Deep tech


EU declares aim to become ‘quantum valley’ of the world

Q-day (the day when quantum computers will successfully actually break the internet) may be some time away yet. However, that does not mean that companies — and states — shouldn’t hop on the qubit bandwagon now so as not to be left behind in the race for a technology that could potentially alter how we think about life, the Universe, and well… everything. 

Spurred on by a discourse that more and more revolves around the concept of “digital sovereignty,” 11 EU member states this week signed the European Declaration on Quantum Technologies. 

The signatories have agreed to align, coordinate, engage, support, monitor, and all those other international collaboration verbs, on various parts of the budding quantum technology ecosystem. They include France, Belgium, Croatia, Greece, Finland, Slovakia, Slovenia, Czech Republic, Malta, Estonia, and Spain. However, the coalition is still missing some quantum frontrunners, such as the Netherlands, Ireland, and Germany, who reportedly opted out due to the short time frame

Ultimate aim: to create a globally competitive quantum ecosystem

“Quantum computing, simulation, communication, and sensing and metrology, are all emerging fields of global strategic importance that will bring about a change of paradigm in technological capacities,” the declaration begins. 

The <3 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

It further states that the bloc’s innovators and industry have not yet sufficiently mobilised to take full advantage of this potential as much as in other regions of the world. As such, it stresses the importance of building domestic R&D capacities for quantum technologies, as well as producing devices and systems based on them. 

In addition, it needs to invest in the whole quantum stack — from hardware to software and applications and standards, so as to safeguard “strategic assets, interests, autonomy, and security.”  

“The ultimate aim is to create a globally competitive ecosystem that can support a wide range of scientific and industrial applications, identify the industrial sectors where quantum technologies will have high economic and societal impact, and foster quantum innovation in small and large companies alike, from promising startups and scaleups to major industrial players — in short, to become the ‘quantum valley’ of the world,” the declaration reads.  

Thierry Breton, whose time as Commissioner for the Internal Market has been marked by a big tech regulation crusade, has declared quantum one of his “favourite subjects.” We can expect to see even more of a push towards greater collaboration across the bloc, should he land the top job of Commission President next year.

Potentially, Breton could get more member states on board to coordinate on a more detailed bloc-wide quantum strategy. With quantum engineering talent notoriously difficult to come by, this could indeed be key to keeping Europe from getting left behind in yet another key technology race.

EU declares aim to become ‘quantum valley’ of the world Read More »


Google’s Gemini AI won’t be available in Europe — for now

Yesterday, Google launched its much anticipated response to OpenAI’s ChatGPT (the first release of Bard didn’t really count, did it?). However, the new set of generative AI models that Google is dubbing “the start of the Gemini era” will not yet be available in Europe — due to regulatory hurdles. 

The tech giant is calling Gemini the “most capable model ever” and says it has been trained to recognise, understand, and combine different types of information including text, images, audio, video, and code. 

According to Demis Hassabis, CEO of Google DeepMind, it is as good as the best human experts in the 50 different subject areas they tested the model on. Furthermore, it scored more than 90% on industry standard benchmarks for large language models (LLMs). 

The three models of the Gemini AI family

The Gemini family of models will be available in three sizes. Gemini Ultra is the largest (but also slowest), intended to perform highly complex tasks; Gemini Pro the best-performing for a broad range of tasks; and Gemini Nano for on-device tasks.

Google says it has trained Gemini 1.0 on its AI-optimised infrastructure using the company’s in-house Tensor Processing Units (TPUs) v4 and v5e. Along with unveiling the Gemini family, Google also announced the Cloud TPU v5p, which is specifically designed for training cutting-edge AI models. 

The Google TPU v5p supercomputer processors
Google’s TPU v5p is designed especially from training advanced AI models. Credit: Google

What is truly an evolution in LLM application is perhaps the Nano, optimised for mobile devices. As told to the Financial Times, Nano will allow developers to build AI applications that can also work offline — with the additional benefits of enhanced data privacy options.

Explained in greater detail by the company in a blog post, Google is also providing the AI Studio — a free, web-based developer tool to prototype and launch apps using an API key. It will make Gemini Pro available to developers and enterprise customers from December 13. 

Just as for Bard, Europe will need to wait for Gemini

A “fine-tuned” version of Gemini Pro launched for Google’s existing Bard chatbot yesterday in 170 countries and territories. The company says it will also be available across more of its services, such as Search, Ads, and Chrome, in the coming months. 

However, users in the EU and the UK eager to test the mettle of Google’s “new era” of AI will have to wait a little longer. Google did not give extensive details, but said it is planning to “expand to different modalities and support new languages and locations in the near future.” 

Indeed, Google is reportedly planning a preview of “Bard Advanced,” powered by the multimodal Gemini Ultra next year. Google first released Bard in March 2023, but due to concerns around compliance with the GDPR, it did not reach European users until June. Let’s see how long we will have to wait for Gemini. 

Google’s Gemini AI won’t be available in Europe — for now Read More »


EU settles on rules for generative AI, moves to surveillance

The tech world is waiting with bated breath for the results from the final negotiations in Brussels regarding the EU’s landmark AI Act. The discussions that commenced at 14: 00 CET on Wednesday failed to reach a conclusion before the end of the day. However, negotiators did reportedly reach a compromise for the control of generative AI systems, such as ChatGPT. 

According to sources familiar with the talks, they will now continue on the topic of the controversial use of AI for biometric surveillance — which lawmakers want to ban. As reported by Reuters, governments may have made concessions on other accounts in order to be able to use the tech for purposes related to “national security, defence, and military.” 

Sources expect negotiations to continue for several more hours on Thursday. 

AI Act: innovation vs. regulation

While the AI Act — the first attempt globally at regulating artificial intelligence — has been in the works since April 2021, the rapid evolution of the technology and the emergence of GenAI has thrown a wrench in the gears of the Brussels machinery. 

The <3 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

In addition to having to understand the technological side to foundation models — and anticipate the evolution of the technology over time so as not to render regulation obsolete within a couple of years — member states have settled into different camps. 

Lawmakers have proposed requirements for developers to maintain information on how they train their models, along with disclosing use of copyrighted material, and labelling content produced by AI, as opposed to humans. 

France and Germany (home to European frontrunners Mistral AI and Aleph Alpha) have opposed binding rules they say would handicap the bloc’s homegrown generative AI companies. Along with Italy, they would prefer to let developers self-regulate, adhering to a code of conduct. 

If Thursday’s talks fail to generate (see what we did there) any definitive conclusions, fears are that the whole act could be shelved until after the European elections next year — which will usher in a new Commission and Parliament. Given the barrage of news of developments, such as Google’s Gemini and ADM’s new super AI chip, regulators may well need to rewrite the rules entirely by then. Oh well, that is Brussels bureaucracy for you.


Back to top

EU settles on rules for generative AI, moves to surveillance Read More »


AI can copy human social learning skills in real time, DeepMind finds

Human intelligence heavily depends on acquiring knowledge from other humans — accumulated through time as part of our cultural evolution. This type of social learning, known in literature as cultural transmission, enables us to imitate actions and behaviours in real time. But can AI also develop social learning skills the same way?

Imitation learning has long been a training approach for artificial intelligence, instructing the algorithms to observe humans complete a task and then try to mimic them. But usually AI tools need multiple examples and exposure to vast amounts of data to successfully copy their trainer.

Now, a groundbreaking study by DeepMind researchers claims that AI agents can also demonstrate social learning skills in real time, by imitating a human in novel contexts “without using any pre-collected human data.”

Specifically, the team focused on a particular form of cultural transmission, known as observational learning or (few-shot) imitation, which refers to the copying of body movement.

DeepMind ran its experiment in a simulated environment called GoalCycle3D, a virtual world with uneven terrain, footpaths, and obstacles, which the AI agents had to navigate.

To help the AI learn, the researchers used reinforcement learning. For those unfamiliar with Pavlov’s work in the field, this method is based on offering rewards for every behaviour that facilitates learning and the desired result — in this case, finding the correct course.

At the following stage, the team added expert agents (either hard-coded or human-controlled) that already knew how to navigate the simulation. The AI agents understood quickly that the best way to reach their destination was to learn from the experts.

The researchers’ observations were twofold. Firstly, they found that the AI not only learned faster when mimicking the experts, but also that it applied the knowledge it had gained to other virtual paths. Secondly, DeepMind discovered that the AI agents could still use their new skills even in the absence of the experts, which, according to the study’s authors, constitutes an example of social learning.

While the authors note that more research is needed, they believe that their method can pave the way “for cultural evolution to play an algorithmic role in the development of artificial general intelligence.” They also look forward to further interdisciplinary cooperation between the fields of AI and cultural evolutionary psychology.

Despite its early stage, DeepMind’s breakthrough could have significant implications for the artificial intelligence industry. Such an advancement has the potential to reduce the traditional, resource-intensive training of algorithms, while increasing their problem-solving capabilities. It also raises the question of whether artificial intelligence could ever learn to acquire social and cultural elements of human thought.

The full study is published on the journal Nature Communications.


Back to top

AI can copy human social learning skills in real time, DeepMind finds Read More »


‘Quantum-first’ microscope could solve chip inspection roadblock

Oh, the wonderful and mind-twisting world of quantum mechanics. However, in order to harness the magic-like potential of bending qubits to one’s will, there is a whole lot of nitty gritty engineering that needs to occur. 

The quantum revolution will not happen unless an entire ecosystem comes together, each part reaching the highest potential of its own expertise. 

And plenty of that development is happening in the Netherlands. Just today, Dutch startup QuantaMap announced it had secured €1.4mn in funding for its quality assurance tech for the production of quantum computer chips.

Quantum chips are not like regular computer chips, on many different levels (let’s set operating principles and data processing aside for now). One of these is that when they do not work like they should, there is not really any way of finding out why, and what has failed. This is to a great extent because it is so difficult to measure properties of the quantum chips without disturbing the qubits in the process. 

QuantaMap, based in Leiden, the Netherlands, has developed what it calls a “quantum-first” microscope that will allow both quantum researchers and chip manufacturers to closely inspect every chip and improve quality. 

What sets its technology apart, the startup says, is a combination of cryogenic scanning technology with quantum sensors, both specifically designed for quantum applications. 

“We are convinced that our technology will be instrumental for making good on the promises of quantum computing, enabling the societal advances that quantum technology can deliver,” said QuantaMap co-founder Johannes Jobst.

QuantaMap was founded in November 2022 by Jobst, Kaveh Lahabi, Milan Allan, and Jimi de Haan. The funding round includes investment from QDNL Participations, a fund that will invest €15mn into early-stage Dutch quantum computing startups in the coming years. 

Ton van ‘t Noordende, the fund’s managing director, said that QuantaMap’s unique combination of cryogenic scanning-probe microscopy and custom quantum sensors would solve the crucial challenge of producing reliable quantum chips. 


Back to top

‘Quantum-first’ microscope could solve chip inspection roadblock Read More »


New AI tool aims to democratise high-res image generation

In the world of AI image generation, tools like DALL-E and Midjourney are holding the crown — and not simply because of their high-resolution performance. The training of these models requires such substantial investment and resources that it inevitably leads to centralised services and pay-per-use access.

A new AI tool developed by the University of Surrey aims to reverse this trend and democratise the technology, by opening up high-res image generation to a wider audience.

Dubbed DemoFusion, the model allows users to generate high-quality images without the need to subscribe to a service, or own a very powerful computer. In fact, the system only requires consumer-grade RTX 3090 GPU that can be found in any mid-range gaming PC or a Mac M1.

The AI is essentially a plug-and-play extension to the Stable Diffusion XL (SDXL) open-source model, which generates images at a resolution of 1024×1024. DemoFusion enables 4x, 16x, or even higher increase in resolution — with a few simple lines of code and without any additional training. The only trade-off according to the team is “a little more patience.” We tried it at TNW and it’s about six minutes.

SDXL vs DemoFusion AI image generator
Credit: University of Surrey
On the left side: the result by SDXL. On the right side, the result by DemoFusion. Credit: University of Surrey

To achieve these high-res results, the scientists first generated low-res images and then enhanced them using a process called progressive upscaling. This improves the SDXL’s detail and resolution by working across images in patches.

“For the first time, our unique technique lets users enhance their AI-generated images without the need for vast computing power, or any re-training of the model,” said Professor Yi-Zhe Song.

“Digital art and imagery is a powerful medium which everyone should have access to — not just a handful of wealthy corporations. That’s why we made DemoFusion publicly available. We believe it can enrich our lives, and everyone should be able to use it.”

The new technique is available online in the paper “DemoFusion: Democratising High-Resolution Image Generation with No $$$.”

Whether DemoFusion will gain enough traction to compete with giants like OpenAI’s DALL-E remains to be seen, but its creation is an important step to opening up AI’s image-generation potential to the public and the wider tech community.


Back to top

New AI tool aims to democratise high-res image generation Read More »


Mistral AI nears $2B valuation — less than 12 months after founding

European contributions might have been a little late to join the generative AI investment party, but that does not mean they will not end up rivalling some of the earlier North American frontrunners. According to people familiar with the matter, Mistral AI, the French genAI seed-funding sensation, is just about to conclude the raising of about €450mn from investors. 

Unlike Germany’s Aleph Alpha who just raised a similar sum, most investors come from beyond the confines of the continent. The round is led by Silicon Valley VC firm Andreessen Horowitz, and also includes backing from Nvidia and Salesforce. 

Sources close to the deal told Bloomberg that Andreessen Horowitz would invest €200mn in funding, whereas Nvidia and Salesforce would be down for €120mn in convertible debt, although this was still subject to change. If it goes through, this would value the Paris-based startup at nearly $2bn — less than a year after it was founded. 

Mistral AI was one of the few European AI companies to participate in the UK’s AI Safety Summit held at Bletchley Park last month. The generative AI startup released its first large language model (LLM), Mistral 7B, under the open source Apache 2.0 licence in September. 

Targeting dev space with smaller size LLMs

The key thing that sets Mistral apart is that it is specifically building smaller models that target the developer space. Speaking at the SLUSH conference in Helsinki last week, co-founder and CEO Arthur Mensch said this was exactly what separates the philosophy of the company from its competitors.

“You can start with a very big model with hundreds of billions of parameters — maybe it’s going to solve your task. But you could actually have something which is a hundred times smaller,” Mensch stated. “And when you make a production application that targets a lot of users, you want to make choices that lower the latency, lower the costs, and leverage the actual populated data that you may have. And this is something that I think is not the topic of our competitors — they’re really targeting multi-usage, very large models.”

Mensch, who previously worked for Google DeepMind, added that this approach would also allow for strong differentiation through proprietary data, a key factor for actors to survive in the mature application market space. 

Mistral AI and the reported investors have all declined to comment on the potential proceedings.


Back to top

Mistral AI nears $2B valuation — less than 12 months after founding Read More »


Ariane 6 rocket set to restore Europe’s space access next year

The European Space Agency’s Ariane 6 rocket is scheduled for its debut launch in mid-2024, its director Josef Aschbacher announced yesterday.

The news follows a successful hot-fire test on November 23 at Europe’s spaceport in French Guiana. The term ‘hot-fire’ refers to the fact that the engine is fired with its propellants, producing actual combustion and exhaust. The only difference from an actual launch is that the boosters are not ignited — keeping the rocket firmly planted to the ground.

“With the latest test complete, Ariane 6 has been through the essential rehearsals required for qualification,” said Aschbacher on X, formerly Twitter. “We have validated our models, increased our knowledge of operations and are now confident for our first launch period for Europe’s new heavy-lift launcher.” 

While the inaugural flight won’t carry major payloads in orbit, it will transport several smaller satellites. If that launch is successful, Arianespace, the company who developed the rocket, will aim for a second launch later in the year. That second launch would carry the CSO-3 reconnaissance satellite for the French military, said the company’s CEO Stéphane Israël in a press briefing.

Following that, Ariane 6 would be put to work conducting as many flights as possible. The long-term objective is to launch the rocket into space 9-10 times per year, said Israël. These would include 18 launches for Amazon’s Kuiper broadband megaconstellation project.  

The <3 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

Ariane 6 was first scheduled to launch four years ago. However, the rocket suffered a series of delays, attributed to technical issues, COVID-19, and design changes. 

With Ariane 6’s predecessor, Ariane 5, officially decommissioned and Italy’s Vega C rocket grounded following launch failure in December, Europe is currently without independent access to space satellites. 

So it is welcome news that Ariane 6 is on track for launch in around 6 months’ time — if all goes to plan that is. 


Back to top

Ariane 6 rocket set to restore Europe’s space access next year Read More »


Silo AI releases checkpoint on mission to democratise LLMs

A year has passed since OpenAI unleashed ChatGPT on the world and popularised terms like foundational model, LLM, and GenAI. However, the promised benefits of generative AI technology are still much more likely to be derived by those who speak English, over other languages. 

There are over 7,000 languages in the world. Yet, most large language models (LLMs) work far more effectively in English. Naturally, this threatens to amplify language bias when it comes to access to knowledge, research, innovation — and competitive advantage for businesses. 

In November, Finland’s Silo AI released its multilingual open European LLM Poro 34B developed in collaboration with the University of Turku. Poro, which means reindeer in Finnish, has been trained on Europe’s most powerful supercomputer LUMI in Kajani, Finland. (Interestingly, LUMI runs on AMD architecture, as opposed to all-the-rage LLM-training Nvidia.) 

Along with Poro 1, the company unveiled a research checkpoint program that will release checkpoints as the model completes (the first three points were announced with the model last month). 

Now, the company, through its branch SiloGen, has trained more than 50% of the model and has just published the next two checkpoints in the program. With these five checkpoints now complete, Poro 34B has shown best-in-class performance for low-resource languages like Finnish (compared to Llama, Mistral, FinGPT, etc) — without compromising performance in English. 

Research Fellow Sampo Pyysalo from TurkuNLP says that they expect to have trained the model fully within the next few weeks. As the next step, the model will add support for other Nordic languages, including Swedish, Norwegian, Danish, and Icelandic. 

“It’s imperative for Europe’s digital sovereignty to have access to language models aligned with European values, culture and languages. We’re proud to see that Poro shows best-in-class performance on a low-resource language like Finnish,” Silo AI’s co-founder and CEO, Peter Sarlin, told TNW. “In line with the intent to cover all European languages, it’s a natural step to start with an extension to the Nordic languages.” 

Furthermore, SiloGen has commenced training Poro 2. Through a partnership with non-profit LAION (Large-scale Artificial Intelligence Open Network), it will add multimodality to the model.

“It’s likewise natural to extend Poro with vision,” Sarlin added. “Like textual data, we see an even larger potential for generative AI to consolidate large amounts of data of different modalities.”

LAION says it is “passionate about advancing the field of machine learning for the greater good.” In keeping with Silo AI’s intentions for building its GenAI model and LAION’s overall mission to increase access to large-scale ML models, and datasets, Poro 2 will be freely available under the Apache 2.0 Licence. This means developers will also be able to build proprietary solutions on top. 

Silo AI, which calls itself “Europe’s largest private AI lab” launched in 2017 on the idea that Europe needed an AI flagship. The company is based in Helsinki, Finland, and builds AI-driven solutions and products to enable smart devices, autonomous vehicles, industry 4.0, and smart cities. Currently, Silo AI counts over 300 employees and also has offices in Sweden, Denmark, the Netherlands, and Canada.

Silo AI releases checkpoint on mission to democratise LLMs Read More »


Tech is bringing ancient ruins back to life. Here’s how

From bringing ancient ruins to life through augmented reality (AR) to 3D-printing centuries-old artefacts, cultural heritage startups are transforming the landscape of heritage preservation and education. By leveraging technology to foster a deeper connection with our past, this breed of companies help safeguard some of the most defining elements of human history.

TNW spoke with three innovative startups in the space to find out how they’re using tech to bridge the gap between past and present.


Over 2,000 years ago, the city of Baia near Naples was the go-to holiday destination for the elite of the Roman Empire. Known for its luxurious and hedonistic vibe, it attracted prominent figures such as Cicero and even Julius Caesar himself.

Today, about half of the ancient town lies beneath the surface of the Mediterranean.

Baia is one of the world’s very few underwater sites open to the public, accessible through snorkelling, scuba diving, and glass-bottomed boat tours. But preserving submerged ruins is no easy task.

archeological park of Baia
Diving in Baia. Credit: Campi Flegrei Sub Center

To help protect Baia, in 2019, the Italian Ministry of Cultural Heritage partnered with Wsense, a spinoff from the Sapienza University of Rome, which specialises in underwater monitoring and communication systems.

“Since GPS, radiocommunication, and satellite signals don’t work underwater, you need to build your own infrastructure for the underwater domain,” Chiara Petrioli, founder and CEO of Wsense and Professor at the Sapienza University of Rome, tells TNW.

Wsense has created a subsea Wi-Fi so that real-time data below the water’s surface can be collected and transmitted back to land. To serve that purpose, the deep tech startup has developed a network of wireless IoUT (Internet of Underwater Things) devices.

Specifically, Wsense’s system relies on multi-sensor nodes, which provide information on various aspects of water quality, from temperature and pressure to pH, salinity currents, and tides.

Data can be transmitted in two ways. Firstly, from one node to the other, a process that is optimised by an AI algorithm that changes the transfer path when sea conditions change. Secondly, it can be transferred to the surface through Wsense’s gateways, which, either integrated into floating buoys or posted on nearby land, connect the underwater network to the cloud — and from there, to the rest of the world.

In the case of Baia, this system allows for remote in-situ monitoring, which doesn’t simply set off alarms in case of unauthorised access, but most importantly provides water information critical for the site’s preservation.

This includes tracking the environmental conditions that could distort the artefacts. It further entails observing CO2 emission levels to understand how the area’s volcanic activity is developing, while enabling the study of climate change’s impact on underwater cultural heritage.

Here’s a video with how Wsense’s system in Baia works:

In addition, the technology has provided a valuable tool for archaeologists diving in the submerged city. Thanks to special micronodes attached to a waterproof tablet, divers can communicate both with each other and with their colleagues above the surface. “Think of it as an underwater WhatsApp,” Dr Petrioli says. At the same time, these micronodes create a type of underwater GPS that helps locate divers in real time.

“We have also been collaborating with a partner to develop an AR app on our tablet, which visitors can use to view 3D reconstructions of Baia while at the site,” she adds.

diver at the underwater city of Baia
Footage of diver using Wsense’s tablet in submerged Baia. CreditL: Wsense

Besides the preservation of cultural heritage, the startup’s technology has multiple areas of application, including environmental and critical infrastructure monitoring and aquaculture. Last January, the World Economic Forum named it “the world’s most innovative company in collecting and managing big data for the purpose of protecting the ocean environment.”

Founded in 2017, Wsense has grown into a team of 50 people, with offices in Italy, Norway, and the UK. In October, the award-winning spinoff completed a €9mn Series A round, raising its total funding to €13mn.


It doesn’t take a knight in shining armour to save a castle in distress — and that’s exactly what Dartagnans has been proving. Named after Dumas’ famous musketeer, the Paris-based startup is fighting to save and promote castles that would have otherwise fallen into oblivion.

“We wanted to save a castle from A to Z.

Founded in 2015, the startup began as a crowdfunding platform connecting donors to owners/managers of historical monuments. By gradually building a community, Dartagnans became France’s leader in crowdfunding for heritage preservation, just after the first two years of operation.

“After a point, we wanted to have our castle and save it from A to Z,” Romain Delaume, Dartagnans’ co-founder and CEO, tells TNW. “We didn’t have enough capital to buy one but we had a growing community.”

So in 2018, the startup reinvented its business model and introduced the collective purchase of castles concept, offering the opportunity for anyone in the world to invest in endangered castles and become co-owners.

“When we launched the first collective purchase campaign, we raised over €1.6mn in 45 days,” Delaume says. “This means that when you give the opportunity to people, they all gather for a cause regardless of their background.”

In the past five years, Dartagnans has helped save four castles in France: the Château de la Monthe Chandeniers in Vienne, the the Château de l’Ebaupinay in Deux-Sèvres, the Château de Vibrac in Charente, and the Château de Boulogne in Oise.

The Château de Boulogne
The Château de Boulogne. The 19th-century castle features a distinct architecture, inspired from history and esotericism. It was commissioned by Count Charles de Boulogne, a rich Belgian landowner. The castle suffered terribly during WWI and, now, nearly 7,500 people have become co-castellans to save it. Credit: Dartagnans

Following the purchase, the castles go through gradual restoration and are opened to the public for touristic activities, such as visits, events, volunteer projects, and hospitality programmes. The self-funded startup now counts over 50,000 co-castellans and an international community of 300,000 heritage defenders. Since its founding, it has raised €15mn for the safeguarding of monuments.

Co-castellans can invest in castles on the startup’s platform and, in return, they receive ownership shares, which they can keep, sell, or pass on to their children or friends. “It’s a share of a company,” Delaume explains. “We create a company for each castle we operate and then we sell the shares to the public.” Each share costs €79.

According to Delaume, Dartagnans owns one-third of the castles, which allows it to pilot the company and carry out restoration, management, marketing, and tourist activities. “I am like a CEO with thousands and thousands of little shareholders,” he says.

The Château de l'Ebaupinay
The Château de l’Ebaupinay. Classified as an historical monument since 1898, the medieval castle is a rare remnant of the mid-15th century architecture. Credit: Dartagnans

Nevertheless, co-castellans have their own say in big management decisions, with each share representing one vote. The community is also involved through activities, meetings, and assemblies, both in person and online. The company’s biggest event is The Night of the Castles (link), when hundreds of castles across France and Europe simultaneously open their doors at nighttime.

Dartagnans currently employs 14 people and operates solely in France, with future plans for international expansion. Within the next decade, Delaume hopes that they’ll have accomplished half of the restoration needed for the castles. Another goal is to keep growing what he calls “a happy community.”


Standing in front of historical ruins or a vase dating back to 500 BCE can cause a feeling of detachment. Even for those with a vivid imagination, reconstructing the past from a centuries-old object is no easy task — but thankfully, technology can help.

Hi.Stories was founded in Sicily in 2017 with the mission to integrate digital technologies into cultural heritage to help facilitate its communication, and in turn, its protection.

The startup offers multiple services. It develops 3D models and prints of museum artefacts; it designs apps for archaeological sites and museums, using storytelling narratives and gamification; and it creates virtual tours based on augmented reality (AR).

Sicily Stories App
The startup developed the Sicily Histories app to help visitors explore the region’s sites using AR and 3D reconstructions. Through storytelling and a gamified narrative, users can help characters explore Sicily. Credit: Hi.Stories

One notable advantage of these tools is that they increase visitors’ interactive experience, and in turn, their engagement with heritage.

“Communication through the realisation of digital use systems allows heritage to be read at different levels: the visitor — on-site or remotely — becomes the protagonist of his or her own visit, being able to choose different degrees of immersion,” Luna Meli, co-founder of the startup, tells TNW.

Another advantage is the improved accessibility of exhibits, which goes well beyond the obvious benefit of accessing sites or museum collections remotely.

The 3D reproduction of objects, for instance, offers an alternative for groups such as individuals with visual impairments to approach works of art through touch. According to the company, this particular service can be used for educational purposes as well, enabling students to develop a direct, physical relationship with artefacts.

3D print of the Capitoline Wolf
3D print of the Capitoline Wolf, the symbol of ancient Rome. According to the legend, the female wolf nurtured the twins Remus and Romulus, the mythical founders of the city. The original statue is displayed at the Palazzo dei Conservatori. Credit: Hi.Stories

Meli says that, after the pandemic, awareness of the need to use digital technologies for cultural valorisation and appropriation has grown. This has led to an increased demand for these services — especially regarding the creation of content and platforms that can be used in multimedia guide applications, webapps with AR, and immersive tours. Meanwhile, 3D models and prints have shown the biggest demand, partly because of their potential to improve the accessibility of exhibits.

In the video below, you can watch part of the startup’s 3D reconstruction and virtual tour of the Necropolis in Via Sant’ Euplio in Catania: