Skip to main content

What AI can and can’t do

Although AI has huge potential, it still stops short on many things. For this post we will categorise on what AI is really good at and what AI is not capable of as of today.

What AI is very good at:

Below are some examples of what AI can do:

Translation

AI is fantastic at translating. Thanks to deep learning AI has gotten better and better at translating written text and has the capability of translating text on imagery also. AI has come a long way in this area, however it has more work to do on instantaneous translation in real time. This is very evident on translating speech into text. Right now, auto generated subtitles are having difficulty in this area but we will see an improvement as the AI model progresses through deep learning.

Driving

AI has become very good at driving based on the trained data it has been fed. Self-driving cars use a myriad of information garnered from sensors, RADAR (Tesla), LIDAR (GM and Ford) and GPS. This information is then used to determine its position, surroundings, road rules (speed, yield, stop etc.) and the direction in which it is heading.

Vision

AI has many uses in its ability to see and determine an action it needs to take. Great examples of computer vision would be in the field of facial recognition for security, detecting defects in products being manufactured or recognising number plates of cars.

What AI is not so good at?

There is still plenty of things that AI cannot do. Many people find AI quite daunting but it is important to remember that without data and instructions to follow AI is stupid. That is the “artificial” part of AI, it can only mimic, copy or make a very highly accurate guess on a potential outcome. Below are some examples of what AI cannot do:

Creating

AI can only follow a set direction or a formula. AI has no imagination it cannot form new ideas, opinions or make an independent observation on a particular topic.

Emotion

AI cannot interpret the emotional element of communication it can never truly understand the meaning of an emotion. Therefore, if one were to converse with an AI model it can only replicate an appropriate emotional response based on key words or phrases it has been told to detect.

If one was to spend two minutes with a chat bot they will run rings around it by presenting common phrases in an altered context (or in other words sarcasm that is not so obvious).

One comment we would make about AI is that it is not there to replace what humans do. What it is there to do is to help us waste less time so that it can free up humans’ time to do what they do best. Innovate and solve problems.

If you have any questions about AI we would love to hear from you. Contact us by e-mailing info@speire.ie.