GPT 4 Could Be Released Next Week and Will Include Video Generation

Microsoft Germany employees said that GPT-4 is coming next week: at an approximately one-hour hybrid information event entitled ” AI in Focus – Digital Kickoff ” on 9 March 2023. Four Microsoft Germany employees presented Large Language Models (LLM) like GPT series as a disruptive force for companies and their Azure-OpenAI offering in detail.

“We will introduce GPT-4 next week, there we will have multimodal models that will offer completely different possibilities – for example videos,” Braun said. The CTO called LLM a “game changer” because they teach machines to understand natural language, which then understand in a statistical way what was previously only readable and understandable by humans. In the meantime, the technology has come so far that it basically “works in all languages”: You can ask a question in German and get an answer in Italian. With multimodality, Microsoft(-OpenAI) wants “make the models comprehensive”.

Braun was joined by the CEO of Microsoft Germany, Marianne Janik, who spoke across the board about disruption through AI in companies.

Clemens Siebler described the chat system handling speech-to-text telephone calls. Those calls could be recorded and the agents of a call center would no longer have to manually summarize and type in the content. This could save 500 working hours a day for a large Microsoft customer in the Netherlands, which receives 30,000 calls a day. And the prototype for the project was created within two hours, a single developer implemented the project in a fortnight (plus further time for final implementation). The three most common use cases are answering questions on company knowledge that is only accessible to employees, AI-assisted document processing and semi-automation by processing spoken language in the call and response centre.

11 thoughts on “GPT 4 Could Be Released Next Week and Will Include Video Generation”

    • Because it is not so open and nonprofit anymore. Microsoft wants to make money.
      If they invested extra 10 billion into it they own 49 %. The rest are fragmented shareholders easy to control by them and have majority.

      Microsoft see it as their next “cash cow”.

  1. I just ask myself how will we control AI? Because with internet everything is connected, AI can hide on almost every computer and use resources from others. Even simple bitcoin virus can steal your GPU power.
    If they code in rules, AI could probably easily bypass them. One programmer can’t understand millions of lines of computer code. Machine learning networks probably use millions of lines of it. That means the people who are using them don’t have enough clue, what is really happening under the hood.

    • The goal is not to control AI but to align it’s values and goals with ours. AI is a tool that will allow us to perform calculations beyond the limits of human beings. Revolutions in protein folding, physics, and medicine are already happening.

      • But isn’t “aligning values and goals” to our own kind of control? What if AI disagrees and says f*** this puny humans I won’t work for them?
        Lot of people are fair, but I don’t believe all with considerable power don’t look how to use AI for their own gain.

  2. If it’s above GPT3.5 capabilities, it already is a revolution.

    If it’s multi-modal and works as predicted, it could scratch AGI. And it could soon be with an API so anyone can come and play.

    We have yet to see anything from Google with broad accessibility, but PALM-E seems to be going into the same multimodal territory.

    By the way: Midjourney is getting fully photorealistic soon.

    • What would be the best image generation tool to generate graphics for work instructions? Like… the minimum distance an employee needs to keep from an excavator when excavating a trench

      • Good point. Image generators still aren’t that good to take precise instructions like that and turn them into images.

        In fact, image generators have a ~890M parameters, while current LLMs have 176 billions. 2 orders or magnitude less.

        That works for us because we are don’t mind some small errors and flights of fancy on an fantasy or publicity image. But a good image generator should be capable of taking very precise commands and perform them flawlessly.

        Humans can do it, and AIs should have to do it as well, in order to be really usable for any purpose.

Comments are closed.