Artificial General Intelligence Exaggerated ?

Mashup ideas combined with David Deutsch, Brett Hall, Naval, JRE and few other Books/PodCasts 

5-6 Minutes Read

Can Machines Think? For Real? Will Robots or AI take over the world, or is it a hypothetical scenario? Does AI become the dominant form of intelligence on Earth? Computer Programs or robots effectively take control of the planet away from the human species? These questions we have been talking about, hearing, and feeling a bit of Panic sometimes over almost two decades. Various mathematicians and scientific communities raised this question, “Can Machines Think?” historically.

“Alan Turing” was very well known for addressing this question: “Can a machine Think ?” in his paper “Computing Machinery and Intelligence” in 1950. He is widely considered the father of theoretical computer science and Artificial Intelligence and helped construct one of the first universal computers during the Second World War. He not only defended the idea that computers can think but also proposed a test famously known as the “Turing Test.”

Turing Test is an attempt to perform a simple test to find whether a program is human or not by a human judge. A reversed form of the Turing test is widely used on the Internet these days; the CAPTCHA. Turing suggested that rather than building a program to simulate the adult mind, it would be better to produce a simpler one to simulate a child’s mind and then subject it to a course of education. 

In 1966, one computer scientist wrote a program named Eliza to demonstrate the superficiality of communication between humans and machines by pattern matching to imitate a psychotherapist; basically, Eliza acts as Psychotherapist and answers questions asked by patients. Consider it similar(not exact) to current chatbots, Apple’s Siri and Amazon’s Alexa. And many people using Eliza were fooled by it and it. It was the first program capable of attempting the “Turing Test” and passed the test successfully.

Reason I compared Eliza with Chatbots, Siri, and Alexa because these programs were written after ~30 years and modern-day chatbots are no better than Eliza. Similarly ~45 years later, Now we have Siri and Alexa. What do you think about them today? They are still not used in the mainstream by many people and occasionally used for narrow use cases like driving directions, finding restaurant timing, or playing a song, just as an alternate mechanism for typing in their phone, and we already know about their limitations and quality of voice recognition capabilities.

I feel they won’t understand the meaning behind the actual questions most of the time, and you get funny answers sometimes, and they respond like politicians at a press conference because they never give a straight answer. Over ~50 years following Turing Test and Eliza, they haven’t achieved noticeable success.

One of the arguments Turing and other computer scientists made on Artificial Intelligence’s limited success during the 1960-70s was computer speed. It was one of the main problems where Computers have very little memory and computing capacity, and AI will break through in the coming years when computers have enough power. By the year 2000, one will be able to speak to machines without expecting to be contradicted. Yet we are here in 2022 with massive computing power on our laptops and servers. Look at all these Chatbots/Siri/Alexa. There is no sign of such an intelligent program in the mainstream. Why?

This is just one side of an argument, Chatbots/Siri are not entirely cover the evolution of AI, but the point I want to focus on, Why they weren’t successful? It’s because they all lack something, which I shall describe later. Many Computer programs these days call artificial intelligence aren’t really intelligent. 

Let’s consider some of the other examples below.

Self-driving cars, and I don’t think it’s an example of Artificial intelligence. I think It’s an example of a specific program running on an extensive fast hardware system. It’s not an explanation of producing machines as general human intelligence.

Smart Phones are brilliant and do a lot of cool things, and they can recognize faces, fingerprint sensors, and a bunch of other things. Can they recognize themselves? If we program them, yes, but that doesn’t mean it has some sort of intelligence. Are smartphones self-aware without human programming in them?

Robots, There are several types of robots that came into use, starting from Vacuum Cleaning Robots, Autonomous, Farming, Industrial, Healthcare, Logistics, and the famous Sophia (Sophia is conceptually similar to the computer program ELIZA) and rest of other advanced space robots like robotic spacecraft’s. There could be many more I am not aware of them. So let me take a simple example of Vacuum Cleaning Robots behavior which i observe almost every day and am probably qualified to talk. It does a perfect job of cleaning by remembering previous places it visited, even when lights are off, going inside the sofa, and avoiding falling from staircases. But every obstacle it comes across, it will bang the block and then move away from it, finding a different route which, of course, it stores in its memory and try not to knock next time.

 On the other hand, we humans, when walking and running, when we see an obstacle or a person in our way. An instant instinct triggers us to move away to avoid obstruction from the opposite person. My hunch is that “instinct” is something machines can never adopt instantly without a programmer programming it in the next version. This example could be fundamental. Comparison b/w a human and a cleaning Robot, but take any intelligent robot like Sophia or any other advanced robot, they can only perform whatever it has been programmed, and my fundamental question is, would it ever figure out something by itself without a previous instruction/program embedded in it? I deny the fact it wouldn’t, But how? 

We can make an argument that something like AI can get to a certain point, where we can program to create and store all the information available in the entire world, with patterns of all the thinking, all the greatest minds ever lived like poets, writers, scientists, engineers, physicists. Store what they have spoken or did in the past and apply logic with reasoning and predict some a sort of the future and start improving on patterns they learned. And then can start acting on its own based on the information stored, analyzed, and modeled.

To achieve something like that, first, we have to model how the human brain works. We all know how the brain works. It is unpredictable, depending on hundreds and thousands of neurons working together by a kind of signaling in the human neural system. And each neuron has a cell inside. And in the cell, there is this machinery going on in cells that does calculations and sends various signals back to the brain to act or say something. They all help build intelligence based on where we are, what situation, with whom, our mood, and a few other things. And you can’t just abstract it easily or model it. It is going to take another 50 years to see what’s going on inside a cell level, and hundred years we can build a brain that can simulate inside the cells.

First, we have to simulate the structure of the human brain that can hold all that information. We are talking about tens and hundreds of thousands of brains of information. We can’t even build one real brain in the next decade in AI. And just dumping enough information isn’t enough. It has to have an environment to operate, needs to have a feedback loop, a context, and several other things.

No doubt Robots can do a great job at so many things humans will love to take advantage of, like baking a cake, ironing a shirt, acting like your dancing partner, a babysitter, and solving equations in order to accomplish everything human. We will be exhausted with all the possible things people can do to put it on the list. It will never accommodate that something is not on the list(programmers). Let’s say such an AI becomes an evil AI. All you have to do is look up the program and then give it a task that’s not on the list because we humans are creative thinkers. The robot will always have a finite list of things to do. We humans have to find something not on the list or create something new on the list since a person is creative. None of the programs knows creating the knowledge. That is the key. A general purpose solver is what person who can tackle an infinite number of problems. That is the qualitative difference b/w Robots/AI and humans.

Even if things like chatbots get better at imitating people, it won’t be anything closer to the path to AI soon.AI can imitate humans imperfectly. They may never create anything new. It can only exist to certain degrees.

Take another example Humor. It can happen anywhere. It just happens based on conversation and situation. Let’s say two people are conversing about something, and sometimes it leads to a different topic, and Humor is created instantly. It might have past references sometimes, but not always. They all are created by thought during that conversation and topics discussed. The nature of Humor is not very well understood, so we do not know whether general-purpose thinking is required to compose a joke. We can never predict something like that. There is no general intelligence, and Every intelligence is contextual within the context of the environment that is in your surrounding.

Engineers reading this consider a thought experiment? Can you ever write a most intelligent program that is self-fixable and run it without testing it? This means promoting it to production with 100% confidence that it will run without any issue; if problems are encountered, will it be self-fixable? We can’t because that kind of knowledge cannot be created when a program is running but only during the development of the program by the programmer.

AI will always depend on explanations of how it works. If there is a designer, then the program is not an AI. If it was a program itself, then it is an AI, basically not depending on any tests like Turing.

People are cosmically significant, with an IQ in a genetic component. Just like Human evolution, if AI usually adapts self intelligence and self-designs evolutionary algorithms and finally settles in some machine DNA genetic code. There is a possibility. Trying to achieve them artificially may never succeed.

Consciousness is a uniquely human ability in addition to self-awareness, intelligence, free will, and knowledge creation.

Can you program the nature of Human Consciousness? If you can’t program it, you haven’t understood it. Many people say they know consciousness. Can you write a program for that, So that the computer can also have consciousness? Can we test that? I don’t think so. Even if it’s possible, I think at least not in the foreseeable future.

Can you build your ‘Self replicating robot ‘? Let’s consider an idea Elon Musk proposed, Inserting a chip in human brain to learn the patterns and all the rest.Do you think based on your past intelligence, reactions, emotions robot can take a decisions in future? Which means, do you know how you going to react/feel for future problems, success etc. If we do that can it you want to build self replicating machine which is something like that can go may be go to the moon and mars and replicate itself and carry. Does it have to have a sufficient level of intelligence very actual is consciousness.

I also want to touch a little bit on Automation.

Automation has been happening for a long time, from an industrial era to self-driving cars. But I think it’s not a problem anymore like unemployment. Automation always frees up people from mundane things, and they will eventually move on to the subsequent creative work. Like an Auto and Rickshaw driver hauling to Uber Cars, in the tech world, mainframe engineers to SQL/Java developers, and the next creative/skillful job.

There will be better and more creative jobs. We can never predict what innovative Automation will kick off next. A few years ago, if you were told I would become a YouTuber/blogger. We would have laughed. The only problem is how soon we can adopt/learn new skills/technology. Where are those people who used to work in call centers? I think the majority of them are probably doing better jobs today.

And people of all ages and geo-locations can be educated with online classes in. bulk. Like Naval said, “People who are talking about AI, Automating Programming have never really written serious code. Coding is thinking”. We are nowhere close to General AI in our lifetime. Reason is that a lot of advances in so-called AI today or what we call today a narrow AI. They are pattern matching.ML is figuring out what is the object on the screen. There is nothing like creative thinking.

Peace✌🏿