Do You Believe AI "Evolve"?

I have read an intriguing article from Business Insider with the title Elon Musk said people who don't think AI could be smarter than they are 'way dumber than they think they are.'

Well, in my dumbest experience of this illogical meltdown, it went like this: I'm talking about the way AI presents businesses with the ability to build machine learning architectures that can link legacy systems and diverse data sets — the so-called 'dark data' problem — to automate processes and produce much more detailed behavioral models of users, customers, and employees. That kind of insight will enable new types of frictionless business and help to mass-market traditionally hard-to-scale services.

I want to build a simple analogy; let's take the idea of personalized prices for consumers; we can see online commerce these days spoiled us with various promotions. It seems wacky, but if a market could predict our online shopping patterns, they could increase our lifetime spend by using variable pricing to incentivize repeat shopping visits more effectively and cross-sell items they know the user needs. There's a role for AI to link together all the different user-facing systems, from online to in-store, and independently for machine learning to define the right decision-making triggers for every user. Thus, rather than one-size-fits-all price promotions, which inevitably drive down user loyalty because price-sensitive buyers Google the lowest price, many sites are trying to build a personalized if we keep shopping there and provide us discounts on everything we buy sales proposition.

I remember my friend Risman Adnan mentioned, "Big Brother is watching us," and I shimmered over it. Then he said new scalable AI-supported service models, for example, everyone could potentially have a personal banker if AI does the donkey work of filling in forms, managing account compliance, researching investments that fit the user's risk profile, or loan products that match their credit scores, and many others. He amazed me with his simple explanation.

Automation AI assistants, and robotic process automation (RPA) mean one human could handle a hundred times as many customers as before (sounds boasting). Well, it doesn't matter. Twice as many are still a 100% increase in capacity, which is enormous. And so on.

Then one of my other friends, Yanuar Rahman, adds, "Yes, but we have to be careful because next thing, we'll be out of a job." Really? In my Idiot Engineering book, I stated that everyone responds enthusiastically and chips in the plot of a sci-fi action movie as though it's a regular business outline. This is like interrupting a discussion about HR processes (I used to twitch this to my friend Erlangga Yudhanegara) by pointing out that there's a risk that people who volunteer for night shifts could be vampires.

That's when it struck me that this problem arises from a fundamental misconception that the theory of evolution applies to machines, that machines live in a society, that they have unconscious desires, and also that the theory of evolution is a detailed set of Samsung manual instructions on how to use Galaxy devices.

I am not agreed with Elon Musk fully, but do we believed that machines would evolve? No brainer, they don't. Yes, controversial. I can think of many who would love to get their teeth into that and argue that they do, but they don't. Not really. Here's my simple argument:

If we could go back in time and throw an abacus (a simple computer) into the pool of amino acids and proteins that organic life eventually emerged in, it would still be an abacus now. It can't evolve. It's not organic. Sweat this, please!

If we broaden that example to include AI, it's possible that machine learning could learn within a narrow set of parameters. Not evolve. Learn. Any advanced models might be able to identify and learn new, distinct categories of a topic. So, let's go back in time and put an AI in the pool of organics we emerged from, with the right hardware. We could learn how to recognize different chains of organic molecules and perhaps extrapolate from other categories of chemicals.

It might, perhaps, if it had the right sensors, learn how to catalog new molecules or even simple organisms. But it's never going to adapt to perform a whole new category of behaviors. Or find another system, mate with it, and produce baby AIs. Unless, of course, it's been cleverly configured to do that by its creator. Touche... a creator? Well, that destroys the whole evolution argument because the idea of a creator and the theory of evolution don't get along. At all. If we create living AI, it's not evolution; it's a curious new branch of intelligent design creationism.

Another thought, intelligent design cases misunderstand evolution as much as people who think machines evolve. It's the equivalent of looking at the horse and cart, then resembling a modern car and seeing the evolution of four-wheeled transportation. That notion of evolution has a cognitive bias that implies somehow, after millions of years of horse-drawn carts having sex with each other, one of them gives birth to a luxury Ferrari (with a horse logo) #LOL.

The simple statement is to simplify the difference between the theory of evolution and the abstract concept of evolving, which is like the difference between quantum theory and the TV show Quantum Leap.

Then if Elon Musk or even Geoffrey Hinton says we have to be careful about AI.... then my argument will be the same as if next time someone mentions they have got a new smartphone, tell them they should be careful it doesn't make calls to people and say stuff without their knowledge. They will respond by saying, "it can't do that," to which we can smugly remind them that it can speak, respond to voice inputs, it can call numbers, and it can auto-suggest words. It can also send emails, manage our calendar, track our location, send strangers detailed information about what we are doing, which other devices we connect to, what we watch on Netflix and how much exercise we do. It also has access to our bank account. Put all that together, and it's equipped to steal our identity. It could...

Why the hell would we use tech that dangerous without serious safeguards in place to make sure it doesn't replace our life? Oh, wait, phones… that's not what phones do. Right. So, getting back to an AI doing our filing for our life… Gosh, then my job in Samsung is safe... but maybe not my career #wink! 

Anyway, enjoy the journey! Stay safe, healthy, and sane!