GPT-4 will appear next week - and it will be multimodal, Microsoft Germany says

The release of GPT-4 is not far off, as mentioned by Microsoft Germany CTO Andreas Braun at the March 9, 2023 AI launch event.

GPT-4 will appear next week: at a roughly hour-long hybrid information event called "AI in Focus - Digital Kickoff" on March 9, 2023, four Microsoft Germany employees presented in detail Large Language Models (LLMs) like the GPT series as a disruptive force for companies and their Azure-OpenAI proposal. The event was held in German and was attended by the news publication Heise. Andreas Braun, CTO of Microsoft Germany and head of the Data & AI STU, mentioned in passing the imminent release of GPT-4. The fact that Microsoft is finalizing multimodality with OpenAI should no longer be a secret after the release of Kosmos-1 in early March.

"We will introduce GPT-4 next week."

"We'll be introducing GPT-4 next week, and there we'll have multimodal models that will offer very different capabilities - video, for example," Brown said. The CTO called LLMs a "game changer" because they teach machines to understand natural language, which then understand in a statistical way what was previously only readable and comprehensible to humans. Meanwhile, the technology has advanced so far that it essentially "works in all languages": you can ask a question in German and get an answer in Italian. With multimodality, Microsoft(-OpenAI) will "make models comprehensive".

Undermining and "destroying the good old days"

Brown was joined by Microsoft Germany CEO Marianne Janik, who spoke about disrupting companies with AI. Janik emphasized the potential of AI to create value and spoke of a tipping point in time - the current development of AI and ChatGPT is an "iPhone moment." It's not about replacing jobs, she said, but about doing repetitive tasks differently than before. One point that is often forgotten in public discussion is that "we in Germany still have a lot of legacy in our companies" and "preserve old treasures over the years."

Disruption doesn't necessarily mean job losses. It will take "a lot of specialists to make the use of AI beneficial," Janik emphasized. Traditional job descriptions are now changing, and exciting new professions are emerging as a result of being enriched with new capabilities. She recommends that companies form internal "competence centers" that can train employees in the use of AI and pool ideas for projects. In doing so, "consideration should be given to good old-fashioned migration."

Additionally, the CEO emphasized that Microsoft does not use customer data to train models (which, however, does not, or at least did not, apply to their research partner OpenAI under the ChatGPT policy). Janik spoke of "democratization" - by which she admittedly meant only the immediate use of the models within Microsoft's product line, particularly their wide availability through the integration of AI into the Azure platform, Outlook and Teams.