Large language models like ChatGPT will change the economy significantly. While the last few decades have seen gradual progress in automated information systems, the latest tools suggest a different and profound shift in the economy over the next few years.
Smarter tools have been around for years. According to Amazon, people who bought this book also bought that book. It’s not a very complicated idea, but it’s a bigger step than having a bookstore clerk say, “I like a certain novel.” Netflix recommends movies based on reviews of other movies. The customer service line chatbot awkwardly asked us why we were calling. And Alexa is waiting for our command to play Jimmy Buffett.
natural language
However, modern intelligent computer programs are a big step forward. First, they are much better able to understand us when we speak or write regular English, or most other common languages. In contrast, the old customer service chatbot searched for specific words like “order,” “shipment date,” and “invoice.” The new approach is called natural language processing. Importantly, ChatGPT and its cousins are trained on the general language we speak and write, rather than on the narrow group of words pertaining to that particular program.
AI that remembers what we say
A large language model can also remember recent conversations (although there are some limitations). A user can, for example, ask a key question about business strategy to prepare for a recession, and then say, “Tell me more about her third point you just mentioned.” Older models started from scratch without knowing what the final answer was. This allows you to drill down for your specific needs. How many times have we used Google to search the web and received results that ignore key elements of our query? increase.
summarize or elaborate
A large language model can summarize long blocks of text or elaborate on short texts. Although limited, a user can, for example, submit a 1000-word essay and ask for a one-paragraph summary. Alternatively, the user can submit a single paragraph of hers and ask for a more detailed explanation of the point made in that paragraph.
Fine-tune AI for business applications
The following two differences will be very important for near future business applications. Models can be fine-tuned with specific information targeted to a specific company, industry, or process. A company manufactures washing machines and provides phone support to its customers. The customer in question probably doesn’t know the names of the various parts of the washing machine, but a large language model can help them understand what they mean. The model then utilizes past customer service call history to tell the customer what to do.
Large language models undergo initial training based on vast amounts of text collected from the internet and digitized books, which is very expensive. However, the cost of data and computational time required for fine-tuning to a specific subject is much less than initial training.
Connect to your application
Combine this with another difference: the ability to connect the new large language model to another computer program through an application programming interface (API). That washing machine company could create a dedicated chatbot to help its customers. You can also connect the AI to your customer database to see which models a particular user owns, to your inventory system to see if replacement parts are in stock, or to your service database. You can also schedule appointments using
Alternatively, companies that create social media posts can add additional training on search engine optimization and link to leading social media sites so users don’t have to copy and paste messages across platforms. You can also do AI-powered virtual assistants with APIs tweaked to the way we speak and write, connected to our emails, calendars, and phones, are coming soon to help us with the mundane tasks that take up most of our workday. It will be processed quickly.
rapid improvement
The last notable difference is the sheer pace of improvement these days. AI has been discussed for decades, but little progress has been made. Since its development in 2018, ChatGPT has come a long way. This is not the slow, gradual rate of improvement that we have seen so far. However, the future growth rate is uncertain. We may have picked the most difficult fruits, which may have smaller future profits. But AI itself can help improve AI, leading to exponential advances in capabilities. Business leaders must consider this uncertainty. The situation may change step by step as current technology is applied to a wide range of problems. Or it could change dramatically over the next few years with significant improvements in AI capabilities.
AI challenges
All AI applications will face challenges such as false AI claims, data security, people’s willingness to adapt to different ways of working, and changing laws and regulations. But some people will take risks and many problems will be overcome.
As of this writing (June 2023), writers are already using ChatGPT to edit existing drafts and create entire articles. Computer programmers help write code. Other programmers are creating “front ends” that allow people to use AI for specialized tasks. However, companies can benefit even more by fine-tuning large language models to their specific needs and adding APIs that make them easier to use for specific tasks.
By the way, all the words in this article were written by me, but when ChatGPT was asked to assess accuracy, they suggested some explanations and details, half of which I wrote in my own words. I have included it in this article. An AI expert then reviewed the draft, and he also provided two suggestions. The end result benefited from both pieces of advice.
This is just the beginning. Connecting large language models to other business activities yields great results. The model itself will continue to be improved, but there are many benefits to be gained from customizing and using the model.
follow me twitter Or LinkedIn. check out my website.
