I got a notification from my music distributor that the June sales report for my tiny record label Supple 9 had arrived. It’s never very much, but I always log in to have a look and see (hope springs eternal..)

-18.57 €

Yes, that’s a minus symbol in front of my sales. Not only did I not earn anything, but in fact they took money out of my account for having music on streaming platforms.

I cannot understand how this is even possible.

I dug into the full report and saw that Spotify charged a 10€ “fraud fee” to two of my tracks (Ebauche – Rozwijać and Dronal – A Body of Water).

June total sales (all platforms): +1.42€
Fraud fees: -20€

And for Spotify alone I had 330 streams earning a whopping 0.96€, which means that having my music on Spotify cost me 19.04€ in June.

Spotify had recently announced that they’d be cracking down on streaming fraud, an announcment I ignored because I don’t do it. I checked my streams for the year thus far and on Spotify up to June I’ve been paid for 6032 streams.

If I was committing streaming fraud, then I’d be doing it very very wrong. I hope it’s clear that I obviously do not use any “stream boosting” services. But I had a look back through my Spotify For Artists stats and I did discover that those two tracks did have a sudden jump in streams on one day each in June (4th and 5th).

So yes, this looks weird.

I remember when it happened (I quickly check my Spotify stats each morning – remember, hope springs eternal…). I had no clue why it had happened, and I was super excited to think that “oh, maybe I’ve been added to a popular playlist”. Sadly the next day my stats returned to normal. I have a vague recollection of digging around Spotify for Artists and finding that all the streams had come from some random playlist, but that indeed, my tracks were no longer included. I didn’t spare it any more thought. And if I look back through the stats now, I can’t see what playlist it was.

So I understand that it’s possible that someone created a playlist, stuffed it with a bunch of random tracks, and some of their own tracks – and then fraudulently streamed that playlist a whole lot, and then deleted it.

What I don’t understand is this:
surely if there was a fraud fee, even though I personally didn’t do anything wrong, how come the fraud fee is far more than those tracks have earned, not just on those random spike days, but in fact far more than those songs have earned in an entire year. For example “A Body of Water” has now earned -9.66€ in 2024..

If there was fraud, shouldn’t the fraud fee just cancel out the fraudulent earnings the track had made?

I’m guessing of course that Spotify will say that they have some minimum fee (perhaps €10?) for handling cases of suspected fraudulent streaming. However surely this then creates a situation where malicious actors could target musicians with fraudulent plays and cost those musicians more than they’re earning from the streaming service.

I decided to dig around further and downloaded my full sales report for the year so far.

My gross income from streams on Spotify = 19.49€
My losses from “Fraud Fees” on Spotify = 20.00€
NET INCOME = -0.51€

So yes, it’s now costing me to have my music on Spotify.
Or let’s put it another way – Spotify are EARNING money FROM ME?!

(I don’t know for sure about this, perhaps Spotify will say that the fraud fees go back into the total pool for artist payouts, in which case that means that now Taylor Swift or someone else is effectively earning money from me)

I’m absolutely baffled. It’s depressing enough trying to at least have my music break even with the costs of distribution etc, but to have tracks COSTING me money by having them on distribution platforms is mind-boggling.

Combine this with the fact that Spotify arbitrarily decided that they wouldn’t bother paying tracks that get less than 1000 streams in the past 12 months: it’s enough to make one just throw in the towel completely.

I have sent iMusician a query about this because the whole thing is utterly ridiculous, but I’d be intrigued to hear from any other musicians that are facing a similar situation.

I recently started watching a series on Curiosity Stream titled “Zenith: Advances in Space Exploration”. Half way through watching it on Curiosity Stream, it disappeared from the service (I guess their rights agreement expired).

Anyway I was delighted to find a full playlist with all the episodes on YouTube.

I think it’s probably a couple of years old as there’s a bunch of “happening soon in 2023” moments, but it doesn’t lessen the relevance or fascinating information contained therein.

If you’re looking for a science documentary that will cover a wide range of topics on space exploration, earth observation, our solar system, and beyond, then I think you’ll enjoy this.

I will admit that I left Firefox for a while about 10 years ago when it was lagging behind Chrome in features and functionality, but I came back to it nearly as my exclusive browser about 4 years ago when I started to disconnect myself from Google’s clutches.

However there was one area which was restricting me from using Firefox 100% of the time, and quite an important area given I live in Germany and am quite far from fluent in German

Automated and inline web page translation is just so important for me to go about day-to-day life without a lot of extra confusion and hassle. For a long time Chrome (and Edge, but that’s even worse than Chrome) were the only browsers where you could get reliable native inline translation.

Then Vivaldi announced they were bringing inline translation with privacy features (anonymised translations) which was great news and led me to have Vivaldi primarily for browsing German websites.

However Firefox have gone one further and don’t send the webpage text anywhere, so it’s completely private.

Unlike cloud-based alternatives, translation is done locally in Firefox, so that the text being translated does not leave your machine.

You can read the details on the project here: https://blog.mozilla.org/en/mozilla/local-translation-add-on-project-bergamot/

A huge congratulations to the Firefox team for bringing this excellent feature to life.

Why not test out Firefox today. Seriously, it’s time to take back some control from Google and Microsoft and have more of a say in how your data and privacy are handled.

Here’s a situation that happens to me on a surprisingly regular basis. I’m online and someone posts a link to an article which sounds very interesting, right up my street. I click on the link and some website loads up and I see the title of the article and some leading image. I start reading and as I make my way through the first paragraph the text slowly fades out from one line to the next until it’s completely invisible. Just below this is a box “Subscribe to keep on reading”.

Yes, I’m sure you’re familiar with this too. It’s the paywall. If you want to read this you have to pay.

Of course, in principle there’s absolutely nothing wrong with this. I totally understand. Investigative journalism takes time and money, people need to earn a living, and plastering adverts everywhere is not always a viable solution (and I personally believe it to be bad in general).

However, here’s the issue. At least for me, I’m self-employed and I have to watch my expenses, especially recurring ones. Sure if it’s a website I visit regularly then I might consider it. I subscribe to The Local as it’s a great source of news for English speaking migrants in Germany. But all these other sites I fall upon every once in a while? All that happens is that I close the tab and forget about it. I’d be happy to pay a once off fee to read the article, but there’s no way I’m committing to spending $5 a month or a week or whatever, when all I want is to read this single article now.

Enter micropayments

I’m convinced that the solution to this is micropayments.

Imagine if you landed on that article and it said you can subscribe to read all articles for X per month, or pay 25c to read just this article now. Well hey, 25c sure, I don’t mind paying that, the article sounded very interesting.

The thing is, for this to work it needs to be baked in, it needs to be universal. It can’t be a piecemeal mess of different approaches from one site to the next. And what’s more – no one wants to be building a plethora of different systems to take digital payments of 25c. Transaction fees probably immediately kill that idea. It’s just not worth it. And it’ll be a huge piece of friction if I go to Site X and I have to create an account, then enter my credit card details, and finally pay the 25c to read that article, and then later on I click on a link to Site Y and I have to do the whole thing all over again.

It has to be a protocol

I’m convinced that the only way micropayments can work is if it’s built out as a full protocol. There are lots of different protocols on the web today. A simple one we all know is email. It’s still around and in use because it’s a protocol. It doesn’t matter who your provider is, it’s all interoperable. Email wouldn’t have worked if you had to have a gmail account to email people on gmail and you needed a hotmail account to email people on there (in fact that’s sort of the situation we have now with instant messengers, everyone has three or four different IM apps installed because some friends are on one platform but not another – and I think we can agree it’s a total shit show).

Anyway, I digress, lets go back to our use case.

You land on Site X, there’s a recognisable button for a micropayment transaction to access the article. You already have an account with your preferred micropayments provider and you’ve connected your account to your web browser. You click the button. A familiar modal dialogue is presented by your browser asking you to confirm payment of 25c to Site X, and saying that you have $X credit in your account. You enter your PIN or scan your fingerprint. Boom, you’re reading the article.

I think we can all agree that this would be pretty great no?

Is it a trivial task to get something like this implemented? Of course not. It’s something that would require a lot of cooperation and planning. And of course much as I hate to admit it, I’m proposing yet another situation where companies can insert themselves in between customers and businesses and leech a percentage without actually ‘making’ anything. But I think in this case it makes a lot of sense.

Of course there’s a lot more to it than the simple paragraph I’ve described above, and it’s not like micropayments haven’t been tried before, but the thing is that no one, from what I can tell, has taken the approach of developing it as an open protocol. I think that’s the key to its success. You avoid the problems with expensive credit card transaction fees by people topping up their account with one of many providers in larger amounts. Providers can be local and integrate into local payment systems, so in places where direct debit is more common, the provider will connect in with familiar direct debit systems, in other countries where people pay by SMS, the provider will offer that option. The link between payment provider and vendor is handled by the browser or the app developer and uses a universally recognised approach defined by the protocol. The only thing I’m not 100% sure on is how the vendor collects their credit from the payment provider – there would need to be some connection between payment providers that would free them from any transaction fees. I’m sure there must be a way.

It’ll simply never work if it’s not a protocol. I’ll sign up for Zibble to make micropayments for Bloomberg and The Washington Post, but then I’ll land on The Atlantic and need a Mizzle account to make micropayments to them, and then when I hit Der Tagesspiegel I find out I need a Dezbler account.

Anyway. I think you get my drift. I am throwing this out into the ether in the hopes that it inspires someone with more technical knowledge and connections than me to make something magically happen.

What do you think? Is it conceivable?

I also figured that a higher sampling rate just enables the tracking of frequencies outside the bounds of human hearing, and that not only would the increased frequency range be inaudible, but that those frequencies would then be removed again when the finished work gets downsampled.

However, I’d never actually experimented to compare the end result of a 96kHz mixdown vs a standard 44kHz mixdown.

So when I went to export stems for mixing the first Exan album I decided to do a test. The core work on the album had been done almost entirely in Reason, with a small amount of hardware jamming recorded on top. This meant that when I was doing the exports I had the option to export at different sample rates. Even though I had been working at 44kHz whilst producing the material, Reason would happily let me export at everything up to 192kHz.

Obviously any stems I’d recorded into Reason from my hardware at 44kHz would just be re-sampled at the higher rate, but all the Reason instrument/effecting processing would take advantage of the higher rates.

So I took a few different tracks and exported all the stems at both 44kHz and 96kHz, and created projects at those sample rates in Studio One. It’s understandably difficult to do A/B listening tests between two projects, as there’s a pause when switching projects, so my best option was to simply use Studio One to sum the stems so I could compare the difference¹.

Once I had rendered all the mixdowns I downsampled a copy of the 96kHz tracks to 44kHz so I could import the all-the-way-44kHz track and the 96-down-to44kHz track into the same project and do immediate A/B testing.

The Result

To say I was surprised by the result is probably an understatement. It was extremely clear to my ears that the track summed at 96kHz and then downsampled to 44kHz was acoustically superior to the one summed at 44kHz.

Most specifically the bass frequencies just sounded better – this was the real surprise. I was expecting that there’d be better top-end sheen or something, but where I really noticed it was in the low end.

Yes, the difference was relatively subtle, but definitely audible. And when you’re striving to get the best possible sound, every small benefit accumulates.

So, my assumption that high sample rate mixing was pointless as later downsampling would just nullify the benefits was quite wrong.

I proceeded to export all my stems at 96kHz and did my first 96kHz album mixing.

The Catch

There were a couple of downsides to this though.

Firstly, working at the higher sample rate of course means just over double the number of calculations that the computer needs to make when processing the audio through software plugins. As a result I found my available DSP getting chewed through far faster, resulting in frequent track-rendering to free up DSP (at least Studio One make this fairly simple, with an option to un-render and tweak). These frequent renderings definitely interrupt the mixing flow.

Secondly, about half the audio inputs/outputs in my setup come from two RME ADI-8s which run on an optical connection to my RME Fireface UFXII. When you switch your sample rate to 96kHz the number of channels provided on the ADI-8 is halved. So all my outboard gear routing was immediately in a shambles, and my Elektron gear, which I use exclusively through Overbridge, won’t record at 96kHz.

So it effectively meant that using any hardware during the mixing session was not really an option. Yes I could have done it either by rendering a temporary 44kHz sum or stems into a new project and doing additional overdubs/effects there, and then upsampling them back into the 96kHz mixing project; or by rewiring things when needed. But that sort of approach is very disruptive to a mixing flow and the main goal of my hybrid setup here was to make switching between hardware and software as fluid and immediate as possible.

The End Result

Unfortunately I don’t think I have my example renders for you to listen to and compare, but of course please feel free to check out my first Exan album, Centic Primer, released on the excellent Errorgrid Records label in Los Angeles.

I would suggest you do some objective listening tests with your own material. See if you can hear the difference as clearly as I could. Then you have to weigh up whether mixing at 96kHz will cause more downsides than the benefits.

For now, I’m back to mixing at 44kHz – I felt the disruption to the crucial mixing flow caused by dealing with 96kHz outweighed the acoustic benefits at the end.

Perhaps if I at some point switch my audio interface setup to MADI, and I get a more powerful computer, I’d reconsider and would export stems for mixing at 96kHz again – although I’m fairly sure we won’t see Overbridge working at 96kHz any time in the near future.

I guess I could consider at least working at 48kHz all the time, although I suspect the difference will be much more difficult to hear.

1. Obviously in hindsight I could have just rendered a full mixdown at the different sample rates from Reason, but as I was going to be mixing in Studio One anyway, I wanted to work through there for my audio summing. ︎

Update: VanMoof have declared Bankruptcy

Download the Bikey app from Cowboy to save your bike’s encryption keys. Other than that we have to wait and see if someone will step in and offer a solution to keep these bikes on the road.

Four days ago TechCrunch reported that Dutch e-bike manufacturer VanMoof was having financial troubles. It sounded like they were trying to raise additional funds to stay liquid. But then the day before yesterday, the Verge followed this with an altogether more ominous report, stating that VanMoof:

has turned to the Dutch courts for legal protection in order to give the company time to pay its bills. The company is exploring all possible routes out of its debt, including a possible sale, according to a source

That is worrying enough but it’s immediately followed by this:

The company is also temporarily closing its brand stores. Amid rumors of trouble, angry customers descended on VanMoof’s flagship Amsterdam store and service center (and former global HQ) on Wednesday to claim their bikes that had been brought in for service weeks ago. The closures are meant to safeguard VanMoof employees.

That sounds downright precipitous.

Why does this matter? Well, two years ago my wife and I decided to take the plunge and purchase e-bikes. Somewhat ironically, considering The Verge’s reporting above, my main reason for even knowing about VanMoof was The Verge’s extremely glowing review of the S3/X3 series. We took a test ride and were immediately impressed.

With the 20/20 vision of hindsight, I now know I should have done a lot more research into this. I guess I was so used to the idea that you buy a bike and it’ll probably keep going for pretty much as long as you like (or until it gets stolen). E-bikes are a different story though – there’s a whole new slew of terminology to familiarise yourself with, different approaches with motors and gears, and of course the batteries – and their potential for limited lifespan.

Our first little surprise

A few months ago we cycled together to the the local art shop and locked our bikes using the integrated kick-lock. When we came out, Ada went to turn on her bike and nothing happened. Totally dead. The app wouldn’t connect to the bike either. Nothing.

So, here’s the first issue: the bike is entirely dependent on its electronics. If they fail you can’t unlock your bike, you can’t change gears. And of course the built-in alarm system no longer functions either.

So whilst waiting to connect to their online chat, we decided our best option was to carry the bike home, or at least, carry it to the tram and bring it home that way. Eventually we got it home and by then finally had a support rep in the online chat. They got me to try a few things with hardware resets and plugging it in which at least allowed us to unlock the bike, but with a plethora of error messages coming up on the screen there was no option but to have it repaired.

Here of course is another obvious issue: you can only have the bike repaired at their brand stores or registered workshops. Now I probably wouldn’t have bought this bike but for the fact that their brand store is just down the road, so I thought it’d be fine; but still, we had to wait a few weeks for a repair appointment. If you were dependent on your bike for a daily commute this would leave you completely stuck.

Anyway, we got the bike repaired and when I went to collect it I asked “if we were outside of warranty, what would this have cost us?”

Here’s another huge issue for me: the replacement of the onboard computer and battery would have cost us €900 – not far off half the price we paid for the bikes when they were new. I had a huge sinking feeling in the pit of my stomach when I heard this – what had I bought into?!

I actually considered perhaps selling the bikes straight away whilst we were only half way through the warranty period, they would probably have fairly good resale value. Ah hindsight – if only I’d taken that option then! Anyway, we kept them – other e-bike options were substantially more expensive than when we bought ours (just before all the supply chain issues and general inflation of everything); and whilst I did go and test ride a normal bike, I quickly realised how much I would miss the boost of the electric motor when pedalling into a heavy headwind.

And the frustrating thing about discovering all these downsides is that these bikes are a real joy to ride – and for the most part have been quite trouble free.

But if something does go wrong, then you’re a bit limited in your options, or should I say option. And that option is entirely dependent on the company you purchased the bike from still existing.

The takeaway

The key thing is – having gone through this experience, I would strongly recommend against buying an e-bike that doesn’t use “off-the-shelf” components. This isn’t necessarily an easy thing to do and it can be a bit hard to find out solid information about it all.

You definitely want a bike that isn’t so dependent on its electronics. In the case of VanMoof there’s an electronic automatic gear-shifter. Sure it’s nice when it’s working, but manually shifting gears really isn’t a problem, and it’s just one more thing that can go wrong (and renders your bike effectively unusable if the electronics fail).

Another thing to consider is that owning an e-bike will probably have a substantially higher maintenance cost over its lifespan than a regular bike. The probability is that at some point the battery will fail (or it will lose range so much that you decide to replace it). In the current climate this will probably cost you €400, and there’s a good chance you’ll have to stump up for this at some point in the first 5 years.

What Next?

We can only hope that VanMoof manage to muddle through this difficulty. Failing that, the hope is that rather than just going bankrupt and disappearing, potentially leaving us with no recourse when the bike inevitably breaks down – we hope that some other company steps into the breach, acquires the assets, and agrees to honour existing customers’ maintenance/warranty agreements, as well as provide repair/spare part provision into the future.

Interestingly, yesterday The Verge reported that another e-bike manufacturer, Cowboy, appear to have spotted the possibility of a PR-win: they are developing an app that will allow you to save the unique key from your bike and access the basic settings.

So there is hope that even if the company does completely fold, either someone like Cowboy will step in to save us, presumably in the hope of fostering customer loyalty and future sales; or smarter people than me will work out DIY options for keeping these bikes running.

And on top of that Zuckerberg has demonstrated time and time again an absolutely horrendous attitude towards user privacy. Remember, this is the man on record as having said “I don’t know why they ‘trust me’, dumb fucks”.

I think you can rest assured that Meta is a company you cannot trust to look after anything more than their best interests. Presented with two options, make more money or respect our users, I’m fairly sure that they’ll pick the make more money option every time.

Anyway, they’ve recently launched a sort of Twitter competitor called Threads. I haven’t really paid any attention to it. I did notice that it’s not actually launching in the EU at present because it simply doesn’t meet the basic requirements for respecting user privacy.

If that doesn’t put shivers up your back, then perhaps this article on Ars Technica comparing the data access permissions required on Threads versus other similar services.

https://arstechnica.com/security/2023/07/how-threads-privacy-policy-compares-to-twitters-and-its-rivals/

I counted it up. Threads requires 76 separate data permissions, compared with 4 on Bluesky and 0 on Mastodon. Twitter themselves were no slouches, looking for 53 separate data points.

Anyway, if this doesn’t fill you with the absolute heebie jeebies, then I don’t know what else to say to you. I recommend just saying no. Check out https://joinmastodon.org/

Some time ago the EU implemented very well intentioned legislation to protect unwitting Internet users from invasive tracking of their browsing habits by websites using cookies – a small text file stored in your browser to which the website has access and and can store identifying things about you.

There are useful features with cookies (persisting login state, preferences etc), but unfortunately they’ve been so vastly abused that it became an absolute imperative to rein them in – and thus it was required that websites ask users if they want cookies or not.

Unfortunately, the implementation of the legislation has caused user experience issues ever since. Anyone in the EU that has used the Internet will of course know what I mean – you land on a website to read an article etc, and instead of being able to read said article you are immediately faced with a large pop up window informing you of cookies and the website’s desire to place them on your computer.

Even more unfortunately, some websites decided it was a great opportunity to flex their dark-pattern muscles, and make the process of rejecting cookies one fraught with confusion and misdirection. Leading to the situation in some cases that to actually fully reject cookies requires clicking through multiple tabs of information and individually turning off cookies from 10-40 different cookie providers, which obviously no one really has the time for.

The Automatic Solution

I’ve been dually frustrated and fascinated by this whole thing. Frustrated because it just is incredibly repetitive and irritating, fascinated because I am amazed the lengths some website owners will go to to trick and confound users who just want a simple way to opt-out of this.

Luckily I’m not the only one who shares this frustration and smarter people than me have already released options to automatically detect and reject these cookies.

There seem to be a few different options around, but from a brief bit of research, Consent-O-Matic seemed like a solid option, and it’s available for Firefox and Chromium based browsers (eg Chrome, Edge, Vivaldi and many more).

Once I had it installed and setup I visited a few websites and boom, not a cookie nag in sight!

The settings are simple – you enable the options you want the plugin to always attempt to select. If said options aren’t available it will default to rejecting cookies where possible.

If the cookie pop up isn’t something it’s programmed to handle, then you’ll have to manually handle it, but after one day with the plugin installed I’ve already found it to be an absolute joy (cookie popups evaporate in front of your eyes ).

I’ve no idea why I didn’t look for a solution like this sooner.

So don’t hesitate, install this plugin in your browser and be a lot happier.

And some more good news

It sounds as though the boffins at Firefox are looking at implementing this feature natively in the browser settings.

According to this article at ghacks.net, the development versions of the browser are already testing the functionality. So with any luck soon cookie popups will be a thing of the past for me.

It will never be 100% perfect, but I suspect they’ll manage to catch the vast majority of cases and auto-reject them on your behalf.

There aren’t any hard rules on this – it’s down to personal preference – but I think it’s worth spending a little time early on thinking about how you’re going to use them, and then stick to it.

I have always felt Categories should be the bigger ticket divisions of content – the main groupings – whilst Tags are a way to group related content across categories.

In my case I split it roughly into two categories, Journal and Technology. Though I think I’ll likely add an Arts category to cover anything I write related to music, books, film and art.

With tags though I try to not go crazy. The goal is to have a relatively minimal number of tags which will be reusable. I think the thing to consider is this, if someone clicks on a tag, it’s not very useful if there’s only one post in that tag archive.

So for me I always pause before I create a new tag – is it something that will help to group a decent number of posts in the future? Will it be something I’ll likely reuse?

I’ll also periodically go into the tags in the WordPress admin area and review them. If I see a tag with only one or two posts, I might decide to get rid of it and re-tag the posts in question.

All of this is only really useful if its exposed to your visitors. In my photoblog for example, I haven’t used categories at all for now. Every post will be a photo and I tag everything under a fairly small set of reusable tags. Perhaps categories would make sense for stricter divisions like nature, street, and other thematic groups, but I think I’d only use that if I was sticking to one or two photo topics.

So I tend to tag photos according to whether they are:

• colour/black & white,
• what the general theme of the photo is (street, portrait, landscape etc),
• and also the country in which the photo was taken.

I figure this makes it easy for curious visitors to see photos grouped into a tag that might be of interest to them.

It’s easy to add a load of random tags to a post, much the same way people might throw a bunch of miscellaneous tags on a tweet (or a toot), but I think it’s worth taking the time to make them useful to your visitors.

Large Language Models (I much prefer this term to AI) – it’s a hard topic to avoid these days, and it’s also a vast and complicated one.

Perhaps if this had come along 15 years ago, I might have had some enthusiasm – but I’ve found a lot of cynicism has crept into my thoughts regarding technology over the past 10-15 years (possibly starting with the Edward Snowden revelations a little over 10 years ago).

But I noticed something since re-starting blogging that reminded me of some thoughts I’ve had of late regarding the impact LLM may have.

A new kind of spam

Of course we’ve dealt with spam in our lives for a long time, but in general, at least for me, it’s mostly been immediately recognisable.

When I restarted my photo blog over at Eidetic Opacity, imagine my delight when I had a number of followers appear after my first few posts. I thought to myself how I missed this, simple pleasure of knowing someone has seen the photos I posted and liked them enough to want to see more.

On further investigation though, I saw a pattern – all of the ‘people’ that had followed me had eerily similar WordPress blogs, with eerily similar posts that followed a similar pattern. In isolation, and without having seen it before, you might not have realised that all of these blogs were in fact being written with the aid of LLMs.

So sadly, I guess these follows were just an effort to get me to look at their sites and hopefully earn some advertising money.

A few interesting articles on this topic

This occurrence reminded me of a couple of recent articles in The Verge one titled “inside the AI Factory” another titled “A Storefront for Robots”. They’re both fairly long reads, but I think well worth spending time on. In the first we see that, despite proponents of the technology talking about it removing dull repetitive tasks, it has in fact spawned a huge industry of even duller and more repetitive tasks; the second, discussing how SEO requirements, now coupled with LLM writing tools, has resulted in a huge amount of garbage appearing on the internet.

So we end up in this strange situation.

We have increasingly convincing content being mass generated by huge spam farms all hoping to cash in on a drip feed of advertising dollars, while at the same time trying to fit in with the often strange dictates of search engine requirements.

Companies find it harder to be discovered amongst all the spam noise, thus spend more money on advertising to drive visitors to their sites, but also start using LLMs to generate more content that matches the weird requirements of SEO.

Google et al are laughing all the way to the bank.

The spam creators then use bots to visit spam farms and “click” on adverts to increase their revenue.

We end up with an internet where substantial portions of it are fake generated content, and potentially much of the ‘browsing’ is also fake – and it’s only going to get worse. This old article from 2018 in the NY Magazine suggests that studies back then show less than 60% of internet traffic is real (and it gets even worse on YouTube).

At what cost?

There are unfortunately numerous costs to consider.

Terrible “real” user experience

Have you ever searched for a recipe online wanting to quickly check what ingredients are needed, only then to be greeted with a huge amount of filler content at the start of the article, in-depth histories of the foods involved, descriptions of the methodologies, and of course, a heavy dose of advertising. Only at the very end of the article do you get a simple “Ingredients” and the short steps involved.

Every time I come across this it irks me – but this is what Google has mandated as the better format, and if a recipe site wants to surface in organic search results, it has to follow this format or languish in obscurity.

It goes further than recipes of course, many search results land you into bland generic keyword stuffed articles.

New Insecure Dull Repetitive Work

As discussed in the AI Factory Verge article, there is now this weird hidden workforce who spend their days chasing unstable work which is dull, repetitive, and entirely void of any obvious reason (due to tasks being parcelled up in such a way that workers don’t really know what they’re working on).

A hugely unregulated area where worker rights no doubt take a back seat.

Massive Energy Waste

Another thing that just doesn’t seem to get discussed nearly often enough regarding this whole situation is the massive waste of energy that is going into all of this.

It’s probably extremely difficult to accurately quantify, and in my own research it’s hard to find good information regarding this topic. I have found some articles discussing the energy costs of training a LLM. I also found some discussion on Stack Overflow estimated between 100Wh and 300Wh per ChatGPT request (the equivalent of charging your phone 20-60 times). If this is the case, it’s enormously wasteful – just take a few moments to write that email reply manually!

I dread to think the cost of generating images with tools like Dall-E and Midjourney.

And all the major tech companies are racing to put these tools right in front of people – write your emails, answer messages, draft document layouts etc etc.

All of these queries are using energy at a time we desperately need to be finding ways to reduce our energy consumption.

And what should we do?

I don’t have any clear answers for you here. I guess I’d like to raise awareness and make people think a bit about what is going on when they use a LLM to generate some stock text. Make some noise about what the environmental cost of all of this is. Is there any clarity on the working conditions of the people making these LLMs possible?

Only by making noise on these issues can it become either a regulatory topic, or a market force that requires tech companies to effect change in order to stay competitive.

I welcome any feedback though – like I say, it’s a complex and difficult topic, and a lot of the companies involve don’t release detailed information on either the ‘supply chain’ or the energy consumption.