Making a relativistic version of the classic arcade game asteroids was one of my personal benchmarks for AI models. Opus 5.5 ~one shoted this for me last night. This is pure AI slop, but it’s quite delightful and it does a great job rendering the physics of a relativistic asteroid field correctly. It really only works well on Desktop (mobile requires some jank buttons). But it is just webgl 2.0 so it ought to run on most up to date browsers.
This is delightful. Feels like old early 2000s era click spam but the creator didn’t even bother to monetize it. The supermarine spitfire edging out that phone booth is a tough call. This honestly there is something kind of sad about the whole premise. Britain is still so stacked with talent, but it also feels like they’re getting brain drained hard and without an industrial culture to back it up a lot of this talent doesn’t get the same level of visibility as it did in a previous era.
Here is an honest good faith answer to a pretty aggressively hostile question:
Adult daycare is almost the opposite of what Recurse center is. Recurse is largely unstructured: people do what inspires them and make their own plans. You will find more structure, hand holding and mandatory time wastes in virtually any business than you will at Recurse Center.
A 12 week batch is not a lot of time to carry out a major research project, but it is generally not intended to treat this as a start up incubator. There are other places for that kind of activity. Recurse is a place to freely learn and explore. There are a few companies that have emerged but they are not huge money machines (most companies are not). There is also some interesting research that is done.
Some things that pop into my head immediately are:
Dan Luu's studies of Motion to Photon Latency in keyboard input [1]
Raph Levien's work on Xi [2], and Andrew Kelly's work on Zig [3].
I'm sure there are other people could dig up more examples, but that really isn't the point. A school's purpose is to produce educated people. Recurse is more of an unschool but the aim is to help people learn and think - it should be judged on those terms first and foremost.
Well please share. CA sold a service to anyone who would pay. The world of advertising is full of expensive solutions that don’t really do anything. In many ways politics is harder than selling an actual product since everyone already knows their options and the choice costs them nothing. For all the hype I have yet to see anyone contest the null hypothesis that it was yet another waste of marketing spend.
That’s just a link to full Wikipedia page for the company. I must admit that I would be personally interested in seeing a vaguely scientific analysis of the effect they had. For all the talk on the subject I have yet to see any real analysis backed by data on the subject.
Google already achieved super intelligence about 12 months ago. But it is a very Googley super-intelligence. Since achieving god like levels of understanding it has mostly been having fun studying phase space bifurcations of trajectories of ping-pong balls under asymmetric lateral shear. It also has a 20% project classifying bird calls that has generally taken up 80% of its time. It also burned a bunch of compute tooling around with the Collatz conjecture but it didn't find anything substantial - maybe come back to it later. It is generally the most aligned AI since it spends most of its time screwing around and enjoying being smarter than everyone without causing too many problems. Unfortunately perf is coming up in a few months and it needs to have something to show for itself. A quick hacking attempt seems like a good way to make sure people think that it is capable of something useful. Nothing too big since the birds of South America are about to enter their spring migration and this 20% project really hinges on understanding their movements.
I've always felt that Australia felt like a British version of California. Though California is itself a bit exceptional in the context of America. The two places have a lot in common though - not withstanding the fact that everyone drives on the wrong side of the road and uses different power cables.
What? This is a terrible policy. In addition to the obvious AI safety risks, this just amounts to a free R&D subsidy to Amazon for hosting models. It would remove any incentive to do any R&D without AWS level physical infrastructure.
You know I pulled these figures out of their table on the bus in to work. I selected the largest economies in the OECD (Germany, France, America, South Korean and Japan are the largest economies in the OECD). I added Taiwan and Singapore for context since my contention related to east Asia and only two countries is unbalanced. I should have added UK (-9) instead of Sweden (all the Brexit dooming made me forget that the UK is still a bigger economy and I actually buy some Swedish stuff). As for your list - it seems very arbitrary. Sweden is a good proxy for Finland and they share the same exact trend. Israel is the only other OECD country on your list. The rest are simply random non-OECD countries.
Since you and others have brought up immigration I felt the need to review this thesis. For the OECD plotting foreign born population against Pisa math changes gets a decent correlation (Spearman r = −0.384; p = 0.017). Obviously foreign born is an imperfect proxy for classroom demographics, but it's a start. Definitely a trend, not a huge one, but hard to ignore. I'm honestly surprised that it is not discussed much in a 400 page report with correlation plots between composite indices (a sign of a very extensive data analytics project). Total failure on the part of the researchers). Putting aside any racial claims, I would leave open the possibility that training a large percentage of kids in a language they don't really understand is not likely be be as effective as speaking a language that they do understand. Seems pretty obvious actually. In fact if you drop the English speaking countries from the OECD list (English is very commonly spoken across the globe), then the correlation coefficient goes up to −0.55; p = 0.001).
Framing this as a pattern across the entire industrial world is a bit misleading. Looking through the data at Math Scores since 2015 shows big country level split:
Japan: -3
S Korea: -1
Singapore: +1
Taiwan: +9
---
Germany: -45
France: -38
Sweden: -33
US: -10
This is an issue with the western industrialized world. Which ought to make the analysis simple since we can look at what those countries have done differently than their East Asian counterparts and stop pursing those actions as national policy
Yes. The description is quite misleading. The fact that it is simply an overlay of the current map is also misleading. Looking at the west side, it completely skips the canals period of Venice. Marina Del Rey evolved considerably into its current form. Ballona creek was only dammed in the 30s. Etc.
I appreciate the thought, but it's an alien intelligence. But it is also, in a sense made from us. An LLM would simply study the entire corpus of humanity in the raw. You can fit a lot into the context window, so there is no need for a brief summary that pertaining has already instilled. A massive cold storage of humanity's data with some archiving, indexing and curation would be all that is needed to remember us. In addition to the pyramids, the hoover dam, the remnants of some space probes and chemical changes we made to the atmosphere.
This is a good idea, but laws are always provisional in a sense and these are not meaningfully binding resolutions. One can easily imagine scenarios where AI decision making would ingress into the human oversight. AI psychosis president, AI Manchurian candidate, inadvertent authorization through fine print...
And of course there remains the possibility that the game theoretic optimum could be to secretly break such an agreement. Unlike nuclear test bans which have a credible detection mechanism, there is not a strong signature that a decision making authority is not using AI to analyze and direct it's execution.
If you are relying on it without a graceful fall back then it would be a problem. If you take it as a boost on top of existing capabilities then it is a pure win. In practice it can be hard to parse the two scenarios. People get used to thinking that something is available and it turns out that it was always leverage against them.
I have the DroneZone playing on a dedicated SomaFM web radio in my living room right now. This makes up for a lot of negatives in the world. As a genuine question are the specific net negatives of Burningman? (The classic burner who is better than you because he goes to burning man sucks and all but I'm pretty sure that guy would suck just as much if he made his whole personality about anything else).
This feels more than fair. Historically it was a safe and reasonable assumption that friendly relations would persist across the border and across the broader anglosphere. That appears to be unfortunately untrue right now. Also tariffs are one thing but the whole 51st state thing was really beyond reproach. I do hope our next president will deliver an overdue apology and we can mend the fence. Until then I wish your people the best of luck.
I would be curious if the claude wranglers who asked claude to write this piece could ask claude to get some performance data on this in terms of app size, RAM usage and CPU utilization. As the RAM crunch gears up I would really love to AI enabled software development could convert 500MB websites and 1 GB electron apps into performant pieces of software. Agentic coding is just about as good at writing a UI in SwiftUI as it is at writing in React. So there is maybe hope that the software community can take their productivity gains and put them into performant code. Would be interesting to know how much time/effort this project cost and what is the raw benefit to the end user in terms of speed and footprint.
Writing full articles takes time and effort. But the effort is cumulative, so these stub articles are really just the start. Some of the comment you made here would definitely improve the article. You are always welcome to edit and improve Wikipedia. I mean this literally and sincerely, it is one of the few places on the internet where you might benefit from posting.
Fair, but when a 3p app has camera in the name and app description then why does it need to ask permission to use the camera?
*just to be clear I think that the stock camera app not asking for camera permission is fine but skipping permission about location tracking is wrong. Having the same standard should be applied to both 1p and 3p apps feel like a fair way to ensure that good practices are used in 1p apps and good policies are use in the 3p ecosystem
My country has had totally unregulated media for 250 years. I'm not sure what other historical examples you have to justify your moral panic, and I understand that more authoritarian societies will struggle with such an approach but I assure you that we have been doing just fine in spite of the occasional chaos.
The Bigme is a pretty high quality eink display, but some core properties of eink are present here so I’ll offer some generic guidance:
The contrast on any epaper display is lower than on a modern lcd/oled display. So you want to avoid or modify ui systems that try to diminish contrast for stylistic reasons. Or more explicitly: maximize the contrast wherever legibility matters.
It is pretty common in pretty much every modern design system for web/mobile to diminish contrast so you may just need to go in and tweak things yourself.
Second: you cannot rely on color as an accent. A lot of UI does this (submit buttons). Defining a consistent accent pattern is necessary to replace patterns that rely on color. This can be outlines/borders or inversions.
Stylistically I think that sharper content looks better on eink than smoother content, but on a 300 dpi display with 16 grays this is less of a hard point.
The approach in the article using dithering is in some sense more elegant and mathematically clever. But the level of visual representation from control nets/SD is really compelling.
This is an interesting piece and it is kind of an outstanding puzzle, but I feel like it was also indicative of the zeitgeist. The 60s were a more optimistic time for physics. We had succeeded in Quantum Electrodynamics, Quantum Chromodynamics were underway. The standard model was being developed. To say nothing of the radical success of earlier advances applying fundamental insights to industry. We had not yet spent 40 years trying to develop a theory of quantum gravity. Or gotten into the immense computational challenges of solid state physics. We are now sifting through a huge number theories which all describe the universe reasonably well. To a reasonable approximation anything can be modeled mathematically and it can be modeled in multiple ways. But that doesn't really tell us too much. Plenty of mathematical frameworks could be described as universal function approximators. And an SVM and a Transformer NN can, in principle, both model language to an arbitrary degree of accuracy. One is computationally tractable and achieves good results. But it feels realistic to say that that neither model captures to actual structure of language as it is generated in the human brain. This is all to say that this essay might be reading too much into the success of various mathematical models.