Why AI Won’t Take Over The World Just Yet

Richard Fang

Definitely not in our current lifetimehttps://img.particlenews.com/image.php?url=2z5ZGs_0Yu6jJWK00

We’ve all seen it in movies.

Either a mastermind AI takes over the world (or tries to) or some unstoppable robot wreaks havoc.

Yet surprisingly, examples of artificial intelligence are surrounding our everyday life more than we know it.

Some include relying on our phones to help GPS the best way to get from point A to Point B, especially if you’re using your voice assistant AI to do this.

If you’re using search (like Google or Bing) or social media, all of these companies are using some kind of AI to help monitor and track your behavior to deliver you results they think you want to see.

With how widespread it is, it’s no wonder people like Elon Musk are scared of the so-called “point of singularity,” which is the point where artificial intelligence surpasses human intelligence.

“The singularity for this level of the simulation is coming soon. I wonder what the levels above us look like.” — Elon Musk on the topic of singularity ontop of a Rick and Morty reference

There is, however, a lack of understanding around AI.

Without knowing the nitty-gritty details around how artificial intelligence is ‘created’ in the first place, most assumptions are based on fictional beliefs, mostly founded by media and movies.

Let’s take a look at some of the details.

Current Machine Learning is improving but not perfect

https://img.particlenews.com/image.php?url=0R7BQ1_0Yu6jJWK00Photo by Franki Chamaki on Unsplash

Brooks, the Chairman, and CTO of Rethink Robotics has stated that much of the misunderstanding has come from the overhyping of certain parts of AI, such as engines beating out professional players in video games like Dota 2 (OpenAI) and board games like Go.

“An AI system can play chess fantastically, but it doesn’t even know that it’s playing a game,” says Brooks, CTO of Rethink Robotics. “We mistake the performance of machines for their competence. When you see how a program learned something that a human can learn, you make the mistake of thinking it has the richness of understanding that you would have.”

Even then, for example, it took many attempts (around 459 attempts and 10, 000 hours of ‘AI’ simulated gameplay) for OpenAI to beat a top e-sports team in a game of Dota 2.

In the end, machine learning is usually a mundane task of using data to form patterns or understandings for the AI to do a specified job. To give an example, for computer vision, data needs to be ‘labeled’. This could be identifying cars within a photo (see the example below with vehicles).

Much of the technology behind self-driving cars require millions of points of data for cars to identify hazards, other vehicles, and even animals on the roads.

https://img.particlenews.com/image.php?url=0sAKRb_0Yu6jJWK00Source: https://venturebeat.com/2018/11/16/hive-taps-a-workforce-of-700000-people-to-label-data-and-train-ai-models/

The developments of machine learning are amazing that we are starting to use different models like ‘unsupervised machine learning algorithms’.

These are models that explore unlabelled data and draw conclusions on its own rather than needing to have someone ‘supervise’ the process. These can help perform more complex processing tasks and potentially help identify new patterns within data.

The biggest drawback of unsupervised machine learning, however, is its inability to get precise information around data.

Possibly the most ‘dangerous’ categorization of machine learning is reinforcement learning. This is when you give the computer a goal through a system of ‘reward and punishment’.

An example could be, the reward for self-driving would be safely arriving at the destination while punishment is veering off the road. The system will want to maximize as much reward and minimize all punishments. Through this type of artificial intelligence, the AI can ‘learn’ how a task should be performed.

This is where fear could be justified. What if an AI is given the wrong goal?

Human intervention may be needed to make sure how AI is developed is guided in the right direction.

For now, however, without human assistance, AI is unable to adapt completely to a new situation. This could be one that is programmed to play chess, suddenly needing to play a completely different game like Dota.

The majority of current artificial intelligence simply do not understand the actual motive behind a task and rather focus on completing simple objectives.

Current technology problems

https://img.particlenews.com/image.php?url=2krrSP_0Yu6jJWK00Photo by imgix on Unsplash

On the other lens, for AI to be ‘truly self-thinking’, this requires significant computing power (think of it like your own laptop or computer but something much more powerful), especially at the scale to create constant decisions.

This requires someone to provide quantum computing power (or beyond) at a huge scale while keeping the data centers cool enough for the machines not to overheat.

At the current time, the only companies that could do this are the tech giants with the right infrastructure to run these complicated algorithms.

The future is still uncertain

There are many articles around this topic, some for and some against.

Although it’s almost certain that the current level of artificial intelligence won’t be dangerous to humankind, the uncertainty comes from thousands or even millions of years in the future.

Even with our current understanding of how artificial intelligence is developing, at the current pace of advancement, we won’t honestly know how it will look like until we reach that point in the future.

That is the unfortunate uncertain answer.

The main obstacle that super artificial intelligence needs to overcome is its understanding of the world around it.

This ‘abstract’ type of thinking may be hard to grasp for machines, and we can only hope that in the future, humans and robots collaborate to enhance the capabilities of one another rather than anything harmful.

Comments / 0

Published by

Editor at CornerTech and Marketing @richardfliu on Twitter


More from Richard Fang

Comments / 0