Wednesday, 22 May, 2024, 11:37 AM
Advance Search

Where have all the captchas gone?

Published : Thursday, 18 August, 2022 at 12:00 AM  Count : 1380

Where have all the captchas gone?

Where have all the captchas gone?

We have all seen a situation while browsing through the internet, suddenly, when we click to browse something, a notification comes up with a tick box saying "I am not a robot". Then we have to write the word we are seeing in a box then click the tick box. It was the word captcha and if we can notice, it has been replaced by pictures. Nowadays, most of the time, we have to recognize something like a bus or train or plane from a set of pictures. The obvious question that arises from this is "How"? How does this test prove that we are not robot and who is taking these tests?

Machine. What defines a machine actually? In theory machine is an assembly of interconnected components that can perform a certain task together. Hammer is a machine and so is the engine of our car. Nowadays machines are everywhere. We have built humanoid machines called "Robot" and to operate these machines, we have written lines after lines of specific instructions that we call "algorithms" and this is the very basic of how a machine learns.

Intelligence comes from the Latin word "Intelligere", it means "to understand". So we can say that intelligence is the ability to understand. A machine can understand something too via the instructions it is given but the problem is, it cannot understand new things or if the subject matter is altered. We can understand just by experiencing with the virtue of our consciousness. That is why if we see a cat anywhere, our brain can calculate by watching its shape, colour, looks etc. and deduce that it is a cat.

The problem with machines or normal AI is, suppose in a picture if the cat is upside down, then the AI will not be able to understand that it is a cat because all the shapes would be out of place, so we have to write a different set of codes again to make it understand that even the upside down cat is actually a cat. Then if in another picture, if a dog is wearing a cat costume, the AI would assume it is a cat whereas it is not. So, we have to write another long set of codes again to make it understand it is not a cat. We can see that it is a rigorous process. Humans can understand what a cat is or what is not just by taking a glimpse based on our experiences and, we do not even know how the process works inside our brain.

The main reason is there are so much variable to consider, our brain is powerful by itself that it can process those variables in an instant and by itself. But machines do not have that power; in fact, even we humans do not have the power to explain all the variables. We just know that it is a cat. This makes the machine learning process more difficult as we have to make someone (or something?) understand something whose process even we do not know.

To tackle this, let us explain everything in a way so that everyone understands without the heavy words. At first, in the words of Elon Musk, humans have "built machines that can build other machines". Let's call them building bots. These bots can only build other bots and most of the time it is not perfect at all. They just do permutations and combinations and build bots with different patterns or algorithms. Then humans also built a teaching bots whose sole job is to teach other machines and it teaches by taking exams.

To put it simply, it provides the new machines or bots question paper where they have to recognize something and whoever can do it the most will stay and the rest will be discarded. The building machine will build new machine based on whatever is left and then again it will be sent back to the teaching machine and the cycle continues with more and more accurate bots. It is a continuous process that is still happening. When we go to YouTube and we see video suggestions, these happen in the same way. A bot is observing the videos I see and constantly trying to find a pattern. Then it shows me the suggestion based on the pattern. We do not watch most of the videos that is on the suggestions because the machine is still learning so it has not perfected the pattern yet.
Now we can come to the topic that I discussed at first, 'Captcha'. We used to see word captcha because during that time AI was not developed enough to recognize word patterns when it was jumbled. The teaching bots have the questions, it has the answer but it does not know how.So when a spam bot tries to enter, even though it is obvious to us, it does not know what is written in the captcha, but we do. Nowadays, AI has become enough strong to understand the words be it in any shape most of the time. So the word captchas have become mostly invalid. So we are taking the next step now. Recognizing objects. That is why nowadays we see pictures instead of words. We are keeping the bots out but at the same time we are basically helping them to find the pattern by themselves. For example, the AI of the future self-driving cars is educating itself through this way.

The AI will constantly become stronger and more accurate. It is inevitable. So much so that in the future may be even the photo captchas will become invalid and we have to find a new way to recognize humans from the bots. Are we going to see any "The Terminator" - esque apocalypse? But as even we humans do not know the extent of our creation, we cannot be certain for sure.
The writer is research associate, Bangladesh Institute of Governance and Management (BIGM)

Latest News
Most Read News
Editor : Iqbal Sobhan Chowdhury
Published by the Editor on behalf of the Observer Ltd. from Globe Printers, 24/A, New Eskaton Road, Ramna, Dhaka.
Editorial, News and Commercial Offices : Aziz Bhaban (2nd floor), 93, Motijheel C/A, Dhaka-1000.
Phone: PABX- 41053001-06; Online: 41053014; Advertisement: 41053012.
E-mail: info©, news©, advertisement©, For Online Edition: mailobserverbd©
  [ABOUT US]     [CONTACT US]   [AD RATE]   Developed & Maintenance by i2soft