We take a break from our Cross Platform Development Series to bring you a fascinating discussion about Artificial Intelligence, led by Guy Borgford, Senior Director of Business Development, and Greyson Richey, Senior Software Engineer.
I’m highly qualified to write a piece about anything for dummies. As far as writing about Artificial Intelligence (AI) for dummies, though—I can’t get over how funny that is. But I’m not here to talk about my own questionable sense of humor; rather, I’d like to initiate a basic discussion about AI, and talk about its pros, cons, opportunities, and challenges, so that you can at least fake it at cocktail parties.
To help me wrap my artificially-intelligent brain around this topic, I’m speaking with fellow L4 peep, Glober, and AI-keener, Greyson Richey.
Guy: Greyson, thanks for joining me. First off, you hear a lot of AI terminology thrown around and used interchangeably: artificial intelligence, machine learning, cognitive computing, neural networks. Let’s clear up what some of these terms mean and discuss how they work together.
Greyson: Thank you for the invite, Guy! I’m always excited to talk about AI, which is a perfect segue into your question: What does that even mean? Artificial Intelligence is a very broad term, one that really captures most of the other buzzwords you’ve probably read. Generally, AI refers to technologies, techniques, and strategies for doing work with less human supervision. Machine learning (ML) is a bit more focused, but context is still important; it is most often (correctly) used to describe an algorithmic approach to parsing data for a conclusion that plays into human usage, like recognizing edges in an image or marking an email as spam. Neural networks (also ANNs, or deep learning) are examples of specific approaches to more advanced AI, using aspects of ML to dig deeper into what kind of work can be done by using more complex approaches to data analysis and decision making. Neural networks use a strategy that mimics the way human minds work, albeit at a very high level.
As a general rule of thumb, it’s safe to use AI to describe automated work. Language that applies to more specific applications of AI can quickly become engineer jargon, and may be misinterpreted by those who are less well-versed in AI’s detailed implementations. Someone who really knows their stuff can expand on their use-case in analogy and specific terms without relying on buzzwords.
Guy: What are some examples of AI commonly used today?
Greyson: Every time you see an ad on Facebook, check your Gmail account, or drive past a speeding camera, AI has affected your life. Those ads are informed by every post you’ve interacted with, every word you’ve posted on your wall. Your inbox is kept clean from not just already-marked spam from other users, but brand new instances that employ the same tactics. That speeding camera used text analysis software to pull your license and send you a ticket. Some of these things are more obvious than others, and they can have very little, semi-annoying, and drastic effects on your daily life.
Guy: Are there any examples of digital experiences that many think are AI but really aren’t?
Greyson: I’ve heard people talking about traffic lights now being controlled by AI. While it is true that some cities have tested this approach in limited areas, the chances are very unlikely that you’ve come into contact with AI-controlled intersections. People also assume that telemarketing is more automated than it actually is. Selecting a number from a list of options isn’t really AI, although perhaps an argument could be made that in a much broader sense, any computer doing any work is a primitive form of AI.
It is worth noting that current AI is decent at solving very narrow problem sets; it becomes exponentially worse as the amount of hidden information or number of complex systems increases. This unreliability results in a drastic increase in cost for the implementation of complex AI, so many problems that could seemingly involve AI, like traffic flow or branching human conversation, are often much cheaper when solved by humans. This is especially true if the outcome of the decision is important to not get wrong.
Guy: What do you see coming from the world of AI the next 1-2 years?
Greyson: Some of the most impressive developments will be in autonomous driving and financial systems. AI is one of those things that can eat up a ton of engineering, R&D, and even legal resources, which means that significant advancements are only going to be found where the money is, and that the money and opportunity are very much in these sectors right now. Sectors like logistics, urban planning, agriculture, manufacturing, and services are probably also going to benefit in a major way from these advancements.
Guy: What will the biggest challenge be for organizations that are facing possible disruption through the deployment of AI in their respective industries?
Greyson: Adaptation to technological advancement in business will probably play out for AI the same way that it has for previous technological advancements like the Internet and mobile phones. Those who can quickly adjust their headings, hire new talent, and integrate changes will perform better than those who can’t, but that’s easier said than done. It comes down to leadership’s familiarity with the advantages and drawbacks that something like AI offers, and the willingness to re-imagine what your business looks like in the LED glow of that umbrella of technologies.
Guy: What do you think is the greatest misconception about Artificial Intelligence?
Greyson: When leaders in this space discuss fear about what AI can do and mean, or hopefulness about what it can emancipate us from, it’s important to remember that this hinges on how we as humans employ the technology. Generalist AI is probably still a long way off, so it’s very unlikely that Skynet will come knocking at your door (or will enlighten humanity to the true meaning of life). In the meantime, AI is yet another powerful tool that we can use to further our progress as a species, or perpetuate the same issues we struggle with today (only with frightening levels of efficiency).
Guy: There are a lot of big problems to solve in our world. Which one comes to mind as a possible candidate for game-changing solutions through AI?
Greyson: In the near term, I think that there are enormous mountains that we can climb with a tool like AI in terms of efficiency in most every system with which we engage, from a less wasteful use of resources to better crop yield and food distribution, or even a global exchange that could automatically requisition and appropriate the items for tomorrow’s breakfast. In the long term, Generalist AI is the ultimate tool, and as we feed it more power and information, the theoretical limits of what that technology could do for us are the dreams of electric sheep.
Guy: Thanks Greyson. Let’s hope you’re right. Until next time!