By Jay Klein – CTO, Voyager Labs
There are specific business-related pain points in mind when a company decides to use products which employ AI technology. Most of you will agree that issues such as the way in which Machine Learning algorithms are utilized by the product, or the number of layers in the Deep Neural Network models mentioned by the vendor, may be meaningless in the ‘pre-decision’ discussions since it does not directly reflect the solution deployment success implications.
Nevertheless, when I joined Voyager Labs a couple of years ago, I noticed these exact discussion patterns with our customers. They were (and still are) extremely curious about the AI capabilities of our products, and were eager to better understand the algorithmic aspects of what we offer. During such discussions, familiar and related terminology is typically thrown into the air – Machine Learning, Deep Learning, Supervised, Unsupervised, etc. A fly-on-the-wall during such a meeting would conclude that customers are trying to conduct some kind of evaluation by going through an AI buzzword checklist.
Don’t get me wrong. Customers have the right to evaluate whatever is needed for them to reach proper business decisions. Moreover, I have nothing against ‘Checklists.’
A checklist is an effective tool for either a comparative analysis, or important reminders and notices. However, a checklist is only good as long as it has the ‘right’ items on it. Let me illustrate with a short story.
A farmer consulted with the wise man of his village regarding his horses. ‘I have 2 horses and it’s very hard for me to distinguish between them’ – said the farmer. The wise man raised some obvious points – ‘Mane length? Ear structure? Tail formation?’ Every time the wise man pointed out some criteria, the farmer would go back home, consider the wise man’s point, and return with an answer.
Disappointingly, both horses had the same mane length, the same ear structure…
The wise man was about to give up and requested a final inspection – ‘Maybe it’s the height?’ The farmer came back a day later with a smile on his face – ‘You were right… one inch of a difference… and it’s the white horse which is taller than the black one…’
Three things can be learned from this story (and, it’s not about the statistical aspects of horse height):
Firstly – a list of obvious and basic items on your checklist cannot tell you the true and complete story.
Secondly – It is very easy to manipulate the list.
Thirdly (and most importantly) – what’s missing from the list, in many cases are the items that are important to you, the customer!
How do we change the checklist? We redefine our goals.
Carefully and Authentically.
As previously mentioned, the reality is that behind a company’s decision to use a product with embedded AI technology are specific business-related pain points in which solutions cannot be gauged directly just by mentioning specific AI models and algorithms.
Simply stated – when creating a checklist, we don’t want to go ‘downhill’ using low-level items as it won’t allow us to emphasize the diversity and to be able to reveal the connection between the items themselves and the business case results. However, on the other hand, we don’t want to go to high-level descriptors that will either have the same effect as before or color the world in black and white solutions. For example, we have all noticed recent discussions about ‘Strong AI’ hinting at the direction of Artificial General Intelligence, which is still far away from any practical commercialization in most, if not at all cases. Putting ‘Strong AI’ on such a list, not only creates this dichotomous effect, but it is completely misleading and meaningless both from the sense of readiness (marketing) and applicability (is it relevant) to the specific business case the customer is assessing.
It is quite apparent that we need to change the discussion points surrounding AI technology and solutions to something meaningful and authentic about the real-life struggles businesses are facing. This is the time to introduce Authentic AI.
Let’s start with the ‘Authentic’ part. The Merriam-Webster dictionary defines ‘Authentic’ as:
- worthy of acceptance or belief as conforming to or based on fact
- conforming to an original so as to reproduce essential features
There’s something about this terminology, ‘Authentic.’ Don’t get me wrong, I’m not trying to hint that there is some sort of ‘Fake AI,’ which needs to be contrasted by the ‘Real’ thing. It’s really about the essential features that need to be acknowledged, and as a consequence will redefine the checklist.
For example, take a CIO/CDO perspective when facing a new problem to be solved. When an AI-based product or solution is examined, several questions may cross the mind regarding the AI aspects of it, which may reveal the previously mentioned essential features:
- Is the AI technology used by the product aimed specifically for my problem, optimally (e.g., performance, cost, etc.)?
- Is it addressing the complete problem or only a part of it?
- Can it be assimilated into the existing ecosystem without imposing new demands?
- Can it address the compelling environmental conditions of the problem space?
Think about; ‘Original,’ ‘Holistic,’ ‘Pragmatic.’
Hold your thought right there, I’ll be back soon.