I remember the old "Windows ain't done until Lotus won't run!" schtick. I wouldn't surprise me at all if the "faulty" code originated with the Edge team to help encourage adoption of their version of the Chromium user tracking software over Google's version, and through doing so nudge a few more people into using Bing and Copilot instead of Google Search and Gemini for even more tracking.

Think of an iceskater pirouetting on the rink with their arms stretched out. If they pull their arms in, they start to spin faster, thereby conserving angular momentum, but they don't gain any additional mass from doing so. More massive blackholes with a faster spin have just consumed more matter to get to that state, but it hasn't fundamentally changed the overall mass of the galaxy they are in.

It was, and is. But there's another factor to this. Look at the figures for Google (which is also doing AI) vs. those for Anthropic.

Google is still mostly an ad-provision driven business, and it is fuelling that business on a relatively low volume of content scraping to actual eyeballs visiting the site and viewing the ads it providing. The cost of providing the data to Google's craweler is, one would hope, largely recovered in the ad revenue generated and other financial benefits from those actually visiting the site (subscriptions, purchases, tips, whatever). In otherwords, they have a viable business model and will survive to provide content to those that want it another day.

Now look at the other extreme in the provided figures with Anthropic. The site gets *nothing* back from Anthropic, quite the opposite in fact since Anthropic is harvesting their content and using it to train their AI engine, any issues around copyright be damned, but is paying for the hosting costs and the quite insane levels of crawling Anthropic is doing. I mean, seriously, WTF is Anthropic doing to need that level of crawling? That's just out of the park and looks very much like a script that is running completely without any throttles even though it's just pulling the exact same content down over and over and over again, which does not give me much confidence in the quality or competence of their developers. For the website though, it's going to completely skew their hosting costs - they had a business model with Google's ratio of 2:1, maybe even at 18:1, but at 60,000:1? That's an increase in an overhead by a factor of 30,000, which is going to break a lot of business models, and potentially they could also be paying Anthropic to help generate some of their content, thereby making their problem even worse... That very likely means more paywalls, more deliberate attempts to poison AI-scraped data, and - of course - lawsuits.

Unless the AI bubble pops soon, I fear that there's only one way this is going to end, and that is with the Internet completely overwhelmed with AI-generated slop, many of the smaller content producers and hosters that can't adapt quickly enough having folded, and any remaining residual content with value that people might actually want to see locked away behind paywalls and other login protections. Maybe it's time for written media to have its "vinyl" moment; Anthropic et al's crawlers are not going to do too well with a printed book...

I'm not following your logic there. AFAICT from TFS, the patent in Canada has *lapsed* and no longer exists so anyone can manufacture this stuff (presumably subject to suitable medical grade quality controls) and sell it in Canada at whatever price they choose, including the government or their appointed sub-contractor if they wanted to. That means *every* version of semaglutide in Canada is going to be at generic prices real soon now anyway, because the market will be saturated with them.

If the government bought out the patent sure, they could then license it for a fee, but what would be the point if the aim is to make the drug available cheaply to Canadians and, I guess, USians that cross the border to fill their prescriptions rather than pay US prices that are still under patent cover? That would just add license costs on top of manufacturing and distribution costs and drive the overall prices up, unless you mean by adding a price cap as part of the license terms? The logical thing to do would be to let the market set the prices and if that's at a realistic level, then great, but if not - because Martin Shkreli isn't the only greedy fsck without a conscience in pharma - then Plan B would be to contract someone to manufacture them at a given price for the use of the Canadian health service.

TFS doesn't quite add up to what the headline implies. AFAICT, the actual sequence of events is that Ms. Horan bought and paid for some toilet rolls on May 8th, after which *human error* at the store resulted in her being added to the Facewatch programme. Because she had been added to the Facewatch DB, the programme then did exactly as it was supposed to and flagged her entry into the stores on May 24th & June 4th, prompting the store staff to react pretty much as you'd expect under the circumstances and ask her to leave, albeit perhaps without sufficient discretion.

There's really only one screw up here, and that was by the staff at the May 8th store who added her to the Facewatch DB, everyone and everything else seems to have done as they/it should have done under the circumstances. Still, on the "lessons learnt" front, users of systems like this *really* need to allow for the possibility of human error in the submission or a mistaken ID by the system (not that this seems to have happened here) when challenging someone like this, and have a clear cut audit trail and process of appeal. If Home Bargains had been able to say, right off the bat, that it was down to a presumed theft of toilet rolls on May 8th and undertake an on-the spot review on May 24th, this could easily have been avoided.

Because it's good PR, for once, as opposed to the expected "Oh, sorry, but you're out of warranty and therefore SoL"? It's basically admitting that yes, they (or one of their component suppliers) screwed up, but that they are going to make good and give you either a credit voucher to buy something else from them or an upgraded model of the product. That some of the these are 10 years old and likely long since lost or discarded means that this is going to cost a lot less than the total ~$30m value of the banks being recalled, which could be a hell of a lot less if they were to get sued, especially in the US, gives them an out on any future issues ("they were recalled, why were you still using it?"), and is probably covered by insurance anyway.

Exactly, this is coming across as more of a sweeter to placate people's concerns (by essentially buying enough off that the remainder are very much a vocal minority) in order to get things done. No one likes things like HT pylons in the rural landscape, and rightly so, but the UK's nimbys have an awful lot of power to ties things up in legal knots and prevent progress that really must happen one way or another.

The devil's in the details, but my gut feeling is that much of the distribution infrastructure for this need not be HT, other than for the short haul near the generation facility to include a grid interconnect. If the generated power is only going locally, and never needs to really touch the national HT grid, then you may be able to get away with mostly 33kV and below "last mile" style pylons and cabling, which is much less visually intrusive and can indeed be more readily buried. This is also presumably the model for the containerised reactors; small, mostly self-contained, networks of both power generation and consumption.

build more power lines,

That's the crux of the problem. The UK's grid was built when large carbon-fueled power stations were king, and so the distribution network is essentially a hub-and-spoke design radiating out from those to the end users, with the load carrying capacity getting smaller and smaller as you go. That means by the time you get to the rural fringes, like Scotland and Cornwall/Devon, which is where a lot of the optimal offshore wind sites are, the grid simply does not have the capacity to take large amounts of power the other way. Baseload capacity has to be pushed around the core of the grid, which - since we're talking overhead HT lines - are generally built along mostly flat routes between the carbon-fueled power stations, and only pass conveniently close to on-shore wind farms or off-shore wind landing stations by coincidence.

You know what else the UK has a lot of besides green energy it can't fully utilise? Nimbyism. As you'd probably expect, Nimbys really don't like the idea of new HT lines running through their local countryside (or nuclear plants, to touch on your other point), so the other goal of this regionalisation is to encourage them to accept the tradeoff in return for (relatively) cheaper local market prices in areas where there is a lot of existing or potential green (or nuclear) energy capacity. Supplementing that with local storage systems - batteries, pumped water, molten salt, or whatever - should hopefully be something that comes along in lockstep, but so far that doesn't seem to be integral to the plan, but is rather being left up to the generating companies to decide on.

That kinda depends on the level of taxation applied to the corporates and those who chose to supplement their UBI by doing some additional work for extra income, which would therefore be taxable. If you have enough revenue from taxation to cover everything else with a surplus, then UBI can be raised above "subsistence level" (which the minimum it would need to be if it's supplanting most current forms of welfare) to something a bit more comfortable. OP is probably right on the "not everyone" front though; if it happens, UBI is going to replace some current unemployment benefits outright, and there are plenty of examples of people who don't manage that income wisely, so we shouldn't expect that to change just because it is - from their perspective - just given a different name?

However, on the face of it, and given UBI experiments have shown recipients still tend to seek out and undertake some work, that would still tend to indicate most of the taxes necesary to fund it are going to have to come from corporates and investment income. In otherwords, reducing the incomes (in the form of bonuses and dividends) of exactly the kind of people who, as TFS demonstrates, are the least likely to want to pay those taxes just so others don't have to work, even thought they've probably never actually done a hard day's work in their life either. So, while I think UBI could absolutely work in principle, I suspect it's going to be a real mixed bag as to how well, and how comfortable the lives on it without supplementary income is likely to be. Where capitalism is king, like the US, UBI would never, ever, get above a level that doesn't even qualify as "subsistence", but I think it would definitely have a realistic chance in places like Scandinavia where they've long since accepted high taxes are a necessity for a higher overall standard of life, and - strange as it may seem - the Middle East where personal taxes tend to be near zero anyway.

This bit amazes me:

a consultant rushed to warn clients to be "extra careful" sharing sensitive data "with ChatGPT or through OpenAI's API for now," warning, "your outputs could eventually be read by others, even if you opted out of training data sharing or used 'temporary chat'!"

I mean, seriously? This is one of a whole bunch of companies that have been blatently hoovering up any data they can get their hands on without any regard to copyright, constraints placed via things like robots.txt, or thought to the hosting costs that can be incurred by continual spidering of vast amounts of website data, and you *honestly* thought you could trust them with the data you *chose* to provide them with or that it might not backfire like this?

Zuckerberg was right all along; "Dumb Fucks" indeed.

...and then there's this.

"This" is a cast iron example of why everyone involved in AI - the content producers, AI companies, VCs backing them, policitians, and users - need to deal with the elephant in the room; copyright law was not designed for the digital age, and certainly wasn't designed for the wholesale ingestation and regurgitation of AI engines. That the media companies, usually the first to cry "foul" and demand outrageous amounts of damages because copyright, are themselves playing fast and loose with other's content while complaining about their own being used as training data more than proves the point it's way past its sell by date.

While amended since, the Berne Convention dates from 1886. AI isn't a crisis for copyright; it's an opportunity to give it a thorough overall, make it fairer for all given it's now so easy to content shift and share data, and generally fit for purpose and fair for the 21st century and beyond. Fail to do so, and it's just a matter of time before the legal fallout (and damages) under the current system are going to give the lawyers on the winning sides of the inevitable disputes a whole new fleet of superyachts.

TFA is vague as hell about what they actually did here since there are any number of ways of interpreting "Take some legacy code, like COBOL or Fortran, and covert it into a human-readable specification". Given what Morgan Stanley does, I'd assume when they say "specification" that they actually mean it, so I'm hoping it's not just converting legacy code on a line-by-line basis and actually producing a usable specification for entire functions that defines the expected input and outputs and leaves it up to a human to figure out how to make that happen efficiently in the desired target language(s).

Either way, leaving it up to a human to interpret that and write the more modern code based function (perhaps also with the aid of AI) seems like a much more practical use of current AI engines than trusting the AI to do the whole thing and then spending even more time than it would have taken a human to do the whole thing from scratch figuring out where the AI got something wrong. COBOL is one thing because it's pretty easy to parse given a little time, but my hunch is the big (and probably overstated) time saving here is more likely to come at least in part from not having younger coders who have never come across it before having to get to grips with something more cryptic like Fortran, or even some early "heavily optimised with obscure coding tricks" C for that matter.

Somewhat old now (it's from 2008), but his might be one of the more honest exposes of life the DPRK, albeit mostly from a moving train. It's a lengthy travelogue by a big fan of rail travel who travels to Pyongyang from Vienna via Russia on a variety of trains over a number of weeks using an "unofficial" tourist crossing, so most of the photos inside the DPRK are not going to be the typical staged locations and setups official tours might take, just someone with no particular axe to grind I can see who got a bunch photos and interactions taken while unescorted between the border and Pyongyang.

It's a *very* long article across many pages and a fascinating read, but if you want to skip to the crossing from Russia into the DPRK start here.. My take is that it confirms a lot you might suspect or have been told about the DPRK, or any other regime that is highly authoritarian and regimented, but also dispels the notion that the country is some kind of dreary, charmless, technological throwback to the dark ages unless it serves the immediate purposes of the military and of the Kim family, but YMMV.

That would be 2028. Eight is an auspicious number to the Chinese, and it'll take advantage of the US presidential elections and the inevitable chaos around Trump refusing to leave office plus, of course, the aftermath from the shit that went down in 2026.

I'd say more, but I've got some sticks that need sharpening.

to climb on one of the death tubes

Well, there's part of your problem. The rest of us get in the death tube.