Microsoft's presentation article for Microsoft 365 Copilot begins with a concise and informative text that provides an overview of what the tool does. The article also includes a video that demonstrates how Copilot can be used in everyday work scenarios.
"Copilot combines the power of large language models (LLMs) with your data in the Microsoft Graph—your calendar, emails, chats, documents, meetings, and more—and the Microsoft 365 apps to turn your words into the most powerful productivity tool on the planet. And it does so within our existing commitments to data security and privacy in the enterprise."
Microsoft, Introducing Microsoft 365 Copilot—A whole new way to work
But how does Microsoft 365 Copilot work?
Microsoft 365 Copilot is made up of three components: an LLM or large language model, Graph data, and Office tools. Essentially, the LLM serves as a general knowledge base, Graph provides specific knowledge about the company and its employees, and the Office tools allow employees to create new documents.
While I'm not an AI expert, it's important to understand what an LLM is in order to effectively use Microsoft 365 Copilot. Without this knowledge, we may make incorrect assumptions or requests that cannot be fulfilled.
In simple terms, Large Language Models are a subset of artificial intelligence trained on massive amounts of text data (in ChatGPT's case, the entire internet) to generate human-like responses to natural language inputs. These models use deep learning techniques, which involve multi-layered neural networks to analyze, process, and make predictions with complex data.
More information about Large Language Models in the following article, What is a Large Language Model (LLM)?
According to Microsoft's articles, AI-powered Large Language Models (LLMs) are trained on a large, but limited set of data. To maximize productivity in business, it's important to connect LLMs with your specific business data. This allows the LLM to better understand the context and specific language used in your business, resulting in more accurate and useful responses.
"Grounded in your business data. AI-powered LLMs are trained on a large but limited corpus of data. The key to unlocking productivity in business lies in connecting LLMs to your business data..."
Microsoft, Introducing Microsoft 365 Copilot – your copilot for work
If you're not familiar with Microsoft 365, you may not be aware of what Graph is. To put it simply, Graph serves as a gateway to all of the information within a company's Microsoft 365 tenant. If you're looking for an official definition, Microsoft's Overview of Microsoft Graph page provides a more detailed explanation:
"Microsoft Graph is the gateway to data and intelligence in Microsoft 365. It provides a unified programmability model that you can use to access the tremendous amount of data in Microsoft 365, Windows, and Enterprise Mobility + Security. Use the wealth of data in Microsoft Graph to build apps for organizations and consumers that interact with millions of users."
Microsoft, Overview of Microsoft Graph
It can't all be so nice
The benefits of Microsoft 365 Copilot are clear and will undoubtedly improve productivity and efficiency for individuals and businesses. It's a technology that is worth exploring and implementing. If you still have reservations, consider trying ChatGPT in your daily tasks to experience its capabilities firsthand.
However, I would like to address some concerns I have about the introduction of this technology. Although I believe in its potential and the positive impact it will have on the way we work, it's important to remain critical and acknowledge its potential weaknesses.
I have categorized the concerns into three parts. The first part deals with queries related to LLM, the second part provides technical insights about Graph, and the last part covers general questions.
1. Large Language Model questions
What type of text is used in the LLM's training data?
The training data for LLMs consists of a vast and varied collection of documents, texts, and writings. However, it may be important to understand the topics present in the corpus and their respective percentages. Although it may not be possible to provide information on all topics and their percentages, this information could affect the accuracy of the LLM's responses. Additionally, it may be useful to know the time frame of the corpus. For example, could the LLM provide accurate responses about ancient Rome or the black plague? Similar questions can be asked about other locations, nations, countries, periods, and languages.
Will companies be able to decide which one they want to use?
I don't know if this could be the case but it may be interesting to know if the AI may have been trained with different texts. Additionally, can a company add its own texts and documents to the corpus used for the LLM? Currently, it seems that company data is combined with the existing corpus in Graph to provide answers.
Can AI be updated with more texts and documents in the future?
Although I am not sure about the feasibility of updating an AI, it is an intriguing concept worth considering. It is common for products to receive updates, and perhaps the current version of Microsoft 365 Copilot is just the beginning. Who knows when version 2.0 will be released?
Is it possible for the AI to generate identical text patterns consistently over time, given that the textual corpus is limited?
For instance, if I request Copilot to generate an introduction for a proposal document, could it provide the same introduction for a different proposal after a certain period of time? While I understand this could happen, I am unsure whether this is acceptable or requires improvement.
2. Graph questions
What will be the impact of the increased workload and new requests on the data centers due to the integration of business knowledge into responses through Microsoft 365 Copilot?
Considering that Microsoft 365 Copilot is a new tool that utilizes Graph to incorporate business knowledge into its responses, it is likely that the number of requests made to Graph will significantly increase, resulting in a higher workload. I am curious about how the data centers will be impacted by this increase in hundreds and thousands of new requests. However, I trust that Microsoft has already taken this into account as the success of Microsoft 365 Copilot depends on properly sizing everything.
How much power will be needed to handle this increase in activity?
I assume that Microsoft has calculated the necessary resources and identified the specific regions/data centers that require additional focus to handle the increased workload. However, will this impact be felt globally? I highly doubt it will have any effect on the planet as a whole. It is crucial to assess the number of working hours saved by using Copilot and determine if it leads to an increase or decrease in energy consumption.
How can Graph be used to facilitate automatic usage of Copilot through custom components or bots acting as end users? Specifically, what endpoints in Graph will enable this interaction and when will they become available?
In the coming months, we will see announcements from Microsoft about different ways to interact through specific endpoints.
What is the plan for integrating external data sources like ServiceNow and SAP into Copilot's responses, and how will Copilot interact with other companies' artificial intelligence?
A fascinating aspect to consider is how Copilot will be able to interact with external data sources, such as ServiceNow and SAP, as well as other applications and technologies. How will Microsoft address the challenge of incorporating information repositories from these external sources into the answers provided by Copilot? Additionally, what will be the implications of the interaction between the AI technologies of different companies once they are released? The potential answer to this question may raise concerns.
3. General questions
How much will companies need to pay for Microsoft 365 Copilot?
Rather than the cost, the relevant factor is which companies can afford to use Microsoft 365 Copilot. We can divide companies into two groups: the first group consists of those who pay for Copilot and benefit from increased productivity, giving them a competitive advantage over the second group who cannot afford it. We cannot predict what competitive advantages will exist in the future for those in the second group.
How will people who want to work for companies that use Microsoft 365 Copilot be trained to use it, and what resources will they have to become qualified?
Companies using Microsoft 365 Copilot will need to train their employees to use it. But what about people who don't know how to work with Copilot and want to work for those companies? How will they learn? Will they be self-taught or guided? It's important to consider how Copilot will evolve and what resources will be needed to become qualified.
What happens to the performance and knowledge of a worker who moves from a company that uses Copilot to a company that does not use Copilot, and how will they be able to find useful material to perform their tasks?
For me, it is a transcendental question for society as we should leave no one behind. Let's see how AI is integrated into our lives to acquire knowledge of its use in a natural way.
Will Microsoft 365 Copilot be introduced to the education sector, and what implications could this have for students who later enter the workforce and may not have access to this technology?
Will Microsoft 365 Copilot be used in universities and schools so that students can use the technology and learn it as they go about their daily tasks, and how will it affect their performance if they are hired by a company that does not use Copilot in the job market?
Microsoft has introduced a new productivity tool called Microsoft 365 Copilot, which combines the power of large language models (LLMs) with Graph data and Microsoft 365 apps to turn words into a powerful productivity tool. LLMs are a subset of artificial intelligence trained on massive amounts of text data to generate human-like responses to natural language inputs. Graph serves as a gateway to all of the information within a company's Microsoft 365 tenant.
While the benefits of Microsoft 365 Copilot are clear, it is important to acknowledge potential weaknesses and concerns related to the LLM's training data, the types of text it uses, and the ability for companies to add their own documents to the corpus used for the LLM. Overall, Microsoft 365 Copilot is a technology worth exploring and implementing, but it is important to remain critical and aware of potential limitations.
Introducing Microsoft 365 Copilot—A whole new way to work: https://www.microsoft.com/en-us/microsoft-365/blog/2023/03/16/introducing-microsoft-365-copilot-a-whole-new-way-to-work/
Introducing Microsoft 365 Copilot – your copilot for work: https://blogs.microsoft.com/blog/2023/03/16/introducing-microsoft-365-copilot-your-copilot-for-work/
Introducing Microsoft 365 Copilot — your copilot for work: https://news.microsoft.com/reinventing-productivity/
Overview of Microsoft Graph: https://learn.microsoft.com/en-us/graph/overview
What is a Large Language Model (LLM)?: https://www.mlq.ai/what-is-a-large-language-model-llm/
Don't forget to share the article with your friends and colleagues if you find it interesting, click on the heart if you like it, or click on the comments to share what you think of the article, if you would add more, or if you want to clarify any of them.
Top comments (0)