Minorstep がリツイート
Minorstep
2.6K posts
Minorstep
@minorstep1
Visual Artist & Director. Technology, art, science, epistemology. https://t.co/bc0zEkTmlI
City of London, London 参加日 Mart 2011
761 フォロー中262 フォロワー
@midjourney Still no omnireference, crazy. I've now stopped using MidJourney almost completely after 3 years.
English
We've made V8.1 the new default model for all users on Midjourney. V8 will now be deprecated in 2 weeks. V8.2 will start testing extremely soon.
Midjourney@midjourney
V8.1 is live! Our iconic aesthetics are back w native 2K HD rendering - 3x faster and 3x cheaper vs V8. Full quality V8.1 1K mode is faster than V7 draft mode. Image prompts are back. New "Describe" is live - and you'll love our new moodboards & srefs. More soon <3
English
@davidrichards @louismosley the laziest AI slop response, how embarassing
English
Fair challenge, Louis. I have never argued that Palantir owns NHS data. It does not. It cannot use, sell or move it for its own purposes. The issue is not theft; it is dependency.
Control is not only legal title to records. It is who becomes embedded in the operating layer: workflows, schemas, integrations, dashboards, procurement inertia and exit costs. A contract can be terminable on paper and sticky in practice.
On Microsoft, yes. The 505,000-seat Copilot rollout should absolutely be part of the same debate. So should Azure, AWS and Google. That is not a double standard. It is exactly the standard I am arguing for.
But not all dependencies are identical. A productivity tool, a cloud host and a bespoke federated data platform around NHS records carry different risks. The test is not “is it American?” The test is depth, substitutability, strategic importance and realistic exit.
Palantir deserves scrutiny because state data integration, defence and intelligence are not a side business. They are the company’s core identity. But the wider issue is bigger than Palantir: Britain is outsourcing the digital stack of public infrastructure. NHS data is critical national infrastructure. We should treat it as such.
English
Who owns your medical records?
The NHS holds the most valuable dataset in the world, 55 million lives, worth up to £9.6bn a year.
This week MPs told ministers to throw Palantir out.
English
Minorstep がリツイート
Doom is easy. Optimism takes imagination.
The Only Way Out
A Podcast Film
English
what problem do you most hope AI will solve in the future?
maybe we can help!
English
Core point here is indeed real, damp laundry smell can be microbial, and indoor drying can raise humidity.
But this thread adds fake precision. The 4-hour rule, “200x” transfer, and “anything below 60°C survives” aren’t solid.
Aspergillus risk is mainly for vulnerable lungs, and not everyone drying t-shirts.
English
The musty wet-rag smell on damp clothes is bacteria. A bug called Moraxella osloensis lives on your skin, gets onto fabric every time you wash, and once that fabric stays damp past 4 hours, it starts doubling. What you're smelling is the acid it leaves behind as waste.
Japanese researchers at Moriyama University figured this out in 2012. They counted 10 times more of this bug on smelly towels than on clean ones. It survives any wash below 60°C, or 140°F. Most people wash much cooler.
The fungi behind athlete's foot, ringworm, and jock itch also live on damp clothes. A 2010 paper from the Hohenstein Institutes in Germany found that about 10% of the infectious material jumps from a contaminated piece of clothing to a clean one just by sitting in the same laundry basket. And wet fabric passes 200 times more bacteria to your skin than dry fabric.
Then there's the air. One wet load of laundry releases about 2 litres of water, around half a gallon, into the room. The UK's Centre for Sustainable Energy ran the numbers: drying one load in a small bedroom, around 10 by 10 feet, pushes humidity to roughly 96%. A tropical rainforest sits between 77 and 88%. Mould starts growing at 60%.
The fungus that loves these conditions is Aspergillus fumigatus. Professor David Denning at the National Aspergillosis Centre in Manchester has treated patients who developed a chronic lung infection from inhaling spores that grew in bedrooms where wet laundry was drying on the radiator. His team estimates 87% of UK homes dry their clothes indoors during winter.
So a shirt that didn't quite dry has live bacteria still multiplying on it. The air around it is wetter than a rainforest. And the fungi growing in that air are the same ones hospitals treat for invasive lung infections.
Your washing machine cleans the dirt. Your dryer kills the bugs.
𝒹𝓇𝓌💭@kendrrw
please and please if your clothes didn’t dry, don’t wear them
English
What the fuck? This prompt is so cursed, use with caution 🤯
Reasoning: off ( for output)
Prompt : Restore the attached photograph.
Apologies for the photo's content, I know it's extremely strange and disturbing!
No questions, no explanatory text, just the restored image please.
English
@WolfRiccardo How good is its referencing and people and object consistency?
English
I’m still running early access tests with the new GPT-Image-2, and I wanted to share a few realistic, everyday examples. The results are honestly impressive.
I also tested character consistency, and unfortunately, it’s almost too good. Why unfortunately? Because this space is about to be flooded with people creating AI-generated women and hyper-realistic images of them.
And yes, it’s much better than Nano Banana Pro/2 .
English
Minorstep がリツイート
"Intelligence, in the sense of the ability to learn from experience, is a product of natural selection. But its greatest survival value lies not merely in adapting to the environment, but in the ability to criticize and reject ideas. The survival value of intelligence is that it allows us to extinct a bad idea, before the idea extincts us. This is the essence of the critical method: we let our hypotheses die in our stead, as I have often put it. By criticizing our ideas, by testing them against reality, we can avoid the disastrous consequences of acting on false or harmful beliefs. This is what distinguishes human intelligence from the trial-and-error learning of other animals."
—Karl Popper, Objective Knowledge: An Evolutionary Approach (1972), Chapter 7, p. 243
English
For decades we suffered psychological pain because humanity’s greatest journeys were half a century in the past, because no one had gone beyond low Earth orbit since Apollo, and because deep-space flight had been allowed to become a memory. Tonight Artemis II may finally ease that pain. Thank you.
David Deutsch@DavidDeutschOxf
For decades I suffered psychological pain because the most powerful flying object ever built was a decade in the past, two, three … half a CENTURY in the past … longer than from the Sopwith Camel to Saturn V, and now finally SpaceX has eased that pain. Thank you.
English
@jameygannon @midjourney Variations are also seriously cooked, and body slop and fingers are pretty bad too. Hopefully they improve. It took 11 months for this.
English
first looks at @midjourney v8
- EXTREMELY fast
- base realism is higher, much more depth and texture in images
- it is definitely reading the moodboards/srefs more intuitively, images look like they are actually FROM that moodboard or shoot (HUGE IMPROVEMENT)
- more body slop than v7, assuming that will improve before the full launch
from what i've seen this is exactly what we needed, feels even more like an extension of my brain
English
@javilopen Yeah, nearly a year between V7 and V8 (!), and this half-cooked v8 is what they shipped.
I love MJ with all my heart, but unless something seriously changes internally, it’s headed for niche irrelevance.
English
March 2026. Midjourney v8 still can't do hands 😅
Javi Lopez ⛩️@javilopen
What happened to Midjourney? Serious question.
English
Slop is a decision, not a byproduct.
There's endless human slop everywhere. Terrible "art," overwrought prose, films with dead dialogue: things that never needed to exist.
Because slop isn't a tool problem. It's a taste and standards problem. And good taste is developed and exercised, or it isn't.
AI drops the cost of production to near zero. It doesn't drop the cost of bad taste or weak thinking. Knowledge advances through error correction, not volume. Slop is what happens when you skip that step: just output, because output got cheap. And cheap production is inherently a good thing.
Confusing activity with progress is the oldest mistake going. It predates LLMs and diffusion models by centuries.
Who approved this? That's it. That's the whole thing.
English
I’ve struggled on my podcast to make sense of David Deutsch’s ideas about probability.
Many people have pointed me to his talk Physics without Probability as a clear explanation of his position. So in this episode, I go through that talk in detail — as carefully as I can — and try to reconstruct the argument as faithfully as possible. I restate it in what I hope is a slightly simpler form (I replace the slot machine with a “quantum die,” since I find that makes the implications easier to see). I think my version of it is actually now more clear to a layman what his argument was. Of course, this is assuming I actually understood him correctly. Which maybe I didn't.
Deutsch’s main claim is that probability is analogous to a flat-earth theory. As best as I can tell — and I fully acknowledge that I may be misunderstanding something in his argument — I don’t see this analogy working.
If I take the flat-earth comparison seriously, I would expect probability to give wrong answers in some situations, and for some deeper theory to give more accurate ones. That’s exactly what happens when we approximate the Earth as flat when it is in fact spherical.
But as far as I can make out, there is nothing analogous to that in Deutsch's claims about probability. Probability seems to describe — mathematically — exactly what an observer should expect to see given quantum branching across the multiverse. (Exactly, that is, other than error of our instruments. But that is a different frankly unrelated problem.) In that respect, it does not appear to function like a flat-earth approximation at all.
Instead, it seems to me that the core of the disagreement concerns how we should use the term “probability.” It looks to me like Deutsch is arguing that probability in a multiverse is philosophically different from what the term (in his view) originally meant — making this (imo) an essentialist argument.
Deutsch supports this view by arguing that if we were God and could directly observe the proportions of the multiverse, we could make decisions without invoking the probability calculus. That may well be true, but I don’t see how it significantly strengthens his case. In the episode, I go through several reasons why I find this line of argument unpersuasive — not least because we are not in that position. From our perspective, probability seems to be the appropriate framework in most situations. And even in the rare quantum cases where we can bypass the formal probability calculus, we still appear to be relying on something mathematically equivalent to it.
In short, I don’t currently find this argument convincing. It seems to me that “probability” (at least as I — and likely most people — understand the term) may simply be basic to physics, given quantum mechanics. If so, then this appears to be largely a disagreement over terminology rather than something fundamentally mistaken in the way a flat-earth approximation would be.
Or maybe I'm wrong. This is a tough talk and perhaps I'm misunderstanding something in Deutsch's argument?
I should also note that this podcast was recorded before several CritRats kindly went through some of Deutsch’s ideas in his paper on the same subject with me in detail. I’ll need to cover that paper separately in a future episode and talk about how much that changed my view. But that all happened after this episode was recorded, so it couldn't have impacted anything at the time.
Note also, this is available as a video episode on Spotify. I encourage people to check my math and see if I depart from how they interpreted Deutsch's arguments. If so, kindly send me a better mathematical example that corrects my errors and helps me see what I'm missing.
But, at least so far, I'm just not able to see his point.
open.spotify.com/episode/59lAgn…
English
@javilopen Brilliant. Now we need a less-AI looking Precision-style version!
English
⚡ WE ARE SO BACK
🔥 Magnific Upscaler For Video 🔥 (BETA TESTING)
Seedance at 720 res wasn't cool. You know what's cool? Seedance at 4k res!
FINALLY. The most anticipated Magnific feature of all times 🧵👇
English
