The Impact of Generative AI and Collaboration in the Microsoft Future of Work

“Instead of ‘How will AI affect work?’, the question should be ‘How do we want AI to affect work?’” This might be the most pivotal statement from the Microsoft New Future of Work Report 2023.” 

Unlike the 2022 edition, which focused on the post-pandemic practice and research of hybrid work, the keyword for the 2023 report revolves around AI. It compiles research data from 2023, outlining the impact of the generative AI boom, including changes in work patterns, current AI collaboration processes, human-AI interaction trends, and what the future office landscape might look like. 

Before diving into the study, be reminded of Microsoft's role as the financial backer of OpenAI and the integration of GPT-4-powered Copilot into its Microsoft 365 office system. From this perspective, Microsoft's AI outlook might seem a bit self-serving. Nonetheless, several points in the report surprised me and are worth following up on, especially regarding collaborative AI and knowledge dissemination. 

Here's a summary of the report in six points: 

1. LLMs boost productivity

Microsoft highlights the potential of Large Language Models (LLMs) to enhance the efficiency of information work tasks but cautions that correctness must be prioritized, with corresponding strategies to address potential inaccuracies. 

The report shows that using ChatGPT increased writing efficiency by 37%. In experiments with Boston Consulting Group (BCG), the quality of consulting projects improved by 40%, although there was a 20% chance of generating incorrect solutions in some cases. 

According to a survey of Microsoft Copilot 365 corporate users: 

73% believe Copilot makes them work faster; 85% say it helps them complete drafts more quickly; 72% agree it reduces the mental burden of repetitive or tedious tasks. 

Interestingly, the benefits of LLMs are more significant for novices or those with lower skill levels than for those with higher skill levels. In experiments with BCG, less skilled workers improved by 43%, while more skilled workers only improved by 17%. This could have implications for human resource management in the future. 

The report shows that AI has a mainly positive impact on work productivity. But the experimental results seem a bit subtle for BCG? (Source: Microsoft.com)

2. AI can enhance critical thinking.

A This may seem counterintuitive. The research suggests that viewing AI as a "challenger," rather than just a work assistant, can promote critical thinking among knowledge workers. In other words, don't take everything AI gives you at face value! 

As AI-generated content becomes more prevalent, knowledge work may lean more towards analysis and critical integration, skills more important than just searching and creating. The study mentions a typical scenario where humans choose and use AI tools to plan tasks. AI generates a draft, then humans check, evaluate, and integrate AI's output. In this model, human's depth of thought and analysis are indispensable. 

3. AI collaboration is dynamically changing. 

The report mentions the symbiotic relationship between humans and AI, noting that more and more jobs may require human oversight, with intervention when necessary. 

However, human attention is limited, so we will likely see more automated monitoring and intervention systems (using AI to manage AI?). This reminds me of the 2010 U.S. stock market crash mentioned in the book Superintelligence, which was caused by a series of problems with algorithmic trading programs. Hence, effective and smart automated monitoring and intervention are crucial for human-AI coexistence. 

4. AI plays a role in the creative process, but humans are equally important. 

LLMs are helpful with creative tasks, but human creativity is still required. The report pointed out that using generative AI efficiently relies on user self-awareness and confidence to calibrate the creative direction. Interestingly, AI can also assist humans with achieving the self-awareness and confidence needed. 

5. AI's diverse applications may revolutionize social sciences.

The report discusses the impact of LLMs in various fields, including software engineering, education, and social sciences. Social sciences surprised me. 

AI can help analyze social science data (mainly economics), and humans can use its text capabilities for real-time understanding via conversations. AI can also be a tool for interviews and surveys and enable those without statistical training to handle data. 

This, however, raises some questions: How do we statistically analyze survey data generated by LLMs? How do we validate the effectiveness of analyses based on LLM-synthesized data? And how do we combine data from humans with that from LLMs? These questions are reminiscent of the discussions around electronic voting, where many people question the verification process of electronic votes compared to manual counting. 

AI has triggered changes in the field of social sciences and also brought new problems. (Source: Microsoft.com)

6. AI has a strong ability to integrate knowledge fragments, potentially changing knowledge dissemination.

Traditionally, knowledge is stored in books and documents, but often, it is scattered in conversations. Nowadays, more and more knowledge is transmitted and received in conversational form, whether between people through digital media or between people and LLMs. LLMs can utilize this conversational knowledge. 

The process of knowledge creation to circulation involves several steps, from turning discoveries into written works and publication to dissemination through publishers and the media. Although LLMs can significantly shorten the process, this could greatly impact the news and publishing industries. In a January U.S. Senate hearing, media experts and scholars discussed the impact of generative AI on the news industry, and Danielle Coffey of the News/Media Alliance pointed out that the media has no business model in the current AI ecosystem. 

The scattered fragments of knowledge in our conversations can be captured and integrated through AI. But I wonder if AI ever gets tired of hearing trash talk? (Source: Microsoft.com)

Although the Microsoft report is optimistic about the application of AI in work and knowledge applications, we must not forget the context in which the study was produced. In the foreseeable future, no one can remain on the sideline of the AI revolution. To paraphrase Superintelligence, if the development of AI is a fast-moving train, "it won't stop at the human station, not even slow down. It's very likely to pass by without stopping." Since we're forced on board, the biggest challenge is how we strive to hold on and adapt to the speed of change.