I’ve been thrown into the world of Artificial Intelligence (AI) with my recent involvement in the North of AI project and XaiPient. Here are some lessons that may help you better understand the technology behind it.
What is AI?
If AI conjures up the images like Terminator or the Red Queen from Resident Evil, I’m sorry to disappoint. AI of today (2021 at the time of writing), is far from capable of self-governing and conscious decision making. Current AI is at best, capable of pattern recognition. This is primarily manifested in applications such as – Image Recognition, Natural Language Processing, and Predictive Modelling. Being able to recognize patterns enables an AI to exhibit behaviours that mimic intelligence; however, current AI is still incapable of independent decision making. Current outputs and possible actions are programmed and AIs are not capable of novel associations/creations (yet, or at least I’ve yet to come across any).
Here’s a simple illustration of the progression of technology:
- Calculators – mathematical in nature, helps speed up singular processes.
- Computers – enter parallel processing, enabling us to multitask better.
- AI – data processing, allows us to account see patterns in much larger volumes of data than we normally could.
With computation, you tell it what to do: IF X, then do Y. This is automation built on rigid algorithms.
With AI, you show it how to: IF X, Y, Z…, then do A, and evaluate if the response is correct. This requires human training, which is why AI is subject to adopting human biases. If it doesn’t require training to develop its own models of pattern recognition/prediction, then it is NOT an AI.
For example, route optimization and planning is not an AI problem, it’s a brute force processing problem. It doesn’t require any training, you simply need a computer to run different simulations of all the possible routes and evaluate time/fuel cost for each route to surface the best possible option.
XaiPient’s lead scoring AI on the other hand is true AI that analyzes the behaviours of your prospects and creates a prediction model and score based on a complex probability matrix of all the activities your prospect has taken. Because it is constantly learning from your data set, it is adaptive by nature and will self-correct to improve for accuracy and account for new trends and changes in behaviour (e.g. consumer behaviour changing with global events like COVID).
So, will it make us obsolete? Will it steal our jobs?
Yes, computers eliminated typewriters.
But no, AI enables us to do more, in less time. Instead of spending hours crunching data and running statistical models, it can help surface these patterns and enable us to make quicker and better-informed decisions with our data. It’s not that we’re “dumb” and need AI to tell us what to do, it’s that AI can help us see more of the picture, more quickly, so that we can make more intelligent and informed decisions grounded in data.
Humans are ultimately still required for decision-making. WHY? Because decisions are made based on our values. We may think that we’re purely logical creatures, but the reality is far from that – pros and cons differ from person to person depending on what they value and what they’re trying to accomplish.
To an extent, I believe AI will usher in retraining and upskilling of our workforce – e.g. parcel sorting could be automated via AI through image recognition. These parcel sorters will no longer be needed on the warehousing floor if a company can afford to purchase an AI that will do the job. Is that a bad thing? I don’t think so, these boring, tedious tasks should be automated anyway.
Watch this video below:
All of these tasks could theoretically be replaced by robots that we can create with the current technology of today, no AI required. What AI will displace will ultimately boil down to cost vs. efficiency – if human labour cost less and is just as efficient as AI, it doesn’t make business sense to adopt an AI solution. AI will only have its place if it is able to improve efficiency by orders of magnitude in relation to the cost of implementation.
How have organizations used it to improve decision making?
Here are some interesting examples from our North of AI kickoff event yesterday:
HelpSeeker Technologies
Utilizing AI to analyze housing. social outreach, and welfare data to predict trends in homelessness or suicide rates so that they can proactively respond to community needs reduce these early indicators that will result in those trends coming true.
Leftovers Foundation and Fresh Routes
Analyze food supply/demand, pricing, and community data with AI to more quickly identify patterns that indicate food waste or a population’s increasing exposure to food insecurity.
Minerva Intelligence
Combining weather data with geological data to better predict landslide/flooding to improve mitigation and disaster prevention.
Is my organization ready to implement AI?
AI problem is often a data problem. Because AI requires training to identify patterns and create models of prediction, the more data you have, the more accurate it becomes. If you aren’t already collecting data, you’ll need to start by outlining a data strategy before you try to integrate AI.
Another common problem is dirty data. Standardize your data entry process and clean up your data set before feeding it to an AI.
Pingback: Advisory Email – January 2022 - ClassyNarwhal