Blog: Decoding AI : the 21st Century Marvel
How often have you got scared of the thought that AI, in its peak form, will bring down humanity on its knees like in the sci-fi marvel “Her”? Do you also think our end will be at the discretion of a bunch of binary soldiers who would decimate humans at will, as shown in “Terminators” series? That AI will be so smart, it would solve problems that we would never have known to be existing, in the split of an eye?
If your thoughts align to these questions above, then that proves to me one thing: You watch a hell lot of Sci-Fi Movies! And what’s best, you believe most of them to be true.
The Reality? Far from it.
Does this sound overwhelming to you? In that case, let us have a quick recap of where AI originated from, and the purpose that it was meant to fulfill.
The term “Artificial Intelligence” was coined way back in 1956 by John McCarthy who chaired the first AI research at Dartmouth College. However, the quest for creating sustainable Machine Intelligence had begun earlier. John Von Neumann, Issac Asimov, and Alan Turing were noted contributors to a similar dream: Enabling a machine to do what humans could.
Cut to today’s date. The dream that the founding fathers of AI saw, to my mind, has already come true. If they were to wake up from a deep sleep for the first time after the early years in AI (like Captain America!) , they would be amazed how AI is being used to boost innovation for the entire humanity.
One of the major differences that exist between the AI in the ’60s and the ’70s and the kind of AI research that we have now is: Specialization. Artificial Intelligence is influencing decision making around the world and seems promising to do so in the future as well.
Wait, wasn’t that the fear you had in mind at the beginning of this article? Feeling cheated already? Hold on for a while.
There is a reason why we call it Artificial Intelligence. In layman terms, it is still pretty much rule-based, or externally programmed. It still must feed and grow on data generated either by us or the systems that we as humans have control over. If we carefully observe, we haven’t yet made it think on its own, we are enabling it to think and process as we would like it to. The remote is still in our hands. And will always be.
Consider these examples:
- Dairy farms in Ireland are using AI for tracking and identify movement and behavior of cows within their premises. This information is used to develop key animal and farm performance indicators which are sent to a dairy farmer’s phone in the form of daily notifications and real-time detailed analytics. It helps in identifying and analyzing inefficiencies and animal health issues which helps to improve productivity and animal welfare.
- HP demonstrated that AI could turn amateurs into Pros when they used their AI lab’s cognitive computing platform to analyze two years’ worth of call data for a client’s call center. The call center was using a queue-based system for routing customer calls, resulting in long wait times and poor-quality customer support. The cognitive computing platform could determine each agent’s unique knowledge of a specific kind of customer request, based on the previous calls. This is now used to match incoming calls to agents who have successfully processed similar requests in the past. As a result, the support center has seen a 40% improvement in first contact resolution and a 50% reduction in the rate of transferred calls.
The list could go on and on, but the question remains: How can we be so sure about AI not turning its back against us? The answer lies in the process carried out while creating any Artificial Intelligence. Let us explore them and see if AI-enabled systems can manipulate these steps:
- Collecting Data: Data is indeed the driving fuel to most of the AI innovations at present or in the days to come. While it can be manually collected, most people rely on Web scraping (a generic term for collecting data from the internet to your local system) to collect this data. So, unless our AI system and the web crawler we created for scraping the internet are mutual buddies and can talk in binary, we can breathe easy!
- Cleaning Data: Often we tend to witness unexpected bias in the collected data. Some of it needs removal, while none of the bias is necessary. We generally perform a range of procedures (EDA, Feature Engineering) to clean our data as soon as we get it from the source. Dedicated libraries such as Pandas/ Numpy in Python and dplyr/ reshape2 in R make this task very easy. (Sorry Thanos, we don’t snap here!) So far, these libraries are still managed and developed by a team of humans. Hence, unless these guys mess it up for us, we can put the fears of an AI takeover to rest.
- Preparing an AI model: This is the part where we provide an environment for the system to learn from the data and perform analysis for us. There are options available on how smooth we want this learning process to be (Learning Rate), and whether we want it to forget something it learnt (Unfreezing layers). As long as we have this amount of control over the system, we can always think of avoiding the doom.
- Training the Model: This depends on the number of cycles we decided to train our machine for, the kind of grouped data we feed to it (Training and Test sets) and the kind of resources (CPU/GPU) that are available for processing. All these parameters are well controlled by humans unless we consider the AI systems to be biased towards any of them.
- Validating the results: No surprises here, as the machine doesn’t get a taste of data apart from what we fed it earlier. It’s like the story of the Elephant and the four blind-folded men. Since each one of them was blind-folded, each derived at a separate description of the elephant. And so, will our machine. It however, would go one step ahead, combine those descriptions and arrive at a successful validation, confirming that the object in question is an elephant, upto a certain accuracy!
So, we see it’s a no-brainer that AI isn’t there to threaten as they show in the movies. That’s just some ignorant human who had too much caffeine to drink, glorifying a tale of fantasy to make huge amount of dollars.
AI might be a black box which might be holding the world’s biggest possibilities, a magic cloak that wraps the reality and lets us live in awe of it. It could even be your soulmate if you want it to be that way!
Whatever it be, it doesn’t necessarily come wielding the hands of wrath as many people have imagined so far. Nor is it like an untamed beast that, if teased and played around with, could bring tragedy to the mightiest of the species: Homo Sapiens.
So, next time you see someone terming AI dangerous in the long run, you may quote this:
“Artificial Intelligence is as effective as the data that fuels it, and as good as the human that feeds the data to it.”