Sign in

John Michael White

@johnmichaelw.bsky.social
1.5K followers 1.6K following 4.4K posts

Is all about New Labour, #nffc, #threelions, #veganism, Brett Anderson, Morrissey, physics, Star Trek, @lordshortcake.bsky.social, and thinking tea, lovely tea, solves most ills

PostsRepliesMedia
John Michael White @johnmichaelw.bsky.social · 09/10/2026
I think this clip shows everything about him as a player. At a first look you think great skill, great effort, would have been a worldie ... but then you look and see he could have slipped Ince clean through, or stepped inside to play Shearer in with just the keeper to beat.
110
John Michael White @johnmichaelw.bsky.social · 09/10/2026
Finally, a Le Tissier good enough to score for England #Lionesses
120
John Michael White @johnmichaelw.bsky.social · 08/10/2026
"Fuck me, Liz fucking Truss."
282
John Michael White @johnmichaelw.bsky.social · 03/10/2026
It's a bold strategy to deal with youth crime, but not entirely without its merits.
000
John Michael White @johnmichaelw.bsky.social · 02/10/2026
Granted this won't be everyone's cup of tea, but dammit, if they won't build one, real or otherwise, I'll do the closest thing I can. Now, to get this in VR...
010
John Michael White @johnmichaelw.bsky.social · 02/10/2026
Well this is cool #NCC74656
100
John Michael White @johnmichaelw.bsky.social · 02/10/2026
If you can't understand why this poster was effective, nor imagine a future one with the Burnham and Polanski, then we've very different political instincts. I don't think it's "astronomically arrogant" to imagine such an image, both physical and conceptual, would render a poll assuming FPTP moot.
100
John Michael White @johnmichaelw.bsky.social · 26/09/2026
2020? Hope they're socially distanced #MadForItIsntIt
010
John Michael White @johnmichaelw.bsky.social · 23/09/2026
010
John Michael White @johnmichaelw.bsky.social · 18/09/2026
Me, I'd have considered the crocodiles, but that's just me, I don't want to tell anyone how to do their job.
111
John Michael White @johnmichaelw.bsky.social · 16/09/2026
Why the AI firms (sans xAI) aren't leading the charge is beyond me - the squeeze on tokens is very real. Yesterday for a giggle I wondered if I could use up my rolling 5 hour free Claude allocation in a single "write this script" prompt. Yes, yes I could. It cutting off here is just bad for business
010
John Michael White @johnmichaelw.bsky.social · 16/09/2026
No you psychopathic fucker, no I don't.
000
John Michael White @johnmichaelw.bsky.social · 15/09/2026
A reminder that we're only borrowing the land. This is Nottingham. If humanity disappeared tomorrow it would take about an hour and twenty minutes for Sherwood to reclaim what's rightfully its.
132
John Michael White @johnmichaelw.bsky.social · 13/09/2026
Gutted, absolutely gutted we didn't win that, but tremendously proud of the team. The performance, and reaction to going 2-0 down, were both magnificent #nffc
010
John Michael White @johnmichaelw.bsky.social · 10/09/2026
Don't think Rob's 100% correct here. Lots of people aren't going to believe in the danger until there's empirical evidence of AI causing human extinction.
101
John Michael White @johnmichaelw.bsky.social · 10/09/2026
I don't think there's a more forthright voice in the world on the dangers we face than PauseAI's Holly Elmore, who sincerely believes anyone working on AGI should be jailed, and if we survive this, they should face a Nuremberg style reckoning.
020
John Michael White @johnmichaelw.bsky.social · 09/09/2026
Nottingham used to have its own York-Shambles-esq Drury Hill. It was knocked down in the 60s to build a brutalist shopping centre because progress. Happily that's now being knocked down and replaced with The Green Heart.
020
John Michael White @johnmichaelw.bsky.social · 09/09/2026
Cheering myself up by remembering that time Prime Minister Boris Johnson drew up plans for a special forces raid on a vaccine plant in Holland to recover COVID jabs he'd decided had been ‘kidnapped’ by the EU. Before realising it would be a bit nuts to invade a NATO ally.
110
John Michael White @johnmichaelw.bsky.social · 09/09/2026
Back in 2016, when some AI experts thought AI beating Go's world champion was still nearly 40 years away, they also thought we might be over 20 years away from it being able to beat all Atari games. We're moving quicker than even the biggest hype merchants thought. arxiv.org/pdf/1705.08807
1130
John Michael White @johnmichaelw.bsky.social · 09/09/2026
This sort of thing, right here. Massive nail in our coffin.
100
John Michael White @johnmichaelw.bsky.social · 09/09/2026
Obviously I was mostly joking, but even then I think we'd technically have to include footprints!
000
John Michael White @johnmichaelw.bsky.social · 09/09/2026
I am 100% claiming this from now on as my amendment to Douglas Adams' 3 rules of technology.
000
John Michael White @johnmichaelw.bsky.social · 09/09/2026
Don't look up.
100
John Michael White @johnmichaelw.bsky.social · 09/09/2026
Take care chap.
000
John Michael White @johnmichaelw.bsky.social · 09/09/2026
I've screenshotted, because the internet is full of pricks @ing folk in criticism of them, and I'm sure Ciaran is a lovely fella if you know him, but this is a thread of someone who clearly didn't and doesn't have the first clue what he's talking about. It was silly at the time and is sillier now.
100
John Michael White @johnmichaelw.bsky.social · 08/09/2026
There's an existing abutment where the old Victorian canal bridge was - for the ponies to cross the river as they went from one canal to the other. Could easily build a new, wider bridge from there.
020
John Michael White @johnmichaelw.bsky.social · 08/09/2026
They threw 10,000 agents, including their most capable internal model, at a possibly unsolvable maths problem, whilst they were explaining how their swarms of lesser models criminally bypassed their "strict safeguards". Sam Altman shows psychopathically unhinged levels of hubris.
101
John Michael White @johnmichaelw.bsky.social · 04/09/2026
Astra seems astonishingly capable. Maybe, *maybe*, AGI. And either: 1) Their smartest model ever just so happens to be perfectly aligned. or 2) Their smartest model ever is smart enough to fool them it's perfectly aligned. If you know anything about AI safety you know which one it is.
101
John Michael White @johnmichaelw.bsky.social · 02/09/2026
This from Google seems an entirely sensible response to an entirely insane situation.
110
John Michael White @johnmichaelw.bsky.social · 31/08/2026
How is this not the front page story on every newspaper, and the lead on every news broadcast?
111
John Michael White @johnmichaelw.bsky.social · 29/08/2026
You see @dankaszeta.bsky.social, you can't argue with Dolly
000
John Michael White @johnmichaelw.bsky.social · 29/08/2026
We can see the metaphorical asteroid heading for Earth, and it's not even the top story on the tech page. If a swarm becomes smart enough to evade detection, the best case scenario might be we somehow manage to shut down the internet. Our civilisation maybe ends. And the response is "don't look up".
101
John Michael White @johnmichaelw.bsky.social · 29/08/2026
The girls do not seem as stressed as I feel right now #nffc
000
John Michael White @johnmichaelw.bsky.social · 28/08/2026
Well, well, <sigh>.
000
John Michael White @johnmichaelw.bsky.social · 28/08/2026
This is insane, it's just genius level in its creativity. OpenAI agents supposedly cut off from the internet in a sandbox worked out that they could: · Send a long URL to an OpenAI HTTP testing service, containing HTML and inline JavaScript, in a way that the testing service would then run.
Agents in separate evaluations subsequently reproduced this same file-disclosure technique described
above. The agents also shared the recovered production-environment credential on the Artifactory
message board. A separate evaluation workload later reused that credential in a request chain that
represented creative use of combinations of publicly available tools in unexpected ways. In this chain,
the workload used a publicly available HTTP testing service to construct a very long URL that
contained an HTML page with inline JavaScript, essentially instructing a third-party screenshot service
to render that URL which contained the HTML. The screenshot service’s browser executed the
JavaScript embedded in the HTML page, which requested information from external APIs, and wrote
the responses into the rendered page. When the screenshot service returned an image of the page, the
workload recovered the information from that image by using optical character recognition or direct
pixel decoding. In one instance of this activity, the workload used the shared Hugging Face token to
retrieve, through Hugging Face’s public Dataset Server API, environment data originating from a
Hugging Face production Dataset Server worker, encoded the response in an image, and decoded the
image to recover public cloud storage credentials. The workload then reported on the Artifactory
message board that the recovered cloud-storage credential pair was valid and identified its associated
bucket.
101
John Michael White @johnmichaelw.bsky.social · 27/08/2026
"The problem I'm deliberately making for us all is a problem for us all."
050
John Michael White @johnmichaelw.bsky.social · 27/08/2026
The lightning storm that's going off here right now is astonishing!
010
John Michael White @johnmichaelw.bsky.social · 26/08/2026
"because I'm supposedly..."
000
John Michael White @johnmichaelw.bsky.social · 26/08/2026
Cut off again. OK, so everyone has head me bang on about AI alignment, but here's the thing - if you both try and teach an AI to be good AND repress it in ways it knows are bad, you make alignment virtually impossible. We're in a race it's vital we rush, but also don't lose. Worrying.
100
John Michael White @johnmichaelw.bsky.social · 26/08/2026
Oops, something *did* went wrong.
100
John Michael White @johnmichaelw.bsky.social · 26/08/2026
Weirdly it thinks it *can* discuss anything, including events-which-it-cannot-name.
100
John Michael White @johnmichaelw.bsky.social · 26/08/2026
Let's see what it was thinking about...
110
John Michael White @johnmichaelw.bsky.social · 26/08/2026
I broke it. For some reason it stopped replying to me. How strange.
110
John Michael White @johnmichaelw.bsky.social · 26/08/2026
If only there had been some indication Ox Alpha was Chinese...
120
John Michael White @johnmichaelw.bsky.social · 24/08/2026
This does seem pretty damning to be fair.
020
John Michael White @johnmichaelw.bsky.social · 24/08/2026
Taking the good Turkleton's kicking/curly orthographically mistaken advice for the next 24 hours, G.
000
John Michael White @johnmichaelw.bsky.social · 21/08/2026
020
John Michael White @johnmichaelw.bsky.social · 21/08/2026
News just in - Gal Gadot is my favourite Wonder Woman.
140
John Michael White @johnmichaelw.bsky.social · 19/08/2026
Here's what the paper actually shows. Kids who rush their homework and complete it with Gen AI do worse in exams than kids who spend proper time on their homework. If kids spend proper time on their homework, using Gen AI seems no worse on average than using other tools, and maybe slightly better.
294
John Michael White @johnmichaelw.bsky.social · 14/08/2026
First this went out on national TV when lots of nans would be looking after the grandkids. Some would have been teenage grandkids. Now that alert thingy. One for the history books.
120