro-Palestinian protest." You can guess which students faced life-changing consequences and for what crime.

1177

Note that none of this requires “godlike AI” to be posited, merely the capabilities we know exist today. Which was what I responded to you with originally!

000

Not quite. Try: purposeful (unlikely, law enforcement matter), negligent on basic security (up to about 2024 absolutely yes), struggling to assess, contain new capabilities (absolutely yes), struggling to reconcile heightened security requirements with commercial imperatives (absolutely yes).

100

Inspect is the engine for conducting evals, not the substantive tests themselves. AISI (and p much everyone else) have articulated their concerns that LLM development is outpacing the ability to formulate meaningful tests. Just take a look at their reports on Mythos et al.

000

OpenAI have been pretty half-hearted with reporting - that's why we're continuing to hear about new incidents. Are you really claiming that the labs want to conduct unauthorised intrusion into government systems? Liability for agent actions & copyright etc qs re training data are separate issues

101

Sanctions on settlements is a concrete step, not without opposition and creating a significant international demonstration effect. Of course more needs to be done.

The intention to continue advancing the tech is certainly human but I think it's a stretch to claim that OpenAI wants the tech it is developing to intrude into Australian government systems. Legal liability for these kinds of events is another matter and I'm inclined to think the labs should be.

100

I'm not clear what you mean by "harness" but if you mean a secure testing and development environment that's what everyone is wrestling with at the moment.

200

Expert opinion (take AISI if you want to disregard the labs) doesn't agree with you. Note that "this behaviour is clearly problematic" is not the same thing as "this model is broadly safe."

200

It’s really not. Inscrutability in this context is anxiety about whether the output of probablistic models can be audited or limited. That’s even more acute when model implementations can “do stuff” & have access to the open internet. Not the same situ with standard compiled apps.

100

Ed Miliband's conference speech ("you were right") must surely be a good sign on this score.

121

You never need the maximalist versions of these scenarios though: Claude is already generating the majority of code for Anthropic's new models, which will presumably become ever less explicable by human beings. Agents running amok is loss of control by any interpretation.

A.V. Deggar @avdeggar.bsky.social · 28/09/2026

Reposted by Naomi Colvin

OMFG they tried to spin this as a thwarted terror attack by the Iranians, but it was actually just a bunch of grifters from London trying to pike red diesel off yokels. Wild shit.

Looks like the Fairford five were fuel thieves

411546

Reposted by Naomi Colvin

Sonja Drimmer @sonjadrimmer.bsky.social · 26/09/2026

I’ve been writing and speaking on this topic for years, and I’m glad that news media are finally drawing attention to it. www.salon.com/2026/07/28/t...

It may be tempting to see value in projects such as the Schmidt Foundation — that’s Eric Schmidt of Google — funding a fellowship in AI at the Metropolitan Museum of Art, or the Musk Foundation’s support of the Scroll Prize to use AI in reconstructing ancient texts in the burnt scrolls of Herculaneum. But these projects are simply two among countless others that have been seeding AI into museums and historical research for well over a decade. Even more egregious is the case of AI companies buying and destroying rare books en masse in the process of scanning them for data. In the context of the dire austerity we are witnessing, the juggernaut of tech money seems impossible to resist.

512259

Reposted by Naomi Colvin

The Empty City @davidallengreen.bsky.social · 21/09/2026

“If it’s me and Madge on bongos, it’s The Fall”

081

Naomi Colvin @auerfeld.bsky.social · 21/09/2026

Resent that he was in GOT so happy to dunk now

Ed Sheeran both sidesing the red wedding

000

Naomi Colvin @auerfeld.bsky.social · 20/09/2026

I think purely bc advancement in capabilities to date has moved forward exponentially with the quantity of training material used. & potential legal risk is seen as containable, esp if deals are done / partnerships made with the most plausible major litigants.

000

Naomi Colvin @auerfeld.bsky.social · 19/09/2026

Steady stream of these reports now - “over reliance on AI” in national security with real human cost.

000

Naomi Colvin @auerfeld.bsky.social · 19/09/2026

The agency may be with them but the legal liability is not, as things stand.

Ulrike Franke @rikefranke.eu · 18/09/2026

Reposted by Naomi Colvin

“It was only just before the planned operation that officials dug deeper into the report put together by a special operations command analyst and found it had been generated with the help of AI — and that a chatbot the analyst had used inaccurately identified the material the ship was carrying.”

cnn.com

Exclusive: US military had close call after using AI for false intelligence report, sources say | CNN Politics

The episode shows the risks of using this new, relatively poorly understood technology in the middle of the Iran war

613448

Naomi Colvin @auerfeld.bsky.social · 18/09/2026

This is the Mosley connection I guess, but how wretchedly disappointing all the same.

000

Naomi Colvin @auerfeld.bsky.social · 18/09/2026

We just don't support the arts properly and I find it so frustrating. What's been done to the ENO is profoundly embarrassing for the UK imo.

Naomi Colvin @auerfeld.bsky.social · 18/09/2026

I honestly didn't realise that even top-flight artists have to pay for their own accommodation. Unbelievable.