Google and Microsoft are bringing AI to office applications. How it can increase productivity for us and cybercriminals

Google and Microsoft are on a mission to take the toll from computing, by bringing next-generation AI tools as add-ons to existing services.

On March 16th, Microsoft announced An AI-powered system called Copilot will soon be introduced to its 365 suite of apps including Word, Excel, PowerPoint, Outlook, and Teams.

The news came about two days later Google published a blog explaining its plans to embed AI in its Workspace applications, such as Docs, Sheets, Slides, Meet, and Chat.

Collectively, millions of people use these apps every day. Augmenting them with AI can dramatically increase productivity—as long as security isn’t a big consideration.

The birth of a new generation of AI

Until recently, AI was mainly used for classification and recognition tasks, such as number plate recognition using traffic cameras.

Innovative AI allows users create new content, by applying deep learning algorithms to big data. ChatGPT And DALL-Eamong others, has taken the world by storm.

Now, Microsoft and Google have found a more concrete way to bring creative AI into our offices and classrooms.

Like other generalist AI tools, Copilot and Workspace AI are built on large language models (LLMs) trained on massive amounts of data. Through this training, systems have “learned” many rules and patterns that can be applied to new content and contexts.

Microsoft’s Copilot is being tested with just 20 customers, with details on availability and pricing published “in the coming months”.

Copilot will be integrated across apps to help speed up tedious or repetitive tasks. For example, it will:

• help users write, edit and summarize Word documents
• Turn an idea or summary into a full PowerPoint presentation
• Identify data trends in Excel and quickly create visualizations
• “aggregate and manage” your Outlook inbox
• provides real-time summary of Team meetings
• gather data from documents, presentations, emails, calendars, notes, and contacts to help write emails and summarize conversations.

Assuming it does these tasks efficiently, Copilot would be a major upgrade from Microsoft’s original Office Assistant, Clippy.

Google’s AI Workspace will provide similar capabilities for paying subscribers.

What’s under the hood?

Microsoft described Copilot as a “sophisticated orchestration and processing engine that works behind the scenes to combine the power of LLMs, including GPT-4 […].”

We don’t know specifically what data the GPT-4 itself was trained on, only that it was a lot of data that was pulled from the internet and licensed, according to openAI.

Artificial Intelligence in Google’s Workspace is built on Palm (Road Language Model), which trained on a combination of books, Wikipedia articles, articles, source code, filtered web pages, and social media chats.

Both systems are integrated into the existing cloud infrastructure. This means that all data to which they are applied will be online and stored in the company’s servers.

The tools will need full access to relevant content to provide contextual feedback. For example, Copilot couldn’t distill a 16-page Word document into a bulleted page without first analyzing the text.

This begs the question: is the user’s information used to train the underlying models?

Regarding this point, Microsoft said, “Copilot’s large language models are not trained on customer content or on individual prompts.”

Google said“[…] private data is kept private and not used in the broader foundational model training data store.”

These statements suggest that the 16-page document itself will not be used to train the algorithms. Instead, Copilot and Workspace AI will process the data in real time.

Given the rush to develop such AI tools, there may be a temptation to train such tools on “real” customer-specific data in the future. For now, however, it seems that this is being explicitly ruled out.

Usability concerns

As many have noted after the release of ChatGPT, text-based generic AI tools Tend algorithm error. These concerns will extend to new Google and Microsoft tools.

The outputs of general AI tools may contain inaccuracies and biases. Microsoft’s own Bing Chatbot, also running on GPT-4, is here under fire earlier this year for making outrageous claims.

Trends occur when large volumes of data are processed without proper selection or understanding of training data and without proper supervision over training processes.

For example, most online content is written in English—likely main language spoken by (mostly white and male) who are developing AI tools. This fundamental bias can affect Writing style and language constructs are understood by and then copied by AI-driven systems.

Currently, it is difficult to say exactly how the issue of bias can appear in Copilot or Workspace AI. As an example, systems may not work well for people in non-English speaking countries or with diverse English speaking patterns.

security concerns

One major flaw in Microsoft and Google’s AI tools is that they can make it a lot easier for cybercriminals to get their victims’ blood.

While in the past criminals might need to scour hundreds of files or emails for specific data, they can now use AI-powered features to quickly collate and extract what they’re looking for. need.

Also, since there is no indication that offline versions are available so far, anyone who wants to use these systems will have to upload the relevant content online. Data uploaded online has a higher risk of being breached than data stored solely on your computer or phone.

Finally, from a privacy perspective, it’s not particularly exciting to see more ways in which the world’s largest corporations can collect and aggregate our data.

This post was reposted from Conversation under a Creative Commons license. Read original article.