Since ChatGPT’s launch back in November 2022, businesses across the globe have leveraged the chatbot’s abilities to streamline their operations and reassign resources and staff to more high-priority activities.
In many ways, however, ChatGPT is the tip of a much larger iceberg. AI programs fulfilling all kinds of crucial functions will soon be at the heart of every business department, and many services we use on a daily basis. Their use will, however, soon be curtailed by regulations designed – in theory – to make us safer without stifling innovation.
Tech.co sat down with Dr. Rasmus Rothe, Co-Founder and Chief Technical Officer of AI R&D and investment platform Merantix (pictured above with fellow Co-founder Adrian Locher), to find out more about the company’s AI campus, regulating artificial intelligence, and the opportunities and challenges on the immediate horizon.
Inside Europe’s Biggest AI Campus
Founded in 2016, Merantix became the world’s first AI platform focused on researching, building, and investing in AI projects and companies.
Since then, the SoftBank-backed company – which currently sits on a venture fund totaling $35 million – has built close to ten AI companies in a variety of AI verticals, with a close focus on those that have “a big positive impact on humanity,” Rothe told Tech.co.
“There’s hundreds of use cases you can go after,” Rothe explains, “so why not focus on the ones that have a good impact on healthcare, or the climate, and so forth? That’s a very pragmatic approach, and that’s at the core of what we do, investing out of the fund. But we basically built a platform around it, and a whole ecosystem that is supporting our mission… beyond our own incubation.”
Initially, the AI campus – the company’s Berlin-based physical collaboration hub – was a way of co-locating its incubator companies. But Merantix quickly realized that it was worth co-locating other stakeholders in the AI ecosystem as well.
“Ultimately we see that now a lot in the policy discussions that it’s about research, it’s about industry knowhow, but it’s also about investment, it’s also about policymaking, it’s about ethics, so we figured let’s try to co-locate all these stakeholders in one place and have them all use one coffee machine,” Rothe says.
Rothe revealed that the AI campus now consists of “around 90 companies and 1000 registered desks” occupied by “everybody along the AI value chain.” This includes the Volkswagen AI team and other big corporates such as Amazon, as well as incubator startups that have partnerships with local universities. Investors like Index and First Minute Capital also work from there.
This year, Merantix is projected to hold 250 events on the AI campus, which range from “super technical paper discussion groups” to “business investor-focused events or policy focus events”, Rothe said.
“It’s really not about marketing,” he continued, “it’s really about discussing content and doing business with each other. There’s now a long waitlist to sign up to the campus so I think there’s quite some demand… people realize also that there’s value in co-locating with other interesting companies.”
There are a number of exciting machine learning ventures Merantix is currently scaling.
Using AI to Save Lives
“I think the breast cancer screening company Vara is very exciting,” Rothe said, pointing to the fact the company’s technology is already used in a third of breast cancer screening centers in Germany. “They’ve also shown in some of their data that AI plus a radiologist can outperform just a radiologist on many tasks.”
Simply put, early screening saves lives – and Vara is now venturing into emerging markets, Rothe said, which have a dearth of trained radiologists.
Cambrium, another Merantix venture, optimizes protein material, and has developed a proprietary protein design language which allows it to design tailor-made proteins for different purposes.
“Nearly a quarter of our global greenhouse gas emissions are because of the chemical industry, and there’s actually a lot of materials which are now petrochemically produced [that] can be produced based on proteins,” Rothe told Tech.co.
Rothe says that Cambrium’s protein optimizations are not just better for the environment, but also “create superior material properties in some cases.” What’s more, excitingly, the process is applicable across any type of material you could produce.”
Rothe also highlighted Deltia, which uses AI to provide insights on workflows within manufacturing settings – and Briink, which operates on the sustainable finance reporting space to make companies more ESG-oriented – as two projects worth keeping tabs on.
Balancing Regulation and Innovation
Along with investment and research, Merantix is heavily involved in AI policy consultation and implementation on a governmental level within Europe. Rothe himself is a founding board member of the German AI association.
Amid a global scramble to regulate AI, spurred on by the launch of ChatGPT last November, the EU was well placed to act quickly – it first introduced AI-focused legislation back in 2021.
A draft version of The EU’s new AI Act has recently been agreed upon and will now be negotiated by the Council of the European Union and EU member states.
“I think generally the idea of putting some regulation in place makes sense,” Rothe says. “I think nobody’s against regulation. It creates clear guidelines and that some of these things should be forbidden, whether it’s manipulating election or misusing personal data.”
However, he pointed out that a lot of AI use cases found in “so-called high-risk areas” are also the ones that may hold the biggest rewards, such as healthcare and mobility. All in all, regulation will be most beneficial if it’s use-case specific – AI is not “good or bad” per se, rather, it’s more about how it’s implemented.
“If you regulate too much in the short term, I think it will hinder a lot of AI innovation, a lot of people then will not build applications in the high-risk areas because there’s too much uncertainty,” Rothe warns. “Uncertainty is bad for investors, for talent – and they won’t build applications at all, or in areas that aren’t impactful.”
“So I think we actually want people developing AI systems in high-risk areas because they’re also high reward… so that should be rather encouraged.”
Major Change Brings Both Challenges and Opportunities
Aside from nurturing its own AI projects and making contributions to legislative discussions, Merantix Momentum – the platform’s research & design subsidiary which supports small and medium-sized businesses that want to streamline and enhance their processes with AI – has given Rothe insight into challenges that arise alongside commercial AI adoption.
“We’ve seen a few places where there’s been tremendous impact on a business in a positive way, but I think it’s super important that a business [doesn’t] just jump too quickly into it with some use case that sounds sexy but doesn’t have the right ROI,” he warned.
While being appropriately cautious with implementation is crucial, “aligning all the internal stakeholders” can be challenging, Rothe told Tech.co. It’s common, for instance, to find that the valuable data you need to implement an AI solution sits in one team, yet those that actually need that data are located elsewhere in the business.
Add the priorities of legal compliance and internal technical teams into the mix, and you can start to understand why getting everyone on the same page often proves to be the toughest task.
Encouraging your staff to use the AI tools already at their fingertips like ChatGPT, always helps – as does pinpointing the areas of your business where AI will truly be most impactful using different frameworks.
Rothe says he refers to two “frameworks” when considering commercial AI adoption – a work-centric approach and a data-centric approach – which is useful for identifying the areas where AI can have a positive impact.
The former involves analyzing your existing workflows for improvements, while the latter involves reviewing your data siloes to see whether you’re leveraging your data in the most productive way.
AI Will Impact Every Industry
With both harnessing data efficiently and streamlining workflows relevant to the vast majority of companies in a competitive, modern economy, it’s hard to envisage any businesses, departments, and even roles that will be left unscathed by mass AI adoption.
“I think [AI] will affect every single industry,” Rothe says. “And I think about it more in terms of business functions. So I think it will affect marketing, it will affect legal, it will affect finance, it will affect customer support, it will affect recruiting. All these functions that you have in most businesses will be affected – so that’s probably any big company or startup that has these functions, and these departments will look very different.”
“ChatGPT in some sense is very generic”, he continued, “I think there will be some even better ChatGPTs in the future, but I think what we will see is a lot of vertical AI plays.”
“You will have 50 different AI tools in a big company, all in different departments… they will all be integrated deeply in the workflow and make it very efficient” – Dr. Rasmus Rothe, Merantix Co-Founder.
It would be unwise to be too drawn in by the idea that we need AI to be perfect to implement it in various sectors and industries. After all, humans make mistakes all the time.
“I think we should use the same standards for AI, not double standards,” Rothe said earlier on in our discussion. “Maybe the flaws are slightly different, and yes, maybe ChatGPT won’t give perfect answers, but it’s pretty good at most tasks. So it’s then more, like, your choice where to use it and how much to trust the result.”
Everything from project management tools, to even website builders, are now jumping on AI to help customers create and save time.
For Rothe, the noticeable gulf in ability between the new, improved GPT-4 language model and its predecessor GPT-3.5 is a sign that it’s “only a matter of time” before these general-use AI tools become extremely competent at completing an even wider range of tasks. What’s more, just this week, Google DeepMind CEO Demis Hassibis claimed that its next language model will eclipse ChatGPT.
Opportunities are going to keep on arising around artificial intelligence, and challenges will gradually be overcome by companies like Merantix, focusing on specific, high-impact business use cases and verticals. Ensuring your business is agile enough to not just adapt to these monumental changes, but also embrace them, is going to become central to growth and success.
