I was doing a program with the World Talent Economic Forum (WTEF) on conversational artificial intelligence (AI) and how it is focused on telesales as an interactive voice response (IVR) replacement.
I argued that I thought this sales focus was problematic for a developing economy, the WTEF focuses on developing countries, because call centers are often placed in these countries due to cheap labor. Not only don’t the economics of replacing cheap labor with AIs often not work, but a successful implementation would result in an uptick in unemployment, increasing an existing significant regional problem.
AIs, in general, have often focused on complicated AI areas like medicine and the military. For an AI to work in medicine, you need massive data access. Yet, medical databases are secured by law and don’t follow a standard structure, making effective AI use problematic. AIs operate at scale, which means should they make a mistake, that mistake will generally be far more extensive than one made by a human, making military use particularly problematic during the early days of AI development.
So I thought I’d discuss three areas where AI could play a significant role based on impact or ease of implementation.
I recently wrote about how AI could be used for internal audits to improve audit results, because, unlike human auditors, AIs shouldn’t have to do exams by sample but could look at the entire population of data. AIs are not only better with numbers than humans, they are particularly good at rules, and, if adequately trained, they are unbiased and incorruptible. They can also flag anything that falls beyond specific parameters: Say, for instance, if someone tries to suddenly buy something from an unapproved vendor or significantly increases the amount of an order that has, up until then, been very stable, or orders a lot of something that typically isn’t consumed during the time of year when it is ordered.
While you’d still need a regular audit review to ensure the AI doing the accounting wasn’t in some way compromised. Still, you could isolate control of that process outside of accounting and have it run in parallel to catch problems as they occurred to make sure a coding error or explanation didn’t do substantial damage before the problem would be caught.
The result should be more reliable books, better forecasting, due to more reliable base data and trends, and far lower fraud. And training the AI should be relatively trivial given the nature of accounting and its synergy with computer processes.
AI In Law Enforcement
AIs are being used in security and law enforcement for facial recognition and identifying people committing crimes, behaving outside of rules, and entering unauthorized areas as well as providing secure physical access to systems and locations.
However, security and law enforcement have one big problem that AIs could help with: dealing with situations escalating violently. Police and security departments can be understaffed and undertrained for these situations.
Still, AIs can be used in conjunction with simulations for practical, low-cost training. They can listen to broadcast sound streams, hear the voice stress, and provide advice on de-escalation, automatically send a supervisor for backup, or get them in communication with the officers and capture the event for future training.
Given the massive economic and life cost of getting this wrong, justifying the cost of an AI should be a no-brainer.
AIs can scale to train individual students and learn how to keep that student engaged over time and optimize a teaching program for the result of a properly trained future employee. In emerging countries and economically disadvantaged municipalities, education is often compromised, if it exists at all. This inability to properly educate the majority of the local population sustains poverty levels. It makes it impossible for the region to dig itself out of poverty or for the population to move and get better employment. There are massive shortages of specific skill sets today. And it will worsen as we work through this 4th Industrial Revolution and new jobs are created, and we increasingly lack the trained people to fill them.
In addition, those who have jobs in these areas are likely to lose them to automation. If these now-obsolete employees don’t get retrained, they will exacerbate the economic issues in their region with their lack of employment. Spinning up AI-driven education could help emerging economies better deal with the changes they need to improve their economic situation and better assure that as jobs are lost to the new industrial revolution, the related employees can find new and better jobs to replace them.
Because educational AI would solve the most problems, it perhaps should be the highest priority.
Re-Focusing AI Efforts
AI promises to allow us to have our cars drive us, to have robots that take care of our homes, and to have super-powered telesales organizations that get us to buy stuff. But the priorities for intelligent technology should instead be to ensure our hard-won wealth isn’t stolen, our trained protectors, security and police, are well trained and don’t abuse their powers, and people are trained to do the jobs that need doing and retrained when those jobs are automated.
AIs can make the world, particularly in economically challenged areas, far better with the proper focus. But to make that happen, we need to change our priorities concerning where we train and deploy it. Accounting, law enforcement, and education are prime areas where AI could make a far more significant difference globally, and there are firms to watch in this space.