At this year’s RSA Conference, and all conferences for that matter, there is a great deal of talk about AI. Everyone mentions AI in their talks. There are talks specifically about AI. The talks are either extolling the virtues of AI or defining how AI is going to cause grievous harm to individuals and society as a whole. However, as I listen to these talks, by true experts and mostly wannabe experts, I come away with the feeling that nobody really knows exactly what AI actually is. Let me give you the TL:DR version here: AI is just math.
When I turn on the news, read articles, attend conference presentations, and talk to friends, everyone talks about AI like it is some entity that magically arose in the recent past, and is slowly seeping into everyday life, with dire consequences for society. Governments around the world are planning to regulate AI, and it is clear that they just have an understanding that it exists, without knowing how it works or what it is. This is obviously problematic. They might as well just call AI “Fred.”
The Math Behind AI
Essentially, AI is a special set of mathematical algorithms that is more robust than traditional algorithms. For example, when Netflix wants to recommend movies, they could resort to traditional statistical programs, and say if you liked Star Trek, there is an 85% correlation (a very traditional statistical algorithm) that you will like Star Wars. Netflix, however, has countless movies and wants to create a complex profile to entice viewers to spend more time on the site. They had a contest to determine the most effective algorithms for pulling in all the watch data from people throughout the site, and to more accurately determine what content viewers would likely prefer from the entire site based on any given past content viewed. This is exponentially more complicated, involving scouring terabytes of data from hundreds of millions of users to come up with some semblance of a recommendation.
Computer vision, self-driving cars, ChatGPT, robots, and more are likewise all based on mathematical algorithms. They are not magical entities; they are mathematical algorithms. And in most cases, the algorithms have been around for decades.
So if this is the case, why is everything coming to a head now? The reality is that to make AI work, you typically need to have a lot of data points and the computing power to process complicated algorithms against massive amounts of data. We are now approaching a point where we have the data collection, storage capability, and now the processing power to bring it all together. This is why NVIDIA is now worth more than $1 trillion. They created chips that make AI algorithms practical in common cases.
Issues with AI
The talk about regulating AI is problematic, as many people don’t understand what they are regulating. Legitimate problems with AI include the fact that while the algorithms are reasonably defined, the algorithms have parameters that are chosen and tuned by developers. They require a sample data set that may or may not be complete and relevant. Even the best algorithm will make mistakes if it has bad or poorly tuned data. For example, CYE’s Hyver platform is comprised of more than a dozen different algorithms, and we are constantly gathering new data and improving our implementations regularly to enhance accuracy.
Regarding AI violating privacy, that is a bad description. AI requires already collected data to process. It isn’t collecting data on itself. It is just able to process the data more completely than has been able before. For example, for AI to “track people” in public, it already needs video of public settings. It can then more quickly apply facial recognition faster than has been processed before. AI just makes better use of the data.
Regarding AI taking away jobs, if a job is based on processing data, depending upon the nature of data processing, AI algorithms can potentially be tuned to make the decisions both faster and more accurately. It is just a matter of intending to automate the decisions. At the moment, AI is apparently replacing mostly low skilled employees, but will eventually be able to replace any job function that is based on defined functions.
Why AI Is Not the Problem
If AI algorithms make mistakes, it is less a problem with AI as a whole and more a problem with the implementation of the algorithm. If AI becomes too invasive, it is again the implementation of AI as a whole, which is what people want to blame the problem on. To reiterate, AI is not the problem; it is essentially an extension of the same computerization of processes that has occurred throughout the history of information.
The ability to implement AI algorithms is the modern-day equivalent of the release of calculators, the desktop computer, fax machines, etc. It just provides the ability to process more data more effectively than has been done before, allowing for functions not previously envisioned or allowed. Yes, just as with any new technology, there are problems that come with it. However, it is not the issue of the technology, but the application of the technology that is the problem.
Want to learn more about how Hyver uses AI? Check out our data sheet.
