outlier-nav

Issue 1

Issue 1

The Sachs Plan for Unity

cyber-crime

Founder of P Capital Partners, Vice Chair of the Open Society Foundation and a founding Council Member of the European Council on Foreign Relations, and patron of numerous foundations for business and political progress, Daniel Sachs is a leading voice promoting the business case for democracy and calling for a unified European growth agenda. Here, he outlines some of his thinking on a unified European innovation strategy.

7 Min Read

“Europe remains competitive by most other measures. If we exclude Silicon Valley and Seattle, Europe is actually more productive than the US.”

Europe presents a plethora of exciting investment opportunities. What would you say are its major advantages?

We have some very strong local clusters in a range of industries: biotech, energy, software engineering, and more. And some extraordinary local talent. The heads of AI at Microsoft and Meta are from the UK and France, and innovation teams at many of the world’s largest companies are European.

We have cutting-edge knowledge in Europe, very innovative people, and a lot of groundbreaking technology. And then there’s the European economy itself. It’s more inclusive than the US, with a more generous welfare state and more focus on sustainability transition and the green economy. There are a number of things that are attractive about the European economy as a contrast. I’m not saying it’s better than the US model—it’s different. And there are real benefits for innovation and entrepreneurship.

Finally we’re used to dealing with regulations. Many business leaders see this as a crutch, but there are ways to use regulations to our advantage. We actually have a headstart in dealing with privacy laws, human rights, and climate legislation.

What do you think needs to change to maximise European investment in technology?

There are three things we lack. One is a broader European market or approach. We’re too fragmented. 27 different state policies won’t cut it. Europe must devise a joint competitiveness strategy and invest in the areas of the economy where we can have a real advantage.

The second piece is that we lack “guts”, for lack of a better word. Some of these American success stories came about because they took for granted that they would dominate the world with their tech and their strategy. That’s harder to come by in a European context.

And then we have a lack of growth funding. There’s lots of funding for mature businesses and we have a lot of seed funding. But most of the best ideas reach the growth stage and get funded by US investors. The kind of innovation we need is partially unicorn-level, and truly disruptive. But a lot is more capital intensive, and slower to develop. Hydrogen technology or hardware development – this can’t be 100% funded by equity.

There’s quite a lot of low-hanging fruit here. Institutions and the private sector can get more involved, but it can also be incentivized by government guarantees, the European Investment Fund, KfW in Germany, and so on. A lot of those systems have been focusing on seed funding, which is now a very well-functioning part of the market. But not enough is going towards scaling up promising technologies and companies.

How do you characterise the European ecosystem in comparison to the US and China?

The Draghi report was a huge wakeup call. In Draghi’s analysis, the comparative reduction of EU productivity stems directly from a lack of investment in technology; but Europe remains competitive by most other measures. If we exclude Silicon Valley and Seattle, Europe is actually more productive than the US. Silicon Valley has been there since the late 1930s and we can see the long-term cluster effects of academia, capital, tech innovation, and people gathering around the innovative economy.

We have clusters in Europe too, but they’re fragmented by country, and building these clusters is about the co-location of academia, innovation, capital, business leadership, lifestyle, and all of that acting together in the long term to attract talent and capital.

It’s how you share knowledge, who you invite to participate. And then these things become organic over time. A lot of clusters have come up through academia and innovation first, and then attracted the business people. We should be interested in deepening the clusters in Oxbridge, Paris, and so on, and making them beneficial at the broader European level.

Do you think a unified Europe is a realistic and necessary goal?

Yes, for Europe to compete with the US and China we need to make the most of our relevant advantages on a European level, not one nation at a time. Authoritarianism and protectionism are both on the rise. Some people in Europe are happy, but the air under the wings of authoritarians is something we should all be wary of. There’s the notion of an ideological battle, but I think it’s largely misunderstood. The meta issue is really the crisis of trust in institutions.

There’s no silver bullet to fix this, but it starts with political parties and institutions getting serious about radical renewal and connecting with people broadly. New faces, new voices, and new ways of interacting with people. Over time, that builds social cohesion, and less extreme, polarizing, and toxic outcomes.

The political forces in the center need to get their act together. They have to get radical about their approach to break through to people. We need a lot of institutional change in the democratic and political setup, and I see it starting to take shape.

Let Innovation Grow

outlier-nav

Issue 1

Issue 1

Let Innovation Grow

cyber-crime

Philippe de la Chevasnerie is Founder & CEO of Papernest, a European scaleup which simplifies utilities and subscription management for households and small businesses. Founded in 2015, Papernest has grown rapidly to over 900 employees. The platform helps more than 1.5 million users gather, subscribe, and cancel services across France, Spain, and Italy which, understandably, means that Philippe supports a more unified Europe.

5 Min Read

“It would be great if Europe could act more as one entity and not 27 different states.”

What do you think Europe can do to promote further growth?

It starts with education. We have a good education system in France. We have excellent engineers and are already leading in areas like AI. But I see very few comrades from engineering school who have created their own companies. Engineering schools are trying to create engineers rather than leaders. They’re not teaching people how to launch and lead successful companies. So we need to adapt the curriculum of these universities.

We need more scientific people in leadership roles and founding companies. And we need to teach more soft skills around management, hiring, negotiation, and running your business. Not just the engineering hard skills that many of them already have.

And I do think we would benefit from a shift in mindset. We can’t do everything: increase wealth, work less, defend ourselves, manage the climate transition, and more. If we want to take the top position, we will have to make hard choices. We need to decide on our priorities and encourage people to push themselves towards them.

Is political change needed to facilitate this?

It would be great if Europe could act more as one entity and not 27 different states. We are far too small to compete with China and the US, so we need to encourage unity.

A good initiative right now is a single corporate status to let companies operate in all EU countries. That would be a great thing and would solve a lot of issues. We also need better equivalencies between the licenses you get in each country. Once you get them, licenses should be valid in all EU countries.

Right now we need 27 different licenses just to operate across Europe, which is a huge hurdle. On top of adapting to different languages, cultures, and client bases, you also have to deal with local regulations.

How can Europe change its reputation for over-regulating

We should regulate smaller and better. GDPR is a classic example: Europeans lose something like 500 million hours per year just clicking on useless cookie banners. Europe could have focused this regulation on the six browser companies most people use. Millions of businesses had to spend hours with GDPR lawyers to set up cookie policies. It doesn’t seem to have had any impact on spam or privacy, and it’s wasting a huge amount of time and energy for companies.

Artificial intelligence is similar. We should regulate only when it’s necessary and not before. We need politicians who will make the tough choices and regulate less and better.

We’re trying to build a complex system to avoid just executing the obvious initiatives (instead of mainly taxing oil, gas and coal, governments create very complex layers of incentives to reduce emissions instead of relying on price signals). Regulations become very complex, and ultimately useless.

What opportunities do you think will emerge from the current economic and political volatility?

Since 2020, from my point of view, volatility has exploded. You had Covid, political instability, wars, economic crises, and energy crises. And it doesn’t seem to be slowing down. To be prepared, you need to diversify, especially internationally. The more geographies you’re in, the more you can cope with issues in any one of them. Having a portfolio attitude to risk is good. You need to be one step ahead and have mitigation plans in mind.

You also have to be obsessive about cash efficiency. Be lean, and don’t just throw money at problems.

Be very willing to adapt. It’s about risk taking and the willingness to try new things. We had to pivot almost right away when we launched.

Finally, create a great team. A small team of highly engaged, hard-working people will really drive efficiency. We have to be proud of working hard and giving a lot for the business to succeed.

Open Models, Distilled Data

outlier-nav

Issue 1

Issue 1

Open Models, Distilled Data

cyber-crime

Phelim Bradley is CEO of Prolific, a technology company building the biggest pool of quality human data in the world – and the ultimate platform to access it. Phelim is also a Venture Partner at Pioneer Fund, and worked in genomic medicine and computational biology before founding Prolific.

5 Min Read

“We’re going to see models that excel in specific niches, and can truly bring the same level of expertise as experienced employees. That means training models on much more focused data sets.”

Has AI investment and enthusiasm peaked, and do you think we’re likely to see AI fatigue soon?

Certainly not. We’ll only see continued investment. The vast majority of this is coming from the US – revenue growth there is much faster than in Europe right now.

We’re going to see continued investment in AI hubs around the world. London and Paris have the potential to be these hubs, as do other cities in Europe. But today, a lot of growth comes from the US, and those networks are critical for our next phase.

What do you think the next steps are for artificial intelligence?

There are still real technical challenges to solve. The most famous models require enormous amounts of training data, which puts strain on servers and energy resources. There’s a trend towards teaching models to be just as good with much smaller data sets. These use less compute and therefore less data. This will make AI tools faster, more efficient, and much cheaper in the long run. Plus, it’s essential for sustainability.

We’re also seeing shifting demands towards the evaluation step in models. Evaluations are like the unit tests: how do you know that the model hasn’t regressed and still provides quality? So beyond building the models themselves, there’s more focus going into their maintenance and continued performance.

But the biggest change we’re working on involves more investment in domain experts and specialists. Large-scale models like GPT are good for broad tasks but aren't perfect for every use case. When training the latest AI models or conducting research, general data isn't enough, and you need responses from people with specific expertise or characteristics.

We’re going to see models that excel in specific niches, and can truly bring the same level of expertise as experienced employees. That means training models on much more focused data sets. And it also means consolidation – infusing AI into industry tools and using real user data from those, rather than broad internet scraping.

High-quality, targeted data from vetted participants leads to more accurate results and helps catch potential issues early. Companies like Carnegie Mellon and Layer 6 have used targeted participant pools to test and improve their AI models.

Does this mean we’ll see an end to LLMs?

No, LLMs will serve as core infrastructure, but we're also seeing a growing ecosystem of smaller, open models and distilled versions. The future will likely have both: large models providing the backbone, while specialized models handle specific tasks. Organizations are recognizing the value of this diversity, selecting models based on actual needs rather than their sheer size and power.

What role does journalism and its “hype machine” play in AI’s next steps?

The "hype machine" shapes both public expectations and research directions in AI. But the focus needs to be on responsible development rather than just technological advancement.

The real challenge is making sure AI systems are deployed responsibly, which requires attention to data quality and diversity of feedback and input. Transformative change often comes in waves of innovation that are more focused on practical applications and specific use cases. These may generate less hype, but can have significant impacts if not deployed responsibly.

How do you focus business strategy in an industry that’s evolving so quickly?

I like Jeff Bezos’ approach: focus on the things that never change. For us that means the breadth and quality of our audience, and the speed of innovation through our platform. We can always improve the data and find new ways of interpreting it.

But this requires a culture of customer obsession. Because it’s not up to us to decide which data is best, we need to understand and solve customers’ problems, whatever they are.

From Curiosity to Core Tool

outlier-nav

Issue 1

Issue 1

From Curiosity to Core Tool

cyber-crime

Dali Kilani is Co-Founder and CTO at FlexAI, delivering AI compute by rearchitecting infrastructure at a systems level. He’s also an angel investor and brings experience from Nvidia, Lifen, and BCG.

5 Min Read

“There are dozens of companies similar to ours in Silicon Valley. We’re a bit special here, and that opens doors.”

What stage of AI investment and enthusiasm are we at right now?

We’re still in the early days of companies deriving real value from AI tools. And for most people, it’s still too hard to do anything. So it remains a curiosity, not a core tool. As soon as things become simpler, the demand is just going to explode. I’ve personally seen this in healthcare. We used AI to simplify workflows and track patient outcomes. The change was completely amazing – people refused to work without it afterwards.

Things have to plateau in certain use cases, but there are plenty of dimensions where we’re not good enough yet. Things like AI software developer agents are coming. We’re never done. If you give researchers more compute, they’ll find greater things to do.

What are the advantages of building your business in Europe?

I’ve lived in Silicon Valley and worked at Nvidia, and I see the advantages of living here in Europe. The key point is, you can’t build a Silicon Valley company in Europe, and you can’t build a European company from Silicon Valley.

We chose our headquarters in Paris for a few reasons. First, there’s strong talent here. Many of them have tasted the Silicon Valley spirit but are still European. They want to be here and build great things in Europe. People are starting companies here all the time, including in AI. Mistral has proved that you can go big from Europe in this industry.

There’s also a lot of institutional support. There’s this rich mix of VC community, government, partners, and academia. For deeptech that’s incredibly important. Their doors are more open here than elsewhere.

Finally, it pays to be in a small pond – for now. There are dozens of companies similar to ours in Silicon Valley. We’re a bit special here, and that opens doors.

What does it mean to be an AI infrastructure provider?

We’re fundamentally a solutions company. Even though we’re starting with software, we need to retool everything. This starts from the workload down: networking, storage, computers – everything.

It’s difficult to defend software in AI infrastructure. The hardware people will keep coming from the bottom and eating up the software. That’s why full solutions are the best approach.

Everybody is running on Nvidia today, but we’re going to see lots of specialization.

Full solutions and tackling use cases completely is where the game is played. Nvidia is now the leading networking and data center, with the best software stack, and they actually built their own systems. You could spend $50 million dollars with Nvidia today to build your AI stack, and not spend a single dollar with anybody else. That’s where the future is, and that’s what we’re focused on.

Are the climate change concerns with AI legitimate?

The energy required today is huge and unsustainable. But that’s because of the way we’ve built models today: we started with solutions that are generic enough to address all workloads. But you can have more customized workloads that optimize things by a factor of 10. There’s a massive movement towards more efficient models. New open source models are very efficient, and companies are now emerging to build the chips to power them.

And then there’s actually using AI to discover new climate tech. Schneider is doing this already here in France. We’re also seeing investment in nuclear energy for AI among the biggest players. The market is incentivized to be more efficient because energy is expensive.

The key thing is that AI is still so new and developing so quickly that the climate footprint today is much higher than it will be eventually. And one of the key imperatives in developing this tech is to make it more efficient. There’s already a huge emphasis on this today, and that will continue.

Scaling Verticals with Deep Expertise

outlier-nav

Issue 1

Issue 1

Scaling Verticals with Deep Expertise

cyber-crime

Nabil Toumi is Co-Founder and CEO of Presti, a pioneering startup revolutionizing product photography for the furniture industry through generative AI and diffusion models. Presti provides tailored, high-quality imagery solutions that significantly reduce production time and costs. Previously, Nabil held various roles as a data scientist and analyst, including at Uber.

4 Min Read

“AI tech isn't just a plug-and-play solution. You have to train models to have deep expertise in order to be effective in complex domains.”

What’s the biggest misconception people have about building an AI company today?

One of the biggest misconceptions today is that you can just simply grab an off-the-shelf model or foundational model, and instantly create a high-value product. But when you deep dive, this is not true at all.

For example, here at Presti, we've run over 1000 model trainings before landing on the right one that achieved the results that really worked for our customers in the furniture industry. And that’s because, if you want to build something relevant, you need to really solve a big problem that provides clear value for your customers. I think this always comes with deep expertise in the domain you're working in.

The AI ‘hype machine’ is in full swing, but you’re in the trenches building. What’s something you think isn’t talked about enough?

The fact that you need to create a tool that solves a very specific problem. For us, it's that the furniture industry has a very specific pain point that needs to be solved: product photo shoots for 10 objects can cost up to $30,000 due to logistical issues and creative costs. Understanding that, and catering to that very specific need is what gives your build relevance and attracts investment.

We’re seeing a shift toward more specialized AI applications. Do you think the era of general-purpose AI is coming to an end?

Absolutely. AI tech isn't just a plug-and-play solution, where you can use a model like Chat GPT, plug it to anything, and it works. You have to train models to have deep expertise in order to be effective in complex domains. The higher the level of expertise, the more specific the focus, the better it will work. Adjusting our model in this way meant we could get a better quality of photorealism into our imagery and offer greater flexibility and control to our users.

How do you stay ahead in such a fast-moving industry?

We've been successful by focusing on one specific vertical. Our model performs very well in the furniture industry, and we've been able to achieve that by focusing our attention on training furniture products, which gave us relevance.

The second thing, on the product side, was to create something bespoke with a targeted UX and feature set for that industry that really meets their needs. It's not just about the AI tech, but the user experience, too.

And finally, on the commercial side, being very focused enables you to engage in good storytelling for your users, where you can help them to recognise themselves and their needs. That really allowed us to build a strong network within a specific industry, which makes you way more relevant than if you created a horizontal solution.

If you had to do it all again from scratch, would you still build Presti in Europe?

Definitely. We're here in Europe for many reasons. I think Europe is a great place to create a startup, especially Paris and France, because you have so much support to build in terms of accessing government funds. And on the tech side of things, the pool of talent here in France is probably the best outside of the US. And that's key to starting a successful business.