With AI showing a lot of potential, albeit paired with a whole lot of room for improvement, technologists are looking at the next big thing; Artificial General Intelligence. How far is it though? When will our Cyborg fantasies become plausible? Is R2D2 nearby?
Let’s put things in perspective first
Artificial Intelligence (AI) as a concept with all the wondrous possibilities and present time uses has become a household name now. Well, that and a topic for many debates and ethical discussions questioning the basis of humanity and what it means to be a human. These discussions and debates, where AI is concerned, are unfounded.
This is because AI aims at equipping robots/software with the intelligence to perform certain tasks. These are based on repeating patterns that allow them to navigate their way through those activities. These activities may revolve around problem solving, making art or answering questions, but all of them are limited to their programming. They are made to carry out certain functions. And so, at best they serve as good assistants and helpers to humans. They cannot function outside of their realm.
Kinda boring, right?
Don’t walk away just yet. There’s something else!
The concern and discussion on the all-mightiness of humans is not completely unfounded when we talk of AGI- Artificial General Intelligence. This field of research and work has not made as much progress as AI has, but this is the one you see in movies like Ex Machina.
AGI aims to make machines that are human like in their intelligence. Instead of specializing in specific tasks, they are expected to function like search engines that can get anything done. The main idea is that they be capable of performing a bunch of different tasks and be intelligent enough to pull out the one that is being asked for. It still does not mean that the machine will necessarily be able to ‘think’ of things it has not been programmed to think. They might have the thinking power to identify a better way of executing a task, but it stops there.
If the difference still is not clear the next part should do the trick.
The coffee test
A good way to distinguish between AI and AGI is a test designed by Steve Wozniak, the co founder of Apple.
Tell a robot to make you a nice cup of coffee, note that this should not be a barista robot designed for the job. It has to the first time you’re giving this order.
If your robot assistant has AGI it should be able to come back with a steaming hot cuppa. This would imply that it looked up the best way to make a cup of coffee. Then it looked for the ingredients in your kichen, put them together and did what it had to do, the beating and everything. Let it heat, poured it, and brought it back to you. It means it can perform more than one task. It can look up a process and be sophisticated enough to follow through.
An AI robot would just look up ‘best coffee recipe’ and tell you. This is what Alexa and the Google Assistant, and Siri all do. This service pales in comparison to what humans are going for now. We are past joking about an assistant’s incompetence. We are looking for ways to fix it.
Power to learn
Where AI can extract relevant information from data even now, what specialists are aiming at is something better.
They want AGI that can learn from experiences. With an infinite amount of time and infinite amount of information/memory any problem can be solved through algorithms, sure. But engineers and technologists are aiming for ‘learning’. It is different from simple problem solving in that learning means you have acquired new information on your own through experience. It allows you to take an experience and give new information to your brain (or program) on your own and then apply that information to other situations.
It is what makes humans unique and this is what AGI development is aiming for. What it looks like on a robot and how much empowerment are we talking about as well as whether it is possible at all, are all valid questions. These are also questions we don’t have definitive answers to yet.
Give us something though. Roughly how long are we looking at?
Very long. Well, in the grand scheme of things, 50 years or so do not constitute a ‘long’ period of time by any means. However, it might not be possible in our lifetime. Although AI is showing getting prowess in analytical processes such as chess games (where robots have already beaten man), and even identifying individual’s sexuality just by a deep analysis of their faces (deep neural networks are behind it), these are still specialized activities they have been programmed for.
We need AGI to manifest in the form of robots that can do everything a human can. That is the basic definition of Artificial General Intelligence too.
According to some AI experts, AGI might appear around 2060. Some actually say 2050 too, but then again it might be never. Let’s refrain from making Back to the Future like assumptions about the future. This is all approximation though so don’t hold your breath.