Earlier this week — in an impressive demonstration of the fragility of our efficacy-optimized economy — a few centimeters of snow combined with temperatures a few centigrades below zero let the Netherlands descend into utter chaos. The railway system collapsed, the national airline stranded thousands (leading to predictable letters to the editor that it be stripped of the K in its name that denotes royalty), and anyone was advised to stay at home to avoid bodily harm.

In the midst of this, I had planned to travel to Warsaw by train. My first attempt — I had decided to take one connection earlier to allow some additional buffer time for the connection in Berlin — failed because the train never showed up. When I went back for my originally scheduled connection, it turned out that the service to Berlin was the only train still running; every other train had been cancelled due to an ICT failure in the systems of the national railways. The ICEs, however, are run by the German Railways.

Somewhat miraculously, the train departed almost on time — we had to wait for some missing staff. Sure, there was no service from the restaurant car — “Unfortunately, the Bordbistro is closed because the staff did not come because of the weather conditions” — and the train was a bit overcrowded with people using it to get to other parts of the Netherlands, but we made it out of the chaos-stricken country only 20 minutes behind schedule.

After the border, a new Bordbistro crew materialized and started serving lunch and coffee. We made it to Berlin with 15 minutes of delay, which allowed me to have a chat over coffee with Berlin-based colleagues before boarding the 1700 PKP intercity service onwards to Warsaw, which after a rewarding meal in the restaurant car — the WARS restaurant cars remain the best in Europe — and an otherwise uneventful trip delivered us right on time at 2300h, into Warszawa Centralna.

Somebody should tell the angry letter-writing KLM customer that there is a train going from Berlin to Amsterdam every two hours every day. It beats flying even when the weather is nice.

So over the wekend we learned that Sam Altman is apparently nervous about the impact of AI on democratic elections:

i am nervous about the impact AI is going to have on future elections (at least until everyone gets used to it). personalized 1:1 persuasion, combined with high-quality generated media, is going to be a powerful force. (@sama on x, 04-08-2023)¹

And that even though he is the CEO of the company that has arguably done more than anyone else to get generative ML tools into the hands of the public, he cannot come up with anything better than “raising awareness”:

although not a complete solution, raising awareness of it is better than nothing. we are curious to hear ideas, and will have some events soon to discuss more. (@sama on x, 04-08-2023)

It seems that Altmans “concern” is something that has already had its impact on OpenAIs public policy narrative. A few hours after Altman’s tweet, Open AI’s new European head op policy and partnerships had translated this into a call to action:

Who are the best thinkers/builders at the intersection of generative AI and elections in Europe? Ideas welcome! (@sGianella on x, 04-08-2023)

Now, there are plenty of reasons to suspect that Altman’s “nervousness” is nothing more than self-serving criti-hype (see here for such an interpretation), but in this case it is worth dwelling a bit on the particular connection being made between AI and elections (especially since both the US and Europe have upcoming elections).

There are indeed reasons to be concerned about the impact of generative ML models on elections and other democratic processes², but it is beyond absurd that such concerns are being articulated by Open AI and its representatives:

The way in which Open AI introduces its models into the public sphere stands in stark contrast to the very norms of openness, transparency, and equality on which democratic elections are based. Instead of these values, Open AI deliberately hides how its models are trained, making it impossible for researchers, policymakers, and the general public to understand “the impact AI is going to have on future elections.”

If Sam Altman and his surrogates are truly interested in limiting any undue interference of the technology they are peddling to the public with upcoming elections, then they should apply the same level of transparency to their publicly available models. Until that happens, we should keep ML systems as far away from the electoral process as possible.

In November 2020 - as the second wave of covid infections crested over Europe - I drove 418 kilometres through the night to report on the hearing in Case C-401/19 at the Court of Justice of the European Union in Luxembourg. At the hearing I was one of two accredited members of the press and the only one writing for a free access publication (the other journalist wrote for a paywalled trade publication). Apart from the two of us, the hearing was attended by a small group of lawyers who were following the hearing to privately report back to their industry clients (both entertainment industry and online platforms) and otherwise the spacious halls of Europe’s highest court were largely deserted.

Before driving down to Luxembourg i had considered to try to live stream the hearing. I had set up a twitch channel ("Radio Luxembourg") and brought some audio equipment that would have allowed me to capture and stream the audio of the hearing from the court’s audio system. Once arrived at the court and set up in the press centre, I quickly abandoned these plans after being reminded that the Court really does not want to have its workings communicated to the public. Usually I am not someone who quickly complies with senseless rules — but even I did not want to mess with Europe’s highest legal authority.

So I settled behind my laptop and engaged in a furious sprint of note taking (not necessarily an easy task given the multilingual nature of proceedings at the CJEU) and published the only(!) publicly available report of the hearing.

All of this is to say that I am extremely happy to hear that yesterday the Court has finally decided to embrace modern technology and will start streaming Grand Chamber hearings and judgements as of next week (pdf).

Coincidentally this means that the first judgement that will be live streamed by the court will be the judgement in Case C-401/19, the case that made me drive through the night to cover back in November 2020. Stock up on the 🍿 and tune in for episode one of season one of CJEU-TV on Tuesday 26 April at 0930h CET (and then join us for our COMMUNIA Salon to discuss the implications of the judgement on Thursday 28 April at 1530h CET).

Apart from this - happy? - coincidence the decision of the Court to finally stream its public proceedings is extremely welcome in general. Justice and democracy do not flourish in darkness and scenarios in which public access to the deliberations on fundamental rights is dependent on some random activist being able to drive across borders are not conducive to trust in our democratic institutions.

Back in November 2020 after the hearing, i expressed my frustration about the senseless aversion to stream the hearing to the court’s extremely friendly and helpful press officer, who turned out to be just as frustrated with this decision as i was. He also hinted that there were discussion ongoing among the judges and that he and his colleagues were pushing for more openness. As it is well known, the wheels of justice turn slowly but now — with a number of additional waves of covid behind us and the acute urgency gone for now¹ — the court has finally realised it does indeed make sense “to facilitate the public’s access to its judicial activity” by streaming its proceedings.

Now let’s hope that the CJEU’s streaming infrastructure will be up to the task to provide access to Tuesday’s judgement, which is highly anticipated by a lot of people.

This year I got myself a Punkt. MP02 mobile phone. I had been eying the Punkt. for a while as a secondary phone that I could use during periods when when I wanted to avoid the distraction potential that comes with my smartphone — vacations or an evening out at a restaurant or at friends for example. Sometime in 2021 the Punkt. — which is essentially a phone that can do calls, SMS and function as a 4G hotspot — gained the ability to send and receive messages over Signal — which given that I do my messaging over Signal and iMessage — made switching for periods of time feel like a viable option (unlike iMessage which will fall-back to SMS, Signal messages just stay undelivered if the receiving client is not active).

I got the Punkt. on xmas eve and while it is indeed a very nice piece of hardware, that feels nice in the hand I was disappointed almost immediately. Setting up Signal (via an app that is called Pigeon) was extremely cumbersome (mainly because is includes a step where you need to solve an image captcha, that is extremely buggy on the phone’s small screen) and once I had managed to set up Pigeon, the app kept crashing immediately after opening it. The other problem was that by signing into Signal on the Punkt. I was immediately signed out of Signal on my iPhone (which was kind of expected) but also on my laptop (i am a heavy signal user on the laptop and had not considered this).

At the time Punkt. had already announced that a new version of Pigeon would be released on the 27th of December so I boxed up the MP02, went back to the iPhone and waited a few days.

After installing the Pigeon update a few days later, I did manage to get Signal up and running on the phone (while suffering thought the same buggy captcha process again) and could receive and send (with all the drawback that come with suddenly being thrown back to T9 text input) Signal messages on the phone. A few days later we went on vacation to a remote corner of Denmark and I was determined to rely on the Punkt. for that week.

During the week in Denmark week I tried to stay away from the internets and anything work related as much as possible and the Punkt. certainly helped with that. Still other limitations of not using an iPhone became painfully clear almost immediately. While I had bought the Punkt. primarily to avoid distractions and to be less reachable there are in fact lots of functions integrated into modern smartphones that are simply helpful without being distracting.

I had expected that I would miss things like recording my runs on Strava and listening to podcasts while running (both of which work perfectly fine on my non-internet connected iPhone which I took along for my runs) but some other limitations where less expected: the fact that the Punkt. does not support contactless payment (a la Apple Pay) means that I needed to carry a wallet with me for the first time in two or so years. But it was the fact that the Punkt. does have no ability to display an EU Digital COVID Certificate is what ultimately made me swap my SIM-card back into the iPhone and abandon the whole experiment. There was no printer in our vacation house in the remote corner of Denmark but more fundamentally — this being the 21st century — I am simply not willing to go back to carry papers around with me as a trade-off for having less distraction in my life.

Tomorrow UPS guy will come and pick up the phone under Punkt.’s Hassle-free 30 days returns policy and I will be a bit sad that this experiment did not work out. The MP02 is a really nice piece of hardware and I would be happy¹ to give it another try if they manage to support Signal in a way that you can run it on multiple devices and switching SIM-cards does not involve silly captcha acrobatics anymore (assuming that by then COVID-19 will be endemic and there is no need to regularly show COVID certificates at random moments of the day).

Billboard magazine has a fascinating piece on the rise of the use of deepfake vocals to create what the piece terms “faux songs”: ‘It’s Fan Fiction For Music’: Why Deepfake Vocals of Music Legends Are on the Rise.

The piece showcases the recent rise of deepfake vocals (vocals produced by computers that closely resemble the voices of well known artists) in new songs that are generally produced without involvement of the artists who’s voices are reproduced¹. It is a super interesting introduction into a world of music production full of delightfull wired-ness. I particularly like this track featuring a recreation of Travis Scott’s voice, which combines glitchy aesthetics with what the agency producing the song “considered to be cohesive lyrics”:

Sadly the main thesis of the piece (“Faux songs created from the original voices of star artists are becoming more popular (and more convincing), leading to murky questions of morality and legality”) does not do this phenomenon much justice. The idea that this type artistic production enabled by digital manipulation techniques can somehow be understood though concepts like “faux” vs “orignial” strikes me as outright silly. Quite obviously the “faux” works are just as orignal as any of the “originals” that the voices are derived from.

Even more problematic are continuing attempts to fit this type of creativity into existing concepts like authorship and copyright. The realities of creative production today are so far removed from the realities in the late 19th century (when modern ideas about copyright and authorship were codified) that it really does not make much sense to try to analyse them throught the lens of these concepts.

We urgently need new concepts that do a better job at recognising artistic innovation while at the same time ensuring that the value created in the process is fairly distributed.

A quick note that i have moved this blog back to my own server. After 2 years or so as a ghost¹ instance hosted somewhere in the clouds provided by Digital Ocean it is back on my own server (provisioned by tuxic.nl). I have rebuild the entire site as a set of static pages generated with Hugo.

Unfortunately this move means that the new version replaces the very first (wordpress based) iteration of this blog which was also hosted here but had a different url structure and now external links to the first version are broken. I have attempted to create redirects for the old urls but have failed to do so (something to do with escaping “?“s in the old urls), which gives me considerable grief in having just read Jonathan Zittrain’s excellent long read on the rotting Internet in the Atlantic.

Lately there has been a lot of talk about reviving the Trans Europe Express (TEE) that had its heyday well before i was even born. If that means more investment into high-speed trans-european rail infrastructure it has my full support (even though it will be next to impossible to match the design sensitivity that came along with the TEE).

Speaking about railway nostalgia, I came across this illustration on flickr today, which illustrates the networks serverd by the TEE predecessor Simplon Orinet Express and its middle eastern equivalent the Taurus Express in the late 1930s:

If there is talk about reviving old railway networks then we should probably skip the TEE and go right here. If revieved, this network would cover more than enough of the world for me to never set foot into an airplane again and spend the rest of my life exploring these parts (as of writing i have visited or lived in 9/11 of the stops on the European side of this network and 7/13 on the Afrian/Asian side).

And since we are discussing reviving express trains here, i am also very much looking forward to the Chungking Express sequel that is apparently in the works.

early last year i wrote about the joy of using JUMP e-bikes during a short stay in San Francisco¹. Turns out that less than one and half years after being introduced, these bikes, while still being perfectly good and much needed, are getting retired and destroyed by the truckload.

This seems to be the outcome of an unfortunate but probably inevitable collisions of the logics of startup consolidation and the dysfunctional way how the US legal system structures liability. As many others have pointed out before me, destroying perfectly good bikes instead of giving them away to people for whom having access to a bike could make a world of difference, is the logical conclsuions of the sociopathic business model of so called “sharing platforms”.

I have always considered myself to be a fairly accomplished urban cyclist. Much of this confidence stems from having worked as a bike messenger between 1993 and 2004, mainly riding fixed with no brakes. Over the years, the rules of the urban cycling game have not changed much even though cycling became massively more popular during at the same time.

However, over the last few months the game is changing by the emergence of a new class of participant, the electrical motor assisted cyclist. As a result i am finding myself learning again, this time trying to master the art of drafting-off e-assisted cyclists navigating though urban traffic¹, which is a fun game to play… (these days on bikes with imperfect brakes which significantly increases the difficulty)

A couple of days ago,  I came across the website generated.photos (via the Verge) a new service that offers 100.000 computer generated portrait images and positions them as an alternative to traditional stock photos. The Verge article highlights the fact that the pictures can be used “royalty free” and the generated.photos website claims that “Copyrights … will be a thing of the past”. This made me postulate on twitter that we might very well be witnessing the emergence of an “auto-generative public domain”.

It has since become clear that the creators of generated.photos do not intended to contribute the output of their algorithms (or for that matter the algorithms themselves) to the public domain: By now the website has been updated to note that the images are available for non-commercial use only. A new terms and conditions page states that “Legal usage rights for content produced by artificial intelligence is a new, largely unknown domain” only to go on to list a number of restrictions on the use of the “materials and software” made available on generated.photos.

As noted in the terms of conditions the copyright status of images (and other types of artworks) that are autonomously created by AI-powered software is largely unsettled. As Andres Guadamuz notes in his excellent overview post on the topic, there are generally two schools when it comes to the question if computer generated artworks are (or should be) protected by copyright. One school argues that copyright protection only attaches to works that have been created by humans and as a result computer generated artworks can by definition not be copyrighted. The other school points out that such works are not created without any human intervention (someone needs to start up the software and set basic parameters) and that whoever initiated the generation of these works should be considered the creator and receive at least a minimum level of (copyright) protection as a reward for their investment.

In the case of generated.photos it is evident that the people behind the project have made a considerable investment into the project. The website states that they have shot more than 29.000 photos of 69 models that have subsequently been used as a training set for the software. Judging by notes on their website, the 100.000 images made available on the website have been created using the open source generative adversarial network StyleGAN that is freely available via GitHub. It remains to seen if creating photos (which are copyright protected) that are then used to train a out of the box GAN does indeed mean that the output of the network is (a) protected by copyright and (b) that the copyright belongs to the entity that trained the GAN.

While it seems to be at least possible that the creators of generated.photos do have a legitimate copyright claim in their output, that does not necessarily invalidate the idea that we are witnessing the emergence of an auto-generative public domain, i.e circumstances in which computer algorithms produce a (possibly endless) stream of artworks, that are indistinguishable from human created works and that are free from copyright and can be used by anyone for any purpose.

In terms of quality, the images provided by generated.photos are still far from indistinguishable from human made stock photos, but it is clear that it is only a matter of time before the technology gets good enough to produce high enough quality outputs at scale. Projects like the next Rembrandt illustrate this development is not limited to stock photography but will likely happen across the full width of human creative expression.

Such a development would dramatically upend a large number of creative professions. It seems like it will only be a matter of time before stock photography and other forms of creative work where the primary draw is not the specific style of a particular creator will be replaced by AI-generated output that will cost almost nothing to create. Once AI powered systems will be able to deliver high quality creative output at zero marginal cost the question if these outputs are protected by copyright or not will be largely meaningless (a single system releasing its output into the public domain will render any attempts to enforce copyright futile).

From the perspective of those making a living by creating stock photos, background music and other forms of creative work that is about to be eaten up by AI, the emergence of this “auto-generative public domain” must feel dystopian. Under these conditions the primary question that we must ask ourselves is not how we can fit works created by computer algorithms within the framework of copyright law. Instead we should ask ourselves how we can create the conditions for human creators to leverage these technologies as tools for their own creative expression. Instead of mourning a future in which humans are no longer employed to shoot endless variations of the same stock photos, we should look out for entirely new forms of creative expression enabled by these tools.