At Rise of AI 2023 in Berlin, Fabian Westerheide, founder and CEO of Rise of AI and founding partner at Asgard Capital, opens the conference with a state-of-AI keynote, preceded by a scripted video of an AI avatar generated with ChatGPT and a text-to-video tool, standing in for host Veronica Westerheide. He welcomes over 1,200 virtual attendees from 57 nations and argues 2023's task is working with existing foundation models and generative AI rather than chasing new ones, with Europe finding its own vertical, industry-specific applications instead of competing on consumer chatbots. He frames AI as augmenting rather than replacing jobs, calls for sustainable, energy-conscious AI development, and discusses the EU AI Act, warning it could have made a tool like ChatGPT illegal in Europe. He predicts AI will widen the wealth gap and raise social tension, and argues Germany's opportunity lies in precise B2B and industrial AI rather than consumer products, since OpenAI has already won that race. The keynote closes with a conversation with Lara, host of the Applied AI stage, on the AI Act's burden on startups, data privacy risks of using ChatGPT at work, and Westerheide's expectation that AI will reach every microchip-equipped machine by 2045.
Key takeaways
- Fabian Westerheide opened the keynote with an AI-generated avatar video, written by ChatGPT and rendered with a text-to-video tool.
- Over 1,200 virtual attendees from 57 nations joined the 2023 conference live, up from mostly curious non-experts in prior years.
- Westerheide argues Germany has already lost the consumer AI race to ChatGPT and should focus on precise, industrial B2B applications instead.
- He estimates 40 to 50% of startups consider themselves high-risk under the proposed EU AI Act, which he says threatens the growing ecosystem.
- He expects AI to widen the wealth gap and cause rising social tension, pointing to Uber driver protests at the Noah conference as a precedent.
Detailed notes
An AI-generated opening and welcome
The conference opens with a video of an AI avatar delivering a short scripted keynote in Veronica Westerheide's place, introduced as an outsourcing joke tied to the European AI Act. Fabian Westerheide explains afterward that he wrote no script himself: he gave ChatGPT bullet points and asked it to write a funny keynote, then generated the video with a text-to-video tool, and says he tested it the day before on people unfamiliar with AI who believed it was a real person. He welcomes the audience in Berlin as the people building, investing in and running AI, and reports more than 1,200 virtual attendees joining live from over 57 nations, up from a mostly curious, non-expert virtual audience in earlier years. He describes a gamified "green room" and leaderboard for online attendees, with a free in-person ticket for the most active virtual participant.
State of AI 2023: working with what exists, finding a European role
Westerheide says the current wave of large language models, foundation models and generative AI is only the beginning of the next wave, with a live contest over whether one dominant model provider, naming OpenAI, will win a near monopoly or whether niche players will emerge. He argues the most important work for the coming year is applying and improving what already exists, especially generative AI for vertical, industry-specific use cases, and that Europe needs to find its own models and base for this rather than building only on systems whose data and biases it does not control. On jobs, he rejects the framing that AI will replace people, describing it instead as a tool that augments human work, and says the priority is educating and reskilling people to use AI tools as time-savers rather than fearing them.
Sustainable AI and the cost of scale
Westerheide argues against simply building ever-larger models with more data, calling for "green AI" that accounts for energy use from the start. He cites the digital sector as one of the largest producers of CO2 among industries, and references a case in Norway where a country could not produce ammunition because a service like TikTok had preferential access to energy capacity, framing it as a choice between "TikTok or munition for troops." He says several talks later in the program address sustainable and green AI directly.
The EU AI Act as opportunity and threat
Westerheide traces the EU AI Act from the 2018 European AI strategy and 2020 white paper to ongoing negotiations, noting the law is being written for technology that does not yet exist and will set boundaries for the next decade. He says that under the law as discussed, a tool like ChatGPT could have been illegal in Europe, which he calls frustrating, but also sees an opportunity: Europe could brand itself around safe, trusted, reliable AI the way German cars are branded, if the law does not crush the continent's growing AI ecosystem. He urges the audience to use their influence in the law's final stages so it does not disproportionately burden smaller companies.
Wealth gap, social tension and the ecosystem's role
Westerheide predicts AI will widen the wealth gap, since companies that successfully deploy it, citing Google's and Microsoft's continued market dominance, need fewer people for the same output, reducing income, taxes and job security for others. He expects rising social tension and eventual anti-AI demonstrations, recalling protests by Uber drivers at the Noah conference, and argues the AI community must keep explaining that AI is a tool, not an inherent threat. He frames AI as unusual among tech ecosystems because it requires tech people, policymakers, enterprises and NGOs to work together, and cites Germany's AI month, in which more than 70 different organizations ran their own events, as evidence the ecosystem can act as one. On Germany's role, he argues the consumer AI race is already lost to tools like ChatGPT, which reached 100 million users in 60 days, and that the opportunity instead lies in precise, industrial B2B AI.
Q&A highlights: conversation with Lara
In a discussion with Lara, host of the Applied AI stage, Westerheide says he did not expect ChatGPT's chatbot interface to become the mainstream breakthrough moment for AI, and that Berlin policymakers were caught off guard despite the industry warning them for years. He says the AI Act is now the biggest concern for startups, since he estimates 40 to 50% of startups consider themselves "high risk AI" under the proposed rules, and argues the ecosystem needs to build support infrastructure so entrepreneurs are not crushed by compliance costs. On end users, he says the public is already adopting tools like ChatGPT without complaint and that education systems, not users, need to adapt, while stressing that people must stay critical of AI output rather than trust it as truth, citing a CTO who says the tool saves 80% of his time but he still checks the remaining 20%. He warns that using ChatGPT at work often violates data policy, noting Samsung banned it internally after an employee leaked sensitive data through it. Asked about the future, he says AI will eventually reach every machine with a microchip and electricity, that the best AI is invisible to its users, and gives 2045 as his personal deadline for that integration to be complete.
Quotes
We can't let AI eat the world.
Fabian Westerheide
I think a good AI is invisible.
Fabian Westerheide
Questions answered in this talk
Why does Fabian Westerheide think Germany cannot compete with ChatGPT in consumer AI?
According to Fabian Westerheide at Rise of AI 2023, ChatGPT reached 100 million users in 60 days, making the consumer B2C race effectively won by American players, so he argues Germany's opportunity lies in precise, industrial B2B AI instead.
How does the EU AI Act affect AI startups, according to this talk?
Fabian Westerheide says at Rise of AI 2023 that he estimates 40 to 50% of startups consider themselves high-risk under the proposed EU AI Act, and calls for ecosystem support so the law's compliance burden does not crush young companies.
Is it safe to use ChatGPT with company data at work?
Fabian Westerheide warns at Rise of AI 2023 that using ChatGPT at work often trains an American system in American jurisdiction with sensitive company data, citing Samsung's decision to ban internal use after an employee leaked data through it.