AI Series Part II
By Bob Deakin
As the trending buzzword, AI has taken on added muscle as the Holiday season approaches. If it has artificial intelligence, it is brighter than other products or services, right? So what is AI and how is it created?
We live in an acronym-dominated world, and business professionals are the kings. If you’re in a meeting and you don’t hear API, CRM, KPI, CTA, ROI, SEO, UI or YoY, someone missed the invite.
Acronyms are the new buzzwords. Only a decade ago, terms like actionable, holistic, ideate, omnichannel, paradigm shift, rockstar, value-add, or anything “centric” were obligatory in commercials and at marketing meetings. If you said “AI” you were talking about NBA star Allen Iverson.
What Is AI?
You can build your own AI. However, you need data scientists and software engineers to commence with the labor-intensive tasks of training machine learning models. You can also purchase services from a dedicated AI vendor to implement into your system. Additionally, you may be using a software vendor that has new AI capabilities in which to upgrade.
A dedicated AI vendor can navigate through the newest technologies and better integrate them into your existing IT environment. Professor of Management and IT at Babson College, Thomas Davenport, recently explained in Forbes that developing AI can cost millions. Training a machine learning algorithm to do what specialized AI vendors already do can also waste valuable time.
"If the goal is an AI project that handles routine activities and delivers immediate value, it’s almost never a good idea to build your own AI," Davenport says. He brings up specialized algorithms for functions like image recognition, which can take thousands of images and months of labor to teach a program to recognize images.
“It’s a cheaper solution than spending millions of dollars hiring data scientists,” Davenport says of AI-specific vendors.
How Is AI Created?
Software development company Koombea breaks down creating your AI system into seven steps:
Identify the problem
Obtain data
Choose a programming language (ex. Java, Haskell, Julia or Python)
Choose a platform (AWS, Azure, TensorFlow)
Write algorithms
Train algorithms
Deploy
It takes skills and experience, an organized team, and financial solvency to build and maintain a robust AI in an organization’s framework.
Gartner estimates that 50 percent of all IT leaders will fail to see their AI plans come to reality. It advises a clear definition of roles to increase the likelihood of success.
“In many organizations, data scientists are still wearing too many hats due to a dearth of talent across other roles,” said Arun Chandrasekaran, Gartner’s Distinguished VP Analyst, during his session at a recent Gartner IT Symposium/Xpo®.
There are currently three general forms of AI, defined as follows:
Artificial narrow intelligence (ANI): Also known as Weak AI, focusing mainly on specific tasks such as facial recognition, language translation, and even ChatGPT.
Artificial general intelligence (AGI): AGI, or Strong AI, refers to systems capable of performing tasks without requiring humans to train the models.
Artificial superintelligence (ASI): Primarily theoretical, ASI refers to a system capable of cognitive abilities that surpasses human intelligence in all aspects while possessing their own beliefs.
Garbage In, Garbage Out
Of the pitfalls associated with AI, you’re likely to hear ‘garbage in, garbage out,’ and yes, GIGO is the acronym. If I ever hear someone say ‘GIGO’ in a meeting, it’s time to GTFO. The expression refers to the quality of the output of your system depending on the quality of the input. So many organizations have infinite amounts of data, but to use it to their advantage is the art of AI.
AI is increasingly less associated with science fiction characters and more with making our lives easier. As with electricity, automobiles, and the World Wide Web, it will take time to gain our trust and be seen less as an intrusion and more as an improvement.