Tech Matters: A guide to the language-based AI taking your world by storm
This week’s topic will help you understand the rapid applications of language-based artificial intelligence features coming to so many consumer products, everything from grocery shopping app Instacart to Microsoft Teams. Here we’ll explore the basics of this exciting (and sometimes frustrating) technology, why companies are jumping on the AI bandwagon and what it means for you.
Currently, the San Francisco-based firm OpenAI is the leader in language-based AI applications, best known for its launch of ChatGPT last November. Early investors included a LinkedIn co-founder, Elon Musk and Microsoft. Earlier this year, Microsoft invested $10 billion on top of its original $1 million, which allowed the tech giant to be first to the market with a version of ChatGPT built into a search engine. Competitor Hugging Face (another U.S. company, named after the smiling emoji with jazz hands) is working with Amazon and China’s Baidu, referred to as the Google of China. And of course, many potential rivals worldwide are looking to get a foothold in this huge new market.
The generic term for these types of language-based AI applications is large language model or LLM. These are based on an algorithm that can recognize, summarize, translate, predict and generate text from huge datasets — think all the content on the internet.
As we drill down, the next term to know is GPT, which stands for Generative Pre-trained Transformer. The generative part refers to the creation of new content that is generated from the relevant data, but that can’t happen without the pre-training bit that allows the model to learn how to use the data set. This method is called unsupervised learning, and it lets the model learn about the dataset in a way that is similar to how we learn language.
The transformer part refers to how the model synthesizes data to understand things like context and word order. Just as a person who masters a language can guess what might come next in a sentence or paragraph or come up with new words or concepts themselves, the GPT can apply its knowledge to predict and generate content. The result is what puts the “intelligence” into AI chatbots like ChatGPT. Along with word-based output, GPT apps can generate code, propose new chemical compounds for vaccine development and create new works of art.
Early chatbots and the ones in use by many websites as on-page helpers are based on a set of pre-programmed questions and answers. There’s nothing generative about those.
It’s no surprise that the virtual meeting space is among the first to integrate generative AI technology. Microsoft already offers this technology in its Teams Premium product and Zoom has announced similar new features coming to its platform. The new features allow you to have your meeting automatically transcribed, generate a summary, provide a list of action items and send emails to the relevant participants. Teams users pay an extra $7 a month for the enhancements for a limited time; the regular price is $10 per month.
The popular grocery and other stores delivery app Instacart announced it is working on a ChatGPT-powered virtual assistant called Ask Instacart to take the work out of meal planning. According to an Instacart press release, the service will allow shoppers to ask specific questions such as “What should I have for lunch?” and “What are healthy alternatives to my favorite recipes?” The idea behind the service is not only to reduce time for meal planning, but to encourage shoppers to try new products and buy more items. Instacart said the feature will be released later this year.
General Motors announced it is working on a virtual assistant with Microsoft to be added to its vehicles. GM said its assistant could recite information found in an owners’ manual, use functions like a garage door opener code and diagnose the seriousness of a check engine light. While the automaker is in the “exploratory” stage, it shouldn’t take too long for a generative chatbot to land on your dashboard.
The rapid application of this technology is on its way to becoming ubiquitous. This is not a fad, it’s the way of the future. But remember, it’s early days. Issues will be addressed, more sophisticated features will evolve and competitors will come and go. Use this time to familiarize yourself with the process and become comfortable with generative AI.
Leslie Meredith has been writing about technology for more than a decade. As a mom of four, value, usefulness and online safety take priority. Have a question? Email Leslie at firstname.lastname@example.org.