Friday , 23 February 2024
Home AI: Technology, News & Trends France and Germany in AI Cash Battle: Competing for the ‘European OpenAI’

France and Germany in AI Cash Battle: Competing for the ‘European OpenAI’

87
Jonas Andrulis

Recently, German artificial intelligence startup Aleph Alpha announced the completion of a B-round financing exceeding $500 million.

Aleph Alpha’s latest financing attracted investment from prominent players, primarily originating from Germany. Leading investors include Bosch’s venture capital division, retail giant Schwarz Group, and the German AI technology hub Ipai. Co-investors include Germany’s largest publicly traded company SAP, investment firm Burda Principal Investment founded by media mogul Hubert Burda, and HPE, a spin-off from the U.S. IT giant Hewlett Packard Enterprise.

Founded in 2019, Aleph Alpha positions itself as the leader in local AI technology in Europe, akin to ‘Germany’s own OpenAI.’

However, just as the German company made its official announcement, French media reported on Mistral AI, dubbed the ‘secret project against ChatGPT’ in France, gearing up for a rumored €450 million new funding round, potentially joining the unicorn ranks.

So, in this Franco-German AI ‘gold rush,’ who will emerge as the ‘European Version of OpenAI’?

Germany’s Own OpenAI

Jonas Andrulis and Samuel Weinbach

Aleph Alpha was founded by Jonas Andrulis and Samuel Weinbach

According to publicly available information, Jonas Andrulis, the current CEO, graduated in Economic Engineering from the elite European university, Karlsruhe Institute of Technology. He held positions in renowned companies such as SAP and Deloitte. Before founding Aleph Alpha, he spent three years as a senior AI research engineer at Apple. The company’s Vice President of Technology, Samuel Weinbach, worked at Deloitte for eight years, one of the world’s top four accounting firms.

Aleph Alpha publicly states that the proceeds from this funding round will be used for fundamental model research, product capability enhancement, and the commercialization of software. The company aims to accelerate its expansion because it believes the next 12-24 months will be a crucial window for establishing cooperation with downstream customers for AI applications.

Distinguishing itself from competitors, Aleph Alpha has not formed alliances with large tech companies. For instance, Anthoropic secured $4 billion in financing from Amazon, and OpenAI previously partnered with Microsoft. Aleph Alpha emphasizes its independence.

Simultaneously, Aleph Alpha champions the concept of ‘data sovereignty,’ where data stored in a country must comply with that country’s laws. In essence, it supports the storage and processing of European AI model data in Germany or the European region, rather than other countries like the United States, to ensure data security. This concept is gaining traction among various European governments and legislative bodies.

Jonas Andrulis previously stated that for companies valuing data sovereignty, Aleph Alpha remains the ‘optimal choice.’ Tanja Rueckert, Chief Digital Officer and board member at Bosch, also noted that developing generative AI is ‘crucially important for Europe’s technological sovereignty.’

Aleph Alpha, with around 70 employees, develops what it calls ‘sovereign AI solutions’ for governments and businesses. The company positions its products in sensitive industries, including healthcare, law, finance, government, and security, with thousands of paying customers in the European region.

Currently, Aleph Alpha’s models support German, French, Spanish, Italian, and English. Its training data includes a large multilingual public document library released by the European Parliament.

Analysis shows that as early as 2021, Aleph Alpha secured €23 million in Series A funding from Earlybird VC, Lakestar, and UVC Partners. Before that, LEA Partners, 468 Capital, and Cavalry Ventures provided €5.3 million in seed funding for Aleph Alpha.

It’s worth mentioning that on June 30th of this year, Aleph Alpha raised €100 million in Series A+ funding, officially joining the ranks of German AI unicorns, with technology giants such as NVIDIA, Intel, and SAP participating in the new round of investments.

Comparable Performance to ChatGPT?

In less than five years since its establishment, Aleph Alpha has been hailed by Federal Minister for Economic Affairs Robert Habeck as an ‘incredible success story.’

So, what sets its products apart?

Researchers claim that Aleph Alpha’s AI language model ‘Luminous,’ compared with OpenAI and Meta, has a comparable accuracy:

Comparison of Luminous and Open AI and Meta Data

The average accuracy of this language model in core functions is 53.4%, with OpenAI’s ‘davinci’ model at 53.7%, Meta’s OPT at 52.2%, and Bloom at 49.3%.

In April 2022, Aleph Alpha released its self-developed Luminous series, with models having 130 billion, 300 billion, and 700 billion parameters, performing similarly to GPT-3.

These models demonstrated impressive results in various tasks such as text classification, common-sense reasoning, and natural language inference in Eleuther AI’s tests.

Currently, these three models can process text in English, German, French, Italian, and Spanish, and support image input. They are suitable for various tasks, including text/code generation, content summarization, translation, and provide API interfaces to help developers integrate the models into their applications.

The interpretability of generated content is a technical highlight of Aleph Alpha’s models. The company states that all three models come with built-in AI explanation features to help clients ensure the secure and reliable processing of their data, providing reasonable explanations for the generated content.

Additionally, users can manually change the weight distribution of various sections in the Luminous model, providing developers with insights into its processing flow by observing how these changes affect the model’s output. This allows developers to assess whether AI-generated content exhibits false or illegal behaviors.

Aleph Alpha says that with the new funding, it aims to accelerate product innovation. The company is currently developing a 300 billion parameter model called Luminous-World, aiming to rival OpenAI‘s GPT-4. Product safety and data privacy are the core cornerstones of development, providing secure and controllable AI services for businesses and individual users.

Franco-German AI in the ‘Gold Rush’

Germany, as one of the industrial giants, is home to global companies such as Siemens, BMW, Volkswagen, Bayer, and Robert Bosch. However, in the field of AI, especially generative AI, Germany lags noticeably behind other countries.
To establish their ‘OpenAI,’ they are placing their hopes on new AI unicorns like Aleph Alpha, which is why many German local institutions participated in this investment.

In a previous press conference, German Vice Chancellor and Minister of Economic Affairs and Climate Habeck considered this investment a strategic national focus. He stated, ‘The idea of having our sovereignty in AI is extremely important. Even if Europe has the best regulation, without European AI companies, the advantage is not significant.’

Due to differences between EU regulatory requirements and those of the United States, the EU also hopes to build its generative AI technology system, minimizing regulatory friction with U.S. tech companies.

Earlier this year, Jonas Andrulis mentioned, ‘In many application scenarios, Luminous is a powerful alternative and an important step toward European technological sovereignty.’

Regardless, although Aleph Alpha still has a significant gap with OpenAI, many EU officials hope that Aleph Alpha can grow to maintain the EU’s innovation position.

The reason lies in the divergence between the EU’s General Data Protection Regulation (GDPR) and the U.S. Clarifying Lawful Overseas Use of Data (CLOUD) Act. Many U.S. companies face regulatory issues in Europe.

Specifically, the EU Data Protection Board believes that data controllers processing personal data must comply with GDPR or other relevant EU regulations, which may conflict with the mandatory data transfer requirements of the CLOUD Act. The report mentions that, according to Article 48 of the GDPR, court orders from non-EU countries must be based on international agreements to be recognized within the EU. The CLOUD Act does not meet this condition.

However, Jonas Andrulis also emphasized that his company has many American employees, stating publicly that ‘I mainly want to contribute to Europe’s tech sector.’

In a horizontal comparison, Mistral AI, dubbed the ‘European version of OpenAI’ and the ‘secret project against ChatGPT’ in France, secured a €113 million seed funding round led by Lightspeed Venture Partners in May this year.

Mistral AI

More recently, insiders revealed that Mistral AI’s latest €450 million funding round is in the final stages, with investors including A16z, NVIDIA, Salesforce, among others. This suggests that, in just six months, Mistral AI’s valuation could rise to €15-20 billion, successfully joining the unicorn ranks.

In any case, Mistral AI’s new funding round has not been officially confirmed yet and is not on par with Aleph Alpha’s B-round financing exceeding $500 million.

However, despite the Franco-German AI ‘gold rush,’ the figures are still relatively modest compared to AI companies across the Atlantic. Europe’s investment in the AI field is still far from matching that of the United States.

For example, OpenAI, backed by Microsoft, raised a staggering $10 billion (approximately ¥72.84 billion). In September of this year, U.S. AI startup Anthropic secured a $4 billion (approximately ¥30 billion) investment from Amazon and, at the end of last month, received a commitment of $2 billion from Google. Inflection, an AI startup co-founded by DeepMind and LinkedIn’s founders, received $1.3 billion (approximately ¥9.47 billion) in funding.

It is foreseeable that the story of the future AI battle between Europe and the United States will continue to unfold.

Leave a comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Articles

AI deciphers burnt scroll

AI Deciphers Burnt Scroll, Reveals Ancient Greek Philosopher’s Musings on Happiness

This breakthrough has ignited the typically slow-moving world of ancient studies. In...

OpenAI sora

OpenAI Unveils New Model Sora: Generates 60-Second Videos From Text!

OpenAI is teaching AI to understand and simulate the physical world of...

AI and writing

Collaborative Creation: Will Human-ChatGPT Partnerships Shape a New Era Trend?

Recently, the novel “Tokyo Resonance Tower” by Japanese author Rie Kudan won...

EU

EU’s 27 Nations Approve Final Text of “Artificial Intelligence Act”

According to reports, representatives of the 27 EU countries recently confirmed the...