Microsoft presents its own artificial intelligence models, reducing its dependence on OpenAI. The company promises efficiency and lower computational power costs than its competitors.
UNN reports with reference to The Verge.
Details
Microsoft's artificial intelligence division recently announced its first proprietary artificial intelligence models.
Their names:
- Mai-Voice-1 AI;
- Mai-1-Preview.
The new Mai-Voice-1 speech model, the company claims, can generate a minute of audio in less than a second on a single GPU.
And Mai-1-Preview "offers a glimpse into future offerings in Copilot."
Microsoft is already using MA1-Voice-1.
Specifically, Copilot Daily:
An AI host recites the day's top news
And also for - creating podcast-style discussions that help explain topics.
As for MAI-1-preview - it was trained on approximately 15,000 Nvidia H100 GPUs.
Designed for users who need an AI model capable of following instructions and "providing useful answers to daily queries."
Recall
It was previously reported that OpenAI plans to release GPT-5. Testers note its capabilities, but consider the leap from GPT-4 less significant than from GPT-3 to GPT-4.
OpenAI, without violating confidentiality, scans and checks user conversations in ChatGPT. Particularly dangerous content related to the promotion of suicide, weapons development, harm, and other illegal activities - the company is already preparing to hand over to the police.