GPT is an Assistant
How to use GPT as your personal assistant.
It used to be that the only people who could afford personal assistants were well-known politicians, businessmen, and expensive lawyers. Many discussions around the use of Large Language Models like GPT revolve around how GPT replaces humans but this isn’t quite true. GPT is a labor saver, it can save hundreds of labor hours normally reserved for personal assistants and secretaries. Writers can use GPT to sketch out ideas, proofread, or for character development. Historians can use it for historical research online and compare the credibility of sources. Artists can use it with stable diffusion models like MidJourney or DALL-E to generate either complete artwork or just draw out basic ideas.
It used to be that in order to do all the research and draft an article you needed to either do it yourself or hire a secretary to do it for you. Now you can do that yourself with the help of GPT and save countless hours. In software and programming, models like Codex and even GPT-4 can be used to translate one programming language to another. The tedious work of transpilation of codebases can be done much quicker and with fewer errors. The benchmarks on the latest model GPT-4 are astounding and even if there are a few errors the results are quite good. It even performed in the top 90th percentile on the SAT math section surpassing many students who take it.
As an example of how we can use GPT to research. We can ask GPT-4 to help research AI issues that have come up recently using its ability to do web searches on Bing. This is a new feature of GPT-4 that is very useful in using it as a personal assistant, below is the prompt I used.
Can you do a web search to help me research real life issues that have come up with using AI in a work or educational setting?
The response GPT-4 returned is a result of its own knowledge base and a web search of recent news articles. It links back to the original articles to generate citations that will be useful if I need to link back to the original articles.
The introduction of artificial intelligence (AI) in work and educational settings has brought about a variety of real-life issues and challenges, as well as some opportunities. These issues span across several areas, including the impact on education systems, workforce dynamics, ethical considerations, and technological infrastructure requirements.
Impact on Education and Learning:
Homework Disruption: AI tools like ChatGPT can solve virtually any homework assignment, leading to a significant shift in how homework is approached and potentially diminishing its educational value.
Educational Equity: The widespread availability of AI tools, such as GPT-4, offers equal access to advanced AI models for students worldwide, potentially leveling the educational playing field.
Detection of AI-generated Content: There is a challenge in accurately identifying whether a piece of content is AI-generated, with varying success rates between different AI models.
Performance Impact: AI assistance has shown a substantial improvement in performance, with notable benefits for students who were previously lower-performing.
Workforce Changes:
Job Transformation: AI and emerging technologies are expected to transform the workforce, automating some jobs while creating new ones. This shift necessitates changes in educational focus to prepare students for future labor markets.
Lifelong Learning: As AI reshapes job markets, there’s an increasing need for lifelong learning and retraining, allowing individuals to adapt to new career pathways and job roles.
Opportunities and Challenges in Teaching:
Personalized Learning: AI can facilitate personalized learning through blended learning approaches, though implementation challenges remain. It’s crucial to support educators in effectively integrating technology into their instruction.
Resource Development: AI can create scalable resources to support large numbers of students, aiding in education, training, and career guidance.
I even used GPT to help do research for this article. It suggested ideas on how to use it as a personal assistant, for example in drafting concise emails or as a research tool to search the internet for information. I wrote the article myself and didn’t rely on GPT entirely but used it as an assistant to help me. So think of GPT as a labor-saving tool that can help finish assignments faster or complete tedious work more easily. There is a danger, if you use GPT without checking the content it generates. Especially in precise and pedantic fields like law or technical writing, it can get a lot wrong.
GPT-4 is often accurate but not perfectly so and so in many cases where precision and accuracy are important it needs to be checked over but as the models improve this issue will subside. The main issue I have with GPT-4 right now is that it is compute-constrained, so you can only do a certain number of requests in a small amount of time. However, I think as OpenAI increases its compute resources this should change and GPT will become an increasingly used tool. If GPT-4 and future versions improve their ability to compute faster and without as many limitations then it will become a very useful tool not just in work but in school and at home. So think of ChatGPT and variants like Bard as a labor-saving tool, rather than a replacement. That way you don’t have to spend as much time in the office.
