There’s so much hype surrounding artificial intelligence you’d think the technology was remarkably advanced.
Sure, there are home automation systems that use AI, but really that’s all the technology is good for right now: making our lives more convenient. Most AI systems cannot solve complex problems or even make appropriate decisions. Alexa and Siri, for example, may jest a bit with you but they can’t actually help you make up your mind — you still need to do that on your own.
It sounds funny to complain about the fact that AI can’t solve complex problems or truly take control, but it shows just how “dumb” the technology is currently.
People are always talking about how much damage AI could potentially do to us in the future. The funny thing is, most AI systems aren’t “nearly as smart as three-year-olds.” Worse yet, AI is almost completely dependent on big data systems. So much data is required, in fact, that only mammoth companies like Amazon, Google, and Apple have these systems in the works.
Even the director for Facebook’s artificial intelligence research — a company that has access to so much usable data it’s insane — has said “the best AI systems are dumb,” right now. This is someone responsible for building a strong AI system that could potentially be the “world’s best,” and even he still thinks they’re dumb.
Without big data, AI just wouldn’t function. Crazy, right?
Why AI is “dumb” by today’s standards
Modern AI systems are dependent on a lot of things — namely, a treasure trove of data that they can tap into. Anytime you want to ask an AI system a question, that system must access an existing source of data to find the answer.
Why does this matter, and how is it any different from the way our brains work? Technically, it’s not — aside from the fact that our brains can store a lot of information. The problem with these kinds of limitations is that it really sets restrictions on what the technology is capable of and where it can be used.
For instance, self-driving vehicles would benefit immensely from an advanced AI system. Not only would the AI be able to control the vehicle, but it could also make split-second decisions in the event of an emergency. Unfortunately, that same system would need to be connected to a network and have access to a data bank that contains all the info it needs. If that connection were to fail, there’s a real possibility the AI would run into some problems.
Again, this sets some strict limitations on how and where the technology can be deployed.
It’s not that AI isn’t useful or that it has no benefits to offer. In fact, in a business setting, AI can improve productivity in many industries, sometimes by as much as 50 percent. It has also been said that AI bots will power 85 percent of customer service interactions by 2020. But none of this lends credence to the idea that AI is truly smart or that it can learn.
The most damning evidence is apparent almost immediately. One only needs to ask: what problems does AI solve? The current answer — at least in the grand scheme of things — is “none.” It doesn’t solve any of our global problems. It doesn’t enrich our lives in any significant way. In fact, it’s mostly a novelty crammed into cute products like Google Home, Amazon Echo, and similar devices.
Where do we go from here?
Ultimately, artificial intelligence needs to become more, well, intelligent. This technology is certainly artificial, but intelligent it is not — which is funny, considering what we call it.
To change that, we must find a way to improve upon this technology so it can break free from its chains. AI needs the capacity to learn and understand, at least in the sense of incoming data. If these systems can analyze information on-the-fly and don’t require access to a huge databank to do so, we can use it almost anywhere. Until then, AI will stay less-than-intelligent.
Photo credit: Metavolution / L’oeil étranger / Surian Soosay / Jim Fischer
Source: Alison Gopnik (Edge) / Dave Gershgorn (Popular Science) / Manek Dubash (ZDNet) / Heather Pemberton Levy (Gartner)
