Google and Microsoft are on a mission to remove the grind from computing by offering next-generation AI tools as add-ons to existing services.
on March 16 Microsoft announced an AI-powered system called Copilot will soon be introduced in its 365 suite of apps, including Word, Excel, PowerPoint, Outlook, and Teams.
The news came about two days later Google published a blog explaining plans to embed AI in its Workspace apps such as Docs, Sheets, Slides, Meet, and Chat.
Collectively, millions of people use these apps every day. Bolstering them with AI could boost productivity, as long as security isn’t an afterthought.
The advent of generative AI
Until recently, AI was mainly used for categorization and identification tasks, such as recognizing a license plate using a traffic camera.
Now Microsoft and Google have found a more concrete way to bring generative AI into our offices and classrooms.
Like other generative AI tools, Copilot and Workspace AI are built on large language models (LLM) trained on massive amounts of data. Through this training, the systems have ‘learned’ many rules and patterns that can be applied to new content and contexts.
Microsoft’s Copilot is being trialled with just 20 customers, with details on availability and pricing be released “in the next months.”
Copilot is integrated into all apps to speed up boring or repetitive tasks. For example, it will:
- help users write, edit and summarize Word documents
- turn ideas or summaries into full PowerPoint presentations
- identify data trends in Excel and quickly create visualizations
- “build and manage” your Outlook inbox
- provide real-time summaries of Teams meetings
- bring together data from documents, presentations, email, calendar, notes, and contacts to compose emails and summarize chats.
Assuming it does these tasks effectively, Copilot will be a huge upgrade from Microsoft’s original Office Assistant, Clippy.
Google’s Workspace AI offers similar capabilities for this paying subscribers.
Introducing a new era for AI and #GoogleWorkspace:
✅ Draft, reply, summarize and prioritize your Gmail
✅ Brainstorm, proofread, write and rewrite in Docs
✅ Bring your creative vision to life with auto-generated images, audio, and video in Slides
And more → https://t.co/vGsTGN3w9i pic.twitter.com/XnkTWvrwgT
— Google Workspace (@GoogleWorkspace) March 14, 2023
What’s under the hood?
Microsoft described Copilot as a
advanced processing and orchestration engine that works behind the scenes to combine the power of LLMs including GPT-4 […].
We don’t know exactly what data GPT-4 itself trained on, only that it was a lot of data that was pulled from the Internet and licensed, according to Open AI.
Google’s Workspace AI is built on Palm (Pathways language model), which was trained on a combination of books, Wikipedia articles, news articles, source codes, filtered web pages, and social media conversations.
Both systems are integrated into the existing cloud infrastructure. This means that all the data they are applied to is already online and stored on company servers.
The tools will full access required to the relevant content to provide contextualized answers. For example, Copilot can’t break down a 16-page Word document into one bulleted page without first analyzing the text.
This begs the question: is the information from users used to train the underlying models?
With regard to this point, Microsoft said:
Copilot’s large language models are not trained on customer content or individual prompts.
[…] private data is kept private and not used in the broader base model training corpus.
These statements suggest that the 16-page document itself will not be used to train the algorithms. Rather, 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 the time being, however, it seems that this is explicitly excluded.
As many people noticed after the release of ChatGPT, text based generative AI tools are susceptible to algorithmic bias. These concerns will extend to the new tools from Google and Microsoft.
The output of generative AI tools can be riddled with inaccuracies and biases. Microsoft’s own Bing chatbot, which also runs on GPT-4, came along under fire earlier this year for making scandalous claims.
Bias occurs when large amounts of data are processed without proper selection or understanding of the training data and without proper supervision of the training processes.
For example, much of the online content is written in English, which is likely the main language spoken by the (usually white and male) people who develop AI tools. This underlying bias can affect the writing style and language constructs that AI-driven systems understand and then replicate.
For now, it’s hard to say exactly how bias issues might arise in Copilot or Workspace AI. For example, the systems just don’t work as effectively for people in non-English speaking countries or with different styles of English.
A major vulnerability in Microsoft’s and Google’s AI tools is that they can make it much easier for cybercriminals to drain victims.
While in the past a criminal would have to sift through hundreds of files or emails to find specific data, now they can use AI-assisted features to quickly collect and extract what they need.
Since there is no indication so far of offline versions becoming available, anyone wanting to use these systems will have to upload the relevant content online. Data uploaded online is at greater risk of being breached than data stored only on your computer or phone.
Finally, from a privacy standpoint, it’s uninspiring to see even more ways for the world’s largest companies to collect and synthesize our data.