HomeIndustriesAI agents, multimodal Phi-3, presented at Microsoft Build 2024

AI agents, multimodal Phi-3, presented at Microsoft Build 2024

Satya Nadella used his keynote on the primary day of Microsoft's Build Developer Conference to announce some exciting recent AI developments that can soon be generally available.

Microsoft Build is an annual conference where developers can see the most recent developments in Windows 11 and Microsoft 365. On the primary day some interesting generative AI tools were presented.

team co-pilot

In 2023, Microsoft released its co-pilot Chatbot that gives real-time intelligent support when you work with Microsoft 365 tools similar to Word, Excel, PowerPoint, Outlook or Teams.

Nadella announced that it could be getting a major AI upgrade with Team co-pilot. team co-pilot expands co-pilot From being a single personal assistant to being a part of a team that improves collaboration and project management.

If you’re employed as a part of a team with Microsoft Teams, Microsoft Loop, or Microsoft Planner, Team co-pilot can facilitate meetings by managing the agenda and taking notes. It can highlight necessary information, track motion items, and address unresolved issues.

It may even act as a project manager, assigning tasks, tracking deadlines, and notifying team members when their input is required.

Custom Copilot Agents

Microsoft co-pilot Studio permits you to create custom co-pilots who act as agents and work independently after you give them instructions.

Easily describe what the agent should do using a natural language prompt, then deploy it across multiple platforms.

According to Microsoft, these agents can:

  • Automate tedious business processes
  • Reasons for actions and user input
  • Use memory to include context
  • Learning based on user feedback
  • Record exception requests and ask for help.

One example of the worth an agent like this might provide is an “order taker” copilot, which Microsoft says could “handle the end-to-end order achievement process – from order acceptance to order achievement to creating intelligent recommendations and substituting out-of-stock items to shipping to the client.”

This feature permits you to create virtual employees to perform menial tasks similar to monitoring emails, data entry, or other repetitive tasks without increasing your headcount.

Phi 3 vision

Microsoft has added a multimodal model with 4.2B parameters to its Phi-3 family of small language models (SLMs). Phi-3 Vision is a low-cost, low-latency model that has audio and visual capabilities and a 128,000 context window.

These smaller models goal on-device solutions where speed, cost, computational power, and web connectivity limitations make larger models impractical. The Phi-3 SLMs display superior reasoning capabilities and outperform several larger models.

Enabling multimodal pondering on the device opens up exciting applications in healthcare, education and agriculture, particularly for rural areas without web connectivity.

You can try it Phi-3 Vision here. It's great for analyzing images, extracting text, and even translating.

Phi-3 vision benchmark results in comparison with other AI models. Source: Microsoft

Advanced Insertion

Windows 11 now offers a wiser method to copy and paste. The recent Advanced Paste feature gives you more options for data you copy to the clipboard. When you press Windows Key + Shift + V, you'll see options to stick as plain text, as Markdown, or as JSON.

You may enter an outline of how the copied text needs to be processed before pasting.

You need one OpenAI API key and balance in your account to make use of this feature. It just saves you the difficulty of pasting the text ChatGPT and asks it to format it there before copying and pasting it back into your document.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Must Read