Content Copyright © 2023 Bloor. All Rights Reserved.
Also posted on: Bloor blogs
ChatGPT hit the IT world like a flaming asteroid, injuring and maybe even destroying some previously well-preserved species of dinosaurs. The first shockwave that spread across the technology world came with the rapid uptake of this new “AI toy” by huge populations of web users. Suddenly, there were millions, then tens of millions, and at the moment I’m writing this, there are over a hundred million GAI users.
Nearly all of them were using ChatGPT as a personal workhorse. Why? It was a better search engine than Google ever provided. It could compose emails for you, letters (job applications anyone), write CVs, blogs, poetry, bedtime stories, articles, reviews, do your homework, help you use software products, and translate between languages, and to be honest, much more.
Accountants, authors, teachers, lawyers, doctors, risk assessors, project managers, planners, researchers, web developers, and programmers all found productive uses for it. It was a productivity gift. There were even murmurs that it would take away thousands of jobs – and maybe it will.
The Corporate Uses of GAI
All of that is marvelous. A whole new productivity app, more useful than the ubiquitous spreadsheet, had been dropped into the laps of users.
But the GAI efforts of the technology world don’t care much for that. They are focused like a laser beam on how to integrate GAI with their technology, to the greater glory of their customers.
So what’s happening in this world? On the one hand, you could say that most vendors (particularly database vendors and AI vendors) are doing the obvious thing – finding ways to integrate the sparkling new technology with their particular workhorse.
One of the most promising areas of integration is between vector databases and GAI products because vector databases can be well suited for some GAI applications.
So vector databases have become prominent. It’s not because they are a new technology per se; Elasticsearch can claim to be a vector database, but it’s not a pure vector database. The first pure vector database was Pinecone, launched in January 2021, built on top of Amazon Web Services (AWS). Others have emerged: Milvus, Vectorwise, and Aneka. A vector database is a type of database optimized for storing and querying vector data – that is data that can be represented as a vector, which is a list of numbers. Text data can be effectively stored in this manner.
The sudden popularity of vector databases is because much of the data that an LLM might wish to examine can be efficiently represented as vectors. So there are a number of database vendors that are already integrating their products with GAI products. For example, Google is pushing BigQuery ML, which is not a pure vector database – it’s an AI database with some vector capabilities. The same is true for Amazon’s Neptune, which is a graph database with vector capabilities. Oracle, Microsoft (which is making a big investment in GAI), Oracle, and IBM are also cozying up with vectors. Streaming-cloud database vendors like SingleStore and AI vendors Dataiku are also involved, both of whom incidentally were quick off the mark.
You can think of this as a bit of a technology race, although perhaps “race” is the wrong metaphor as we do not yet know what the endpoint is.
What we do know for sure is that if LLMs can query and provide insights into all the existing corporate data, there is probably going to be a productivity and knowledge generation bonanza that follows. The expectation is that GAI will have a significant impact in the following areas:
- Product development: GAI will be used to generate new product ideas, design prototypes, and even write marketing copy. Time and money will be saved on product development and bringing new products to market.
- Customer service: GAI is already being used to create chatbots that can answer customer questions and resolve issues. It can work 24/7 and free up customer service staff to focus on complex issues.
- Marketing: GAI is the marketer’s dream. It can be used to generate personalized marketing messages, create targeted ad campaigns and social media campaigns, and even write product descriptions. Ultimately, it means more bang for the buck.
- Risk management: GAI can be used to identify and assess risks, develop risk mitigation strategies, and even predict future events. This will help companies make better decisions and protect themselves from financial losses.
- Fraud detection: GAI can be used to detect fraudulent activity, such as credit card fraud or insurance fraud. This can help companies protect their customers and their bottom line.
- Content creation: GAI can be used to create blog posts, articles, social media posts, and even marketing materials. This can help companies save time and money on content creation and produce high-quality content that is more engaging and informative for their target audience.
- Data analysis: GAI can be used to analyze large datasets and identify patterns and trends. This can help companies make better decisions, improve their products and services, and identify new opportunities.
- Research and development: GAI can be used to conduct research and development on new products, services, and technologies. This can help companies stay ahead of the competition and bring new innovations to market.
And that, by the way, is just scratching the surface.
Unfortunately, the corporate use of LLMs is not entirely simple. There are a host of issues, including:
- Whether to use an in-house LLM.
- Which LLM to use, and why.
- The unreliability of some LLM responses.
- Regulatory compliance (in many areas, but personal information and health data are chief among them).
If you want to know about this topic in-depth, read this ebook which I co-authored.