Frontier models have seen more Mathematica and Powershell code than I ever have in their training, yet they really struggle to produce working output. They seem heavily tuned to Linux too. Despite adding skills to rectify this, they still fail to realize they're running on Windows and waste tokens. A human with this much training wouldn't have this problem. There are evidently still some pretty big holes still.
This is probably a harness problem rather than a model problem. GitHub Copilot will happily and effectively use Powershell while Claude Code struggles in my experience.
I have spent a small fortune in tokens trying to get Anthropic models to attempt proofs of things like the irrationality of the Euler-Mascheroni constant, and to make lattice reduction / SVP algorithms that rival SOTA methods like the G6K sieve, and all sorts of difficult frontier problems. It says 'honest' as a euphemism for giving up, because it knows I'm being a crank, and because the training data contains a lot of pessimism.
But I can get around this by first having it write a prompt for itself discussing the problem followed by a rousing speech about never giving in and admitting defeat, but pressing on towards results regardless of how hard matching SOTA results is.
Maybe a dumb question, but what's to stop people from communicating e2e encrypted over totally insecure channels using steganography techniques?
You don't need a special app to do this, or maybe you just need a companion app that you type your message into and it gives you the thing you just paste into whatever messaging app / social media you use. The steganography makes it hard for the operator to determine that you're "abusing" the service by not transmitting your message in the clear so they can read it.
1) Alice uses steganography to embed her public key in an otherwise innocent or mundane looking image e.g their profile picture.
2) Bob uses the public key to encrypt a short message to send her.
3) Bob embeds the encrypted message in his own mundane looking image (could generate these from a pool of images or on the fly using stable diffusion)
4) Bob sends the image to Alice.
5) Alice recovers the encrypted message and decrypts using her private key.
(Could also use the process to do key encapsulation too, instead of using the raw key pair)
Getting the criminals is not the point here, mass control is. How many of your friends do this now? The criminals might do it, but why, when they can just meet in person and talk there, without a digital recording of what they talk?
Combine the 'age verification' (show your ID when you register) with this (we can read what you type), add some AI (to profile the people), and you have all the info you'd ever want on anyone anywhere.
Any evidence what so ever for the "mass control" claim?
I do not want E2E comms broken, but these posts on hackernews always bring out such outlandish claims.
You can't be applying the "mass control" claim for the attempts at stopping the spread of CSAM? So can you point to any evidence of wide spread "mass control" with Chat Control 1.0?
Because it's control of what we send and it's done en-masse.
I was born in a communist country (with red stars, a communist party and a dictator), and even back then, the government requiring post office workers to read (and photocopy) our mail would seem absurd, but now, somehow it's ok, because "it's on the internet". And yes, CSAM could be sent via normal mail too.
Same for every repressive thing... if you went to a bar, and had to show your id, have it logged, and someone would write down when you came, who you sat with and how long you've talked, that would seem absurd even then... but now, since "it's on the internet", storing metadata is somehow ok.
Stop protecting the governments, if they wanted to stop pedos, they have the whole epstein list, and they don't care about it.
Well when it comes to ending encrypted traffic, I would assume if they can’t read your traffic you will be in violation and the police will show up at your door to kindly imprison you for a few years
If they can't distinguish traffic containing hidden encrypted messages from humdrum non-encrypted traffic then they'd have to ban the whole thing for everyone.
Steganography fundamentally requires you to be able to know where the data is, which requires you to have the original image to compare against. The only other strategy I'm aware of is setting known pixel positions to exact data, which is very easy for basic tools to spot and decode. Or to add the data to non-visual data blocks if the image format supports those, which is also quite easy to spot.
Zia Yusuf : "... criticised sections of the legislation that allow ministers to direct regulator Ofcom to modify its rules setting out how companies can comply with requirements to crack down on illegal or harmful content, saying it was “the sort of thing that I think (Chinese president) Xi Jinping himself would blush at the concept of”."
Reform and Restore will say anything to get into power, and then they will turn into ultraconservatives. See also: council tax in pretty much all sizeable councils they got to control, which they promised to keep fixed and then raised immediately after getting in. Their instincts are traditional ultra-right.
And yeah, I agree that there is no natural home in UK politics for social liberism. Hence why quipping "they keep voting for this" is just pointless - there are no realistic alternatives, what we have at the moment is the least worst and it's still pretty bad.
Where is the evidence for that? Though I grant they may get into power and suddenly change their mind, at the moment those parties are against digital dystopia, most likely because the centrist parties are most interested in using those tools to silence criticism of current immigration policy and other failings.
And the left are also being surveilled and censored, e.g pro-palestine / anti-war / anti-capitalist groups, though the strategy used there is less censorship and more often bogging those groups down with infighting about identity issues. What seems more dangerous is letting the increasingly tyrannical centrist establishment, dead set on stagnation and "managed decline", give legitimacy to censorship tools (which will be available to future extreme governments!) rather than fixing things properly.
"Commissioning the battery requires a device capable of running the BYD Be Connect App - available for the Apple App Store and Google Play Store." - No thanks. Just have a control panel on a lcd screen ffs. They also do an inverter "Fronius Primo GEN24 Plus" that requires an internet connection. Completely unnecessary.
Subscriptions make companies lazy and it degrades the product. I'm looking at you: Foundry, Adobe, Maxon, heated seats on BMWs ...
They rest on their laurels, enjoy the increased cash flow, say it allows them to work on regular updates. But this goes from being useful bug fixes, to merely shuffling the UI around, changing the fonts, introducing nonsensical features nobody asked for or can make use of, and gutting useful features for "streamlining" purposes... while longstanding bugs that actually need fixing are still unfixed.
Eventually customers become dissatisfied with the product and make up for lost features and degraded user experience with a smörgåsbord of perpetually licensed or FOSS alternatives from various competitors because they too will want to improve their cash-flow instead of being bled dry every month.
Companies that choose to offer lump-sum permanent licenses have to make a bigger effort to convince customers to upgrade, which means the product improves. Also it makes your customers more committed to your product. You should invite this kind of challenge and forgo the temptation to boost cash-flow because it keeps you on your toes. Subscription-only will seem great for a while but eventually you'll atrophy and fail.
Something similar happened when software went from being released on CDs/DVDs to regular patches and downloads. Not saying we need to go back to that era, but QAs had to work harder back then because distribution was expensive. Nowadays you can release things in an unfinished and broken state.
Man I think you're spot-on. Back in the day the biggest motivator companies had to make good products was that they would be competing in the market with their own old products.
Wow, really great insight--I've never seen it framed like this before, but it makes so much sense. Typically software becomes worse over time, and therefore uncompetitive with previous versions of the same software. The fix? Never make the previous software available, and force users to upgrade to the worse, new version.
This explains a lot about why subscriptions have become the norm.
If you have a product and a team working on it, it’s tempting to just keep adding things to the product. You can’t just sit around fixing bugs, they say. When that is precisely what customers would want. Can’t make upsells though and when you’re a business that kind of thing really matters… nobody from within can see that the ship is so heavy it can’t stay afloat anymore. Until it’s sunk.
Subscriptions can create an illusion of a deal, because in principle, you’re ostensibly able to benefit more for a fixed price. But are you?
Netflix is a good example. You can watch as much as you want for a flat rate, but how many people watch enough to justify the monthly fee? (Putting aside the question of whether watching so much is actually a benefit in the first place.) Companies recognize the distinction between potential use and actual use, and so in practice, many are paying more for less and subsidizing the outliers that consume more. When actual use exceeds predicted use, the company will raise the price of subscription.
Subscriptions make sense for situations where there are regular maintenance costs or where the benefits are received at a steady and proportional rate.
>I don't understand why most people don't just torrent? [...] It's actually more convenient than streaming services.
I think your technical sophistication means you're somewhat out-of-touch with what "most" people do.
Most normal people watch Netflix/HBO/etc on smartphones/tablets, or stream devices like Amazon Fire Stick, Google Chromecast puck, Apple TV cube, or the "smart tv app" built-in with their Samsung or LG tv. All of those "mainstream devices used by most" don't make it easy to access torrenting sites or files. Sure, one could hypothetically sideload a torrenting app on a Google Chromecast but now you're beyond the demographic of "most people" because you have extra complexity of also adding some USB storage to save the torrent or point to a local network share.
The type of situations that makes "torrenting more convenient" are people watching everything on a laptop or have a dedicated HTPC media server hooked up to their tv.
I'm technically savvy and it was not easy to sideload Kodi player onto Amazon Fire Stick to legitimately play DVD ISOs. It required a lot of google searches to finally figure it out. (E.g. after realizing VLC app for Fire Stick doesn't work, and then finally stumbling across a "developer setting", and then getting the SMB network path correct, and so on...) Thinking that most people could just torrent is being unrealistic.
Torrents nowadays have <1% the number of people they had a decade ago. It didn't used to be considered technically sophisticated, just a new version of file sharing that everyone used two decades ago.
I just stream files from my Macbook via AirPlay to my tv. works pretty seamlessly. No media server setups. They don't even have to be on the same wifi network. I think most TVs support several streaming protocols nowadays? I got the cheapest Samsung smart TV.
You are probably right I'm out of touch with technology, but I also think that many people do much more advanced technical stuff like using VPNs - became pretty mainstream.
>I just stream files from my Macbook via AirPlay to my tv.
Ok, explaining your situation with a laptop clarifies where you're coming from. (Which my prior reply anticipated and covered in my 3nd paragraph about torrents being easy for people using laptops.)
In any case, most normal people do not use AirPlay from their laptop, nor cast from a Chromebook, nor cast/mirror a Windows to their tv to play Netflix/HBO/Disney. Instead they just use the mainstream hardware streaming devices or the built-in tv app. Torrents would be much less convenient for the way most non-techie people watch tv (Roku/FireStick/SamsungTVapp/etc). Netflix has stated many times that the majority of their customers' watch time comes from smartphones/tablets/tv and not desktop/laptop web browsers.
Because you’d be denying cast and crew of their royalties?
If people pirated on a mass scale, the losses would add up. Whatever you may think about streaming platforms, don’t punish the people just trying to make a living.
The ayes have it. Motion passed, now let's discuss the subscription tiers. How many stickers should we include with the premium 'founders' subscription tier?
The video game industry is plagued by this problem. With live-service games becoming commonplace, there has also been a recent trend of games being released in an incomplete state. The shocking part is that multi-million dollar "AAA" game studios are engaging in this behavior. There's also a strong "own nothing" component to the issue.
I'm sorry to write that all those were problems before subscriptions too.
Managers trying to hit targets and ratchet performance metrics, product managers trying to clear queues, the exponential growth of complexity from size (ala mythical man month)... Fundamentally it's a misalignment of incentives and and as yet unsolved problem of scaling social knowing.
I continue to maintain that the problem here is not greed but laziness. Make money with less effort or make even more money over a longer term with more effort. Building for the future requires effort and investment and has the potential to make more money than focusing purely on the current quarter.
> Companies that choose to offer lump-sum permanent licenses have to make a bigger effort to convince customers to upgrade, which means the product improves.
Why can't they expand their customer base instead? With a great product, you sell millions of copies, pay everybody's salaries and pay investors.
Something like for example Affinity should in a rational market eat at least half of Adobes customer base with their current offerings. So maybe it's a problem of marketing?
I have never in my life seen an advertisement for any app with a pay-once offer, even though I have bought most of my apps as pay-once. And they're always several levels higher in quality than other offerings.
Affinity did have a growing userbase, partly because you could just buy the app for $50 and use it perpetually.
I assume with Canva buying them out and making it "free", Affinity will fade away and eventually just be folded into subscription-only, cloud-only "products".
This is exactly why I bought Affinity. That, and the fact that the old version did not require an "account" and "activation". You send them money, they give you an unlock key, and that's the end of your relationship with them.
>Something like for example Affinity should in a rational market eat at least half of Adobes customer base with their current offerings. So maybe it's a problem of marketing?
I suspect Adobe's customers look at their tools in a different way to the typical HN poster. They don't want too many new features because that disrupts their existing workflow. They would prefer to get annoying bugs fixed over something that causes them to relearn the software. They aren't even that worried about subscriptions because the software is a means to an income.
Graphic designer here. Adobe’s products are the golden standard in the industry and are often the only products used by entire graphic teams, being the odd one out using Affinity’s or (heaven forbid) Corel’s tools makes collaboration difficult (if not entirely impossible) due to how unfriendly and unpredictable portability between apps can be, and unpredictable is the last thing you want in an environment where precision is absolutely essencial.
It’s sad, really. Adobe’s apps are powerful, yes, but they’re stuck in time. They perform horrendously and have terrible UI/X that aren’t even standardised between apps (even something as simple as the icon for their own assets library cloud is not the same between apps).
How I wish someone would come along and eat Adobe’s lunch to force it to pull itself together for once, but alas, that’s unlikely at best.
Considering how cheap Affinity is (and now it's free). Why don't studios just get both apps? There's almost no learning curve in switching between them.
I foresee a lot of smaller actors switching to Affinity in the future, just like tons of people went to Canvas.
Canva is a nice tool for light work only, but no studio, agency or design department worth their salt relies on Canva for more serious work—and rightfully so, Canva is not a player in the same space as Adobe and Affinity, for example.
Affinity is nice, and I really do wish they would corner Adobe a bit more so they could get their act together. But the fact is Adobe has encroached upon every corner of this industry, pretty much all of their apps are the standard tools for any given design job (from both a software and hardware perspective). It would take a massive amount of effort to convince designers used to Adobe’s tools to switch to an unproven new tool, and it would take an even bigger effort to actually make these new tools become the standard.
I would love to give Adobe the boot, but the truth is that no one has been able to meet Adobe’s apps head on as of yet.
But that's what I mean with studios installing both. You're not forced to choose either Adobe or Affinity. Since you can install different software on a machine, it would be in the interest of every studio to work with both apps. No need to give Adobe the boot.
Now that Affinity is free, you will have many more clients and small studios sending you material in their formats.
As for Canvas, I expect every pro studio to be able to work with that format, since legions of clients will make their stuff in Canvas and send it to you and vice versa.
It's like Visa and MasterCard. You don't have to pick only one, and if you're smart you will accept both.
The thing is, studios might “offer” it, but designers aren’t likely to go for the new kid on the block with Affinity’s option when Photoshop is sitting right there; and Canva doesn’t play any part in this little software war, really. Canva is not aimed at pro work, professionals do use it if needed but it is not a tool of choice for any serious design work, it’s really just a stop-gap tool to enable non-designers to do some decent design work, but nothing more than that.
The truth is Adobe has a massive upper hand here in both market share and public perception. Affinity’s offerings might be 100x better, but they will never be “THE Photoshop”—that’s how deep Adobe’s claws are buried within this industry.
Updates were fewer and far between back then, but they were bigger updates with more features. Now, many subscription services seem content to milk fewer feature upgrades staggered over years.
This is how I've felt about Apple, and it doesn't just include a model with a jack, but a worsening of the iOS system, models that are no for most users of previous models. Apple has run out of ideas, and maybe that's the entire smartphone category.
People will still buy smart phones, but I think their will be less enthusiasm for the latest and greatest model.
reply