Overwhelmed by questions about ChatGPT? So were we. So, we asked ChatGPT to help us sort it out and pulled the results into this FAQ.
Our recent webinar, “The ChatGPT & AI Revolution: An Executive’s Guide,” drew an unprecedented number of participants — and generated an unprecedented number of questions. We loaded over 80 questions into ChatGPT and asked it to consolidate them into broad ChatGPT FAQ categories.
The categories and questions below are the results, with minor editing. We pulled these responses from the webinar transcript and our fellow AI experts.
AI and ChatGPT Fundamentals
AI Applications and Use Cases
ChatGPT Competitors and the Market
AI Tools and Integration
AI Policy, Ethics and Security
AI and the Work Environment
AI and ChatGPT Fundamentals
Q: What is AI, and how is it defined in the context of ChatGPT?
A: Everyone’s definition of AI is a little different. One challenge in defining AI is that everything we’re doing now is algorithm based. But compared to other deterministic systems with a predefined set of requirements or a predefined outcome, AI is really good at finding patterns that we couldn’t find ourselves because there’s too much data, too much noise. AI takes data, identifies patterns, learns from those patterns, and figures out which patterns are useful and which are not.
Within that broad category of pattern recognition, ChatGPT is an AI-powered large language model (LLM) based on GPT-4 architecture developed by OpenAI. Everyone is familiar with tools like autocorrect (type-ahead search), which predict the word you may be trying to type before you finish typing it. ChatGPT is a much more sophisticated version of that technology. It uses 1,536 dimensions for every word to identify language and create a response, not research that response.
Q: What are some ChatGPT capabilities?
A: Unlike the search engines we’re all familiar with, Chat GPT is conversational. You provide a prompt, and ChatGPT delivers a consolidated response as text (data), which is a huge productivity boost. Then, if the response isn’t quite what you were looking for, you rephrase your prompt, and ChatGPT responds again, refining its answer based on the additional context you provide.
ChatGPT’s capability to understand language in context allows it to deliver results specific to many tasks. We’ve seen ChatGPT used for everything from coding applications and games to writing Excel formulas, creating content summaries or drafting almost any kind of business communication to helping you start a business, conducting competitive analysis or even developing new products.
Q: What are some ChatGPT limitations?
A: Because ChatGPT creates responses rather than researching them, it can make mistakes or even generate false information, especially if used as a research tool. For example, in one recent case, a lawyer used ChatGPT to generate his legal brief, but ChatGPT created legal precedents that didn’t exist, complete with plaintiffs’ and defendants’ names. ChatGPT was not being evil — it provided an output that looked like a valid brief, based on its training. The lawyer’s job was to research the result.
The main point: You need experts to produce legal documents. You must know the question you are asking and validate the output before using it. If you’re not qualified to validate, find someone who is. In this example, the lawyer should have validated the cases ChatGPT provided in the brief. If he had, he would have realized they didn’t exist.
AI Applications and Use Cases
Q: Can you provide practical examples of good ChatGPT use cases?
A: People are discovering new uses for ChatGPT every day, and we are working on blogs to address use cases for specific industries and roles, like this one. But a short list of what ChatGPT can help you do might include:
- Content writing: Generate ideas, brainstorm keywords or topics, build outlines and get opposing perspectives.
- Market analysis: Compare markets, such as the Northeast vs. the Midwest, based on the factors that are important to you.
- SWOT analysis: Identify strengths, weaknesses, opportunities and threats in your business, plans or opportunities.
- Employee reviews: Write employee reviews that even include tactful, appropriate language when describing growth opportunities.
- Job descriptions: Use ChatGPT to write job descriptions based on lists of job requirements, duties and more.
- Table output: Output ChatGPT results in a tabular format, or input tabular data and have ChatGPT conduct a basic data analysis.
The results of these use cases still need human review and may not be 100 percent what you need, but imagine the time savings if ChatGPT gets you even 70 percent toward your goal in a matter of seconds.
ChatGPT Competitors and the Market
Q: Are there alternatives to ChatGPT?
A: We expect 25,000 new AI-related tools and services to be released this year alone. Google launched Google Bard, HuggingFace released HuggingChat and the opensource community has released several LLMs designed to compete with ChatGPT. Other tools provide twists on the technology, such as allowing designers to create graphics from written instructions. However, currently, the best ChatGPT-like tool is Microsoft’s Bing and its Copilot functionality.
Q: What’s Microsoft’s role with ChatGPT?
Microsoft purchased a 49 percent stake in OpenAI this year, which allowed it to implement ChatGPT into its Edge browser. Anyone with a Microsoft account using Edge can have an experience very similar to ChatGPT, with some slight differences. For example, Bing adds a counter indicating when users have entered 20 prompt iterations.
Copilot is Microsoft’s ongoing initiative to provide natural-language search using the OpenAI model within all Microsoft 365 products. For example, you could be working in Excel and ask Copilot to perform a task without leaving the spreadsheet program. Currently, only 625 companies have access to Copilot, but we anticipate explosive growth once fully deployed.
Q: Why would someone use paid ChatGPT when free ChatGPT is available?
A: While not an OpenAI salesperson, I can share that the paid ChatGPT Plus — which gives users access to ChatGPT-4 — provides access to plugins and bigger windows to work in, and it is always accessible regardless of traffic. With free ChatGPT, you may have to wait for access based on demand. That said, if you’re technically inclined or need to integrate ChatGPT with your own applications, the paid ChatGPT API (application program interface) may be cheaper in the long run. But whether you pay or not, my best advice is to get used to using these tools every chance you get to use them.
AI Tools and Integration
Q: How can you integrate ChatGPT with other tools, such as Excel spreadsheets and calendars?
A: Until Microsoft Copilot allows full integration of ChatGPT across the Microsoft 365 suite, users can access more than 250 certified plugins for ChatGPT integrations with Microsoft apps such as Excel, as well as calendars, travel planners, investment advisors and more. Outside of Microsoft, ChatGPT API gives you more integration options. For more information and to keep up with the newest plugins, visit https://openai.com/blog/chatgpt-plugins.
AI Policy, Ethics and Security
Q: What are some ChatGPT copyright infringement concerns?
A: AI copyright law is an evolving field. Because ChatGPT generates responses from information already on the web, it can be difficult to say ChatGPT output belongs to any one person, but that does not mean you are at no risk of copyright infringement if you publish ChatGPT output verbatim. Similarly, we would not advise you to take ChatGPT output and publish it under your name, your company’s name or anyone else’s. Doing so is not only unethical but could make you responsible for any inaccuracies or false information in the ChatGPT results.
Q: What measures are in place to ensure ChatGPT data security?
A: ChatGPT security begins with you and your employees. Everyone must be careful not to enter any confidential or proprietary information into ChatGPT. This is known as “leaky data.” Employees can make this mistake unknowingly, which makes governance, ethics and acceptable use policies critical.
From a technical perspective, the best solution now is for organizations to implement ChatGPT internally, within their own organizations, rather than using public ChatGPT. Doing so can help prevent the unintentional release of both private, personally identifiable information — such as personal health information — and proprietary or confidential organizational data.
ChatGPT and Microsoft’s Azure OpenAI services have also published data privacy policies regarding their use of data. Both ChatGPT and Azure OpenAI will retain your data for 30 days for abuse and misuse monitoring, but both promise not to use your data for future training.
But if you use the ChatGPT browser version, the default data privacy policy does allow the information you input to be used for future training unless you opt out. However, you may lose some of your early access special features when you do so.
Q: Are there differences between how Microsoft and ChatGPT secure data?
A: Microsoft encrypts your data when it’s in motion and when it’s at rest. The company also provides flags to alert you when your data is being used for abuse and misuse monitoring.
Microsoft also reserves the right to use your data for troubleshooting, but they will submit a request to you in a lockbox requesting permission from you to do so. You must approve the request before Microsoft can use that data for troubleshooting or debugging.
The bigger question is whether a standard set of data privacy and security policies will emerge. We currently don’t have standard policies in the U.S., which makes it even more important to put your own governance, ethics and security policies in place. We expect the European Union to pass laws more quickly on this than the U.S. and we would expect the U.S. laws to be a close approximation of what comes out of the E.U. – similar to what happened with Data Privacy Laws of GDPR and CCPA.
Q: How does AI handle bias, and what are some examples of this?
A: Humans create AI, which means ChatGPT and other forms of AI always have the potential for AI bias. For example, an employee might ask ChatGPT to help with mortgage underwriting. If ChatGPT uses a heavy concentration of data from the 1950s, it might produce loan recommendations that we could interpret as redlining. Only a human would be able to recognize the problem, and only strong governance policies would keep it from happening again.
Those policies should address fairness, transparency, privacy and non-discrimination, as well as bias that simply results in unreliable responses. As a result, employees must understand the data they’re using and how they’re using it. You should create active policies you then communicate clearly to ensure your users understand so your organization can monitor for bias in your AI output.
AI and the Work Environment
Q: What are the potential downsides of AI in the workplace, such as skill erosion?
A: Fear of AI is real. According to a recent CNBC/SurveyMonkey WorkForce survey, 24 percent of workers are worried about AI taking their jobs, and 43 percent of them think AI will significantly alter their job — they just don’t know how. Whenever you have that kind of uncertainty in the workplace, you face risks of turnover, lower productivity and decreased employee satisfaction.
Employees need to understand how AI will alter some jobs. Our job as leaders is to make sure folks are ready for those changes and that people are empowered to become AI augmented. Also, while some skills will go away with AI, you need to know which skills to keep in case AI fails.
A workforce planning initiative that looks two to three years into the future can help identify those skills and answer other questions, like, “What skills will my AI-powered workforce need to be effective? Can I upskill and reskill folks to prepare for that change?” Being ready to answer those questions is just as important as implementing and rolling out AI tooling.
Q: How can I prepare my workforce for ChatGPT and other AI tools?
A: Start with an AI literacy campaign around your tools — how to use them, what they are capable of and what they are not capable of. You should also include AI training as a part of your onboarding for any new hires.
Hands-on workshops with real use cases for your organization can help, too, provided you have a secure space where employees can work. You should also establish an internal AI community of practice and implement a robust organizational change management campaign.
Finally, you need a strong culture that embraces using AI in a safe, secure and ethical manner on an individual level. It can’t simply be a document that sits on the shelf. It must be a proactive and pervasive living policy for the organization to succeed with AI in compliance with your organizational values and the law.
Q: Can ChatGPT replace certain roles, such as a marketing team?
A: It’s more a question of augmentation, not replacement. Depending on the study cited, AI can potentially increase personal productivity from 14 percent to 40 percent, all while developing new capabilities for employees.
We still need human experts to ensure quality AI output. We will always need people for collaboration, logic, reasoning, and deep, critical thinking. It’s these things that AI can’t do that make a company successful.
In short, ChatGPT is a multiplicative force when it comes to individual productivity. It can help employees reduce their time in meetings, help them create task lists, teach new skills or topics, generate presentation outlines, draft content and even suggest images. ChatGPT can make employees more valuable, not less valuable or obsolete.
Are you ready to soar to new heights with Microsoft Copilot? In this on-demand webinar, our Microsoft AI experts will help you learn how to assess and level up your readiness for using the powerful AI tools that Copilot provides.
