Artificial intelligence is big news in 2023. Companies are rushing to use artificial intelligence to gain a competitive advantage. But can AI really help? Or does it simply generate a ton of substandard blog posts and meta descriptions?
ChatGPT, Bard, and other language models will undoubtedly produce tons of poor quality blog posts. But AI is entering a new phase that could create many new opportunities. IBM described the 2023 progress as “a gradual shift in AI performance and the potential to increase enterprise value.”
Understanding the developments that enabled these advances could help managers and owners of retail, e-commerce, and direct-to-consumer business use AI to their advantage.
foundation model
Ask someone how ChatGPT works. You may hear phrases like “large language models,” “generative AI,” and “vectors.” Both discuss aspects of ChatGPT and similar platforms. Another answer is that ChatGPT is a foundational model.
In the past, an AI that predicted the best-selling price of an item on an e-commerce site required training by modeling thousands, even millions, of transactions. That will get the job done, but it will take longer.
The underlying model takes the process back one step. It is trained in an unsupervised manner over a much larger set of information – the entire Internet.
This general-purpose approach resembles a virtual jack-of-all-trades, unlike traditional AI models trained for a single specialized task. It leverages its broad knowledge base to perform a range of tasks from creating human-like text to recognizing patterns in complex datasets.
Such a foundation model is flexible. Gain a foundational understanding of many topics with our first training on comprehensive and diverse data.
This foundation has the potential to be transformative, as it can be fine-tuned for specific applications, such as predicting the best-selling price of a product on an e-commerce site, with a fraction of the time, data, and resources previously required. I have.
Generation AI
The next concept to consider is the generation of code, copies, etc. Take a look at a working example on GitHub Copilot. Use OpenAI to process natural language across sections of code.
Platforms such as GitHub Copilot and ChatGPT are said to be “generative” when they create something that never existed in that form.
GitHub Copilot is a code-writing generation AI derived from the OpenAI foundation model.
Google is evolving to generate AI search results. Search engines no longer just display links to web pages or snippets of those pages. Instead, an answer containing the link will be generated.
In an e-commerce context, a seller can share a number of product specifications with a generative AI tool and ask it to create a description for those items.
Compare this to my 2016 article How to ‘Create’ a Product Description for Ecommerce, which describes a 5-step (manual) process.
In my article, I used a soup spoon as an example product and ended up with this product description.
Craving hearty chicken noodles or creamy clam chowder? This soup spoon has a large bowl for serving bisques and soups to your mouth. In fact, this soup spoon can hold about 3 times more soup than a standard tablespoon. You can eat three times as much or buy this soup spoon for a big sip.
I pasted the article into ChatGPT and asked him to learn the process and write a product description for the hammer. This is what it generated.
Ready to tackle your next DIY project? This hammer is made with a premium steel head to ensure durability. Designed to produce a powerful, high-impact strike with every swing. Get the job done right and add this trusty hammer to your toolkit today.
AI created new hammer descriptions. It combined knowledge of the underlying model with a specific process that I gave.
Generating something new is part of what makes the underlying model so flexible.
large scale language model
AI uses images, sounds, and videos to generate things. But text is one of the most important forms of generative AI for business.
Tools like ChatGPT, Google’s Bard, and Jasper AI bring us another notion of large-scale language models.
LLM uses its underlying knowledge to predict which words will follow another.
Last week, I saw the engineering director of a software company giving a short AI presentation at their company. He expertly explained his LLM.
“I’d like to talk to you about how these models work,” he began. “I’m not quite sure what I’m talking about, so think of this not as a presentation of facts, but as a string of words strung together. Each word has a meaning based on the word that precedes it.” but it’s not 100% correct.” That’s my number one goal. ”
Give “Don’t cry if you spill it…” and your LLM will probably come up with the word “milk”. Since we have the underlying model, we can guess the word.
application
Understanding underlying models, generative AI, and LLM can help you think about how artificial intelligence can create business opportunities. Therefore, we generally do not ask ChatGPT to develop products. But you can also ask us to analyze your market gaps for potential product opportunities.
