 Scott Orlosky has over 25 years of experience in marketing, sales, and application support in a B2B environment. Scott’s career has involved the application of technology solutions to a variety of manufacturing and customer support issues. Scott is passionate about customer service as a strategic core value for business success.
|
Artificial Intelligence in BusinessLike a lot of “Overnight Successes”, what we call Artificial Intelligence today was 85 years in the making. In the 1940’s with the advent of digital computing came the idea that an electronic brain was a possibility. In 1956 Dartmouth the field of Artificial Intelligence was founded. Millions was spent in research to make a brain as intelligent as a human one; a feat that turned out to be much harder than expected. From 1974 to 1980 work on AI dropped off dramatically and veered off into “Expert Systems”. This was the notion that, for example you can create a “library” of symptoms and associated afflictions and make it capable of diagnosing particular afflictions. And a lot of different expert systems were successfully deployed. In the early 2000’s after an “AI Winter” the industry started picking up steam again with the advent of machine learning and language models. Language-based learning involved training computers to use words, syntax and rules of language to emulate human speech. It became a super search engine connected to the whole internet and capable of very human-like responses. It also highlighted the incredible lack of common sense that humans use to navigate their way through an average day. Nonetheless, the ability to collect massive amounts of data, process them quickly and produce a result with a high probability of being correct and reasonable was enough to activate the market. One of the issues (putting aside ethical concerns) is that the process of grinding data through Large Language Models (LLM) to extract the nearly infinite queries the world needs is huge. If we look at current data sensor trends LLMs could emit the equivalent of five billion U.S. cross country flights per year. Knowing the history of AI is useful to understanding how it is being used as a tool. Since an AI engine is taught via a data dump, the training data needs to be relevant to the intended task. If the desired result is a chatbot, then the AI may have speech recognition, facial recognition and language as the source data. If the objective is text-to-image, then its library should include a wide range of visual elements, composition rules and texturing elements. In essence there is no universal AI that is optimized for “everything”. It would just be too large and hard to train. This doesn’t prevent people from working towards that solution, but that is a whole different story. The artificial intelligence that is most widely used today is a type known as narrow generative AI, meaning that it is limited in scope and that the output is generated in response to prompts. However, within these constraints all sorts of business oriented programs are available, each with their particular mix of AI elements and algorithms. In the coming months we will expand on this theme with more detail about AI applications. This means that when upgrading your software to add AI capability it is important to know the hardware and software solution that you are buying and how the vendor will handle upgrades. This latter point is important, because the LLM that underlies most AI engines is only as good as its latest update, and the information on the internet is continuously changing. To get a snapshot of where we are today, here are the top ten business applications of AI courtesy of Chat GPT, a widely available AI: 1. Chatbots & Customer Support
2. Marketing Automation
3. Sales & Lead Generation
4. Accounting & Financial Management
5. Inventory & Supply Chain Management
6. HR & Recruitment
7. Cybersecurity & Fraud Detection
8. Personalized Customer Experiences
9. Voice Assistants & Virtual Assistants
10. Data Analytics & Business Insights
Read other technology articles |