With the market feeling highly optimistic on the potential of artificial intelligence (AI), there’s a rush of companies looking to cash in on the hype. So much so that AI startups in the US raised a record $9.33 billion in 2018 alone, per the latest PwC MoneyTree Report.
So what defines ‘true’ AI?
Real or “sentient” AI uses a combination of neural networks and machine, transfer, reinforcement, and deep learning in production models to simulate or predict the outcomes of real-world simulations. One example of this is OpenAI, a technology nonprofit co-founded by Elon Musk whose bot GPT-2 has been able to write entire articles that are so good, people can’t tell them from human writing.
Technology that isn’t quite as “predictive” as OpenAI tends to be referred to as Machine Learning DevOps, or ML Ops for short. It essentially follows a formula or behavior patterns based on human user inputs. These inputs “teach” the machine how to automate workflows or processes the way humans conduct them, enabling people to focus their time on high-value tasks that can’t be programmed so easily.
In short, there’s no reason to panic about an AI uprising or human enslavement to the machines. And here are the 5 biggest reasons why.
1. Machines can’t think for themselves.
Much of our current AI technology is not nearly as futuristic as you might think, given it can’t think for itself. So when you hear companies announce new AI features, they’re often using ML Ops to build neural networks.
As a result, machines (or platforms) are essentially learning to be better at specific tasks through predefined algorithms. In no way do they put on their thinking caps like people, become self-aware, or make autonomous decisions.
2. Machines depend on human teachers.
AI as it’s depicted in science fiction and comics usually appears as an all-knowing or superhuman brain that understands and perceives its environment as naturally as humans do. Technically, this is a case of Artificial General Intelligence (AGI).
In real life, even Siri and Alexa are dependent on algorithms, databases and neural networks (however complex the amalgamation may be) that work behind the scenes to power their output. Without the necessary ingredient of human input, current AI systems are obsolete.
3. Machines can’t solve emotional problems.
Imagine you’re waiting on a vital medical report that can go either way. You wouldn’t want to hear the bad news from a robot, would you? AI has only developed to the point of executing tasks in the most efficient manner possible. Since it does not understand the emotional implications behind these tasks, it becomes easy to see them as pragmatic beasts.
There are numerous cases of experiments gone wrong due to this intense state of apathy. Most notable was the instance at Facebook where two AI programs were shut down after they developed their own secret language to “talk” to each other. That language was essentially a blend of user inputs unique to each system, proving that AI is purely logic-based yet highly limited.
4. Machines can only decode algorithmic or formulaic logic.
Let’s say you show 1 trillion photos of red apples to an AI program, then code in a function that makes a robot find and peel that apple. What if it then finds only perfectly ripe green apples? Simply put, it will ignore them and not perform its task.
Machines are programs incapable of thinking for themselves, and can only function within the box that they are taught to live in. To them, that box is their whole existence and trying to process anything outside of it is staring a system malfunction in the face!
5. Machines are controlled by people.
This last reason is the most obvious one. Since all of today’s AI technologies cannot function autonomously, human intervention remains the primary barrier between success and failure. No resolution is bereft of the possibilities of human flaws or errors in judgement, taking true intelligence away from the underlying AI tech. And putting the onus of responsibility back on us.
AI still has a long way to go.
Modern computing power is starting bridging several gaps and achieve things we never thought possible. But this does not necessarily mean that machines themselves are intelligent.
True AI equally prioritizes data generation, collection and processing to build a strong machine foundation. But any indication that the machines can leverage that information to start altering the course of history is far off the mark.