Artificial Intelligence /AI/ is all around us, and this is not a joke. It is literally everywhere. From chat-bots to smart assistants who are the most visible of its incarnations, through healthcare and architecture, AI has impacted all industries. Today AI is capable enough not only to help designing machinery much faster and more efficiently but also to create actual art.
Let’s face it – AI and its abilities are a significant part of our daily lifestyle, even though we might not suspect this at all. It is quite possible that your personal credit score is defined by AI. Just like the fruits you’ve just put in your basket – their logistic might be designed by AI. Preferences suggestions? Quite likely, they are products of AI. PC & console games, public administration, finance and pharmaceutical businesses, all of them rely on AI for one thing or another. Even the smart vacuum cleaners are advertised as the Artificial Intelligence that sweeps the floor.
We have to keep in mind that many people are scared of AI. Sci-Fi did a lot to create this perceptuion. Is there any rationale for such a feeling? Not at all, at least not yet. We have already covered this topic here with our 5 AI Myths. Busted blog post, which summarizes our fears. All of this is coming out of the „technological singularity“ term, coined in the early ’90s by Vernor Vinge, math professor, and sci-fi author. This term refers to the possible point in time when the technological growth will become uncontrollable and irreversible. Hello Skynet! The remarkable win of IBM supercomputer Deep Blue over the chess world champion Garry Kasparov in the late ’90s raised a lot of questions and even fear for the near future. Brilliant minds such as Steven Hawking and Elon Musk already warned humanity that a prevalence of artificial intelligence could result in human extinction.
So what is AI exactly?
It is a system of approaches which makes computers capable of learning through the analysis of past data and then extrapolating past trends and tendencies to perform strictly defined tasks. The decisions are made within a strictly defined framework. AI is extremely good at tasks requiring logic, even the most complex ones, while it is entirely unable to generate human-like creativity. This is why the first AI breakthrough came from the chess world, where complex calculations, rule-based decision making, and advanced math are crucial.
How smart is AI?
There are two kinds of intelligence. The first one is the so called convergence thinking, defined as the ability to answer questions correctly and predominantly displaying logic and memory. The second one is divergent thinking, which is the ability to generate many potential answers to a single problem or question. The divergent thinking shows a flair of curiosity and ability to think outside the box. Knowing the capital of Australia and figuring out how to start a successful business in Canberra without knowing English is a good example of the difference between them.
It is widely known that AI might be extremely good with convergence thinking. A research focused on the abilities of digital assistants as Siri and Cortana found out that the most brilliant of them all is Google Assistant. Its score on IQ tests is measured at 47.28, which is just bellow that of a six-year-old human score which is about 55.5. This was measured in 2016. Just a couple of years before, the Google Assistant scored only 26.4 or almost half the later score.
OK, it is evident that the Google Assistant became more and more intelligent in a short period of time but does this mean that it will keep the pace? Rather not because AI doesn’t think, it is evolving .
You won’t believe how AI is solving some problems!
There is an extremely interesting paper, written by a few computer scientists focused on the history of AI. It has a few gems within it, and we are going to share two of them. The first one is about an algorithm that was supposed to figure out how to land a virtual airplane with minimal force. But the AI soon discovered that if it crashed the plane, the program would register to forge so large that it would overwhelm its own memory and count it as a perfect score. Another remarkable example of AI solving problems is a locomotion test, performed by a simulated robot that was programmed to travel forward as quickly as possible. Instead of growing legs and starting walking, it built itself into a tall tower and fell forward. Well, it covered a horizontal distance pretty quickly, no doubt. But at the same time, AI took its task exceptionally literally. Janelle Shane, a research scientist, says that there is pure genius in such a “falling strategy”. She is emphasizing the fact that the wheat uses exactly the same strategy to propagate. At the end of each season, the wheat stalks fall over, and their seeds are spread a little bit farther from where the actual plant has grown.
What is AI perfect for?
Robotic Process Automation is one of the fastest-growing applications of AI. This is the case because it is outperforming humans on repetitive and boring tasks without needing a lunch break. Today AI combined with automation is better than any of us at such workrelated process and the bad news is it is doing it faster, more efficiently and more accurately. This is a brief list of areas where AI is successfully applied:
- Consumer demand forecasting
- Fraud prevention
- Scoring of credit risk
- Logistics and routes management
- Pricing
- Chatbots
- Sales leads scoring
- Promotions optimization
- Product recommendations
- Work schedule management
- Services personalization
- Business processes automation
- Portfolio optimization, etc.
Combining the efficiency of AI with its ability to process vast amounts of data is turning this technology in a perfect data-driven decision tool. Nothing is better than a machine in accounting for millions or even billions of variables before suggesting the best possible decision. So why don’t we outsource the decision making to AI? Because it is awful with unstructured data where we, the humans, are still the best. We easily incorporate in our decisionmaking process conversations, web pages, distant memories, experience-based intuition. This is currently impossible for AI.