Humans generally underestimate technology in the long term and overestimate it in the short term. Part of my job at Innovalab includes reading and researching about the state of tech (4–6 hours a day), talking to specialists from different areas and then distil their commercial and personal interests from what is real or not so we can do a better job teaching students how to position themselves in this tech driven world. Below is my take on the most talked technologies of late. This is part 1: Artificial Intelligence.
Artificial intelligence, or AI, is the ability of a digital computer or computer-controlled machine to perform tasks commonly associated with intelligent beings, such as visual perception, speech recognition, decision-making, or translation between languages. You see applications of AI in your everyday life when you do a Google Search, buy something at Amazon or watch a show on Netflix.
However, more often than not, sensationalized articles around the Internet promote a dystopian view of AI as something magical that will solve all the problems in the world in a decade or replace most human jobs very soon. In this article, I analyze the current state of AI in the most common assumptions.
Assumption: Computers are soon going to become smarter than humans
The current applications of AI are very narrow in scope, which means computers can do just some tasks better than humans such as playing board games and videogames with superhuman capabilities, understanding natural speech reasonably well (Alexa, Siri), translating words in real time (try Google Translate) or recognizing faces (iPhone 11) and images (Google Lens).
Although the industry is progressing at an exponential pace in these areas, there is no evidence we will have general artificial intelligence (being better than humans in all tasks) in the next two decades. AIs still struggle to do tasks that humans consider very basic such as understanding context in sentences. Unless there is a software breakthrough in the next decade, don’t expect really intelligent robots soon.
Assumption: Cars are going to drive themselves by 2020
Autonomy was probably the most overhyped application of artificial intelligence in this decade so far. By 2015, everyone was sure we would have self driving cars by 2020. If you ask most engineers working on the problem in Silicon Valley in 2019, their answer varies from it is going to happen in 3 years to it is going to take decades (which I personally find preposterous). One of the reasons for the pessimism is that autonomous cars face many of the hardest problems in computer science (perception, decision making, prediction) and law (regulation, ethics, liability).
We currently do have trials of autonomous vehicles accepting passengers in several cities in the US and China, but in geofenced and controlled areas. Waymo, the self driving unit of Google, for instance, takes passengers experimentally in the Phoenix metropolitan region which has been mapped in high resolution and does not present complicated traffic and pedestrian patterns.
On the other hand, Tesla seems more confident in their approach to reach autonomy using only cameras. The reason is that 500K+ cars in their fleet — a hundred times more cars than any other company — are currently collecting and analyzing data in order to figure out the most difficult corner cases. Elon Musk promised full autonomy for 2020 but I think 2022 is more attainable, if at all.
Tesla and Waymo are the two companies ahead in the field so it is likely one of them will be first when autonomy finally arrives. My bet is on Tesla.
Assumption: AIs are going to replace doctors in analyzing exams
There are a few dozen AI algorithms right now approved by the FDA that may be better than doctors to evaluate medical exams. It will probably take a very long time for AIs to be better than doctors to evaluate all the exams. It turns out that trained humans are pretty good on what they do.
The consensus right now is that, in the next decade, AI will be a powerful tool to help doctors to identify fractures, tumors and other ailments but replacing human judgment whatsoever remains far away.
Assumption: Most startups are solving problems using artificial intelligence
This is not true. Artificial intelligence is often misused by entrepreneurs all over the world to dupe unsophisticated investors and raise money for their startups. In my own experience, I would say that most companies who claim they’re using modern AI for solving a specific problem are actually using old AI techniques (pre-2010s) or not any real AI at all (some startups still fake it by using humans behind the scenes).
When you hear an entrepreneur bringing too much importance to buzzwords such as machine learning, neural networks or genetic algorithms, it is generally a bad sign. Nonetheless, using AI does not automatically equate to solving customers’ problems in an efficient way.
My 2 cents
My conclusion after extensive research about these areas is that modern artificial Intelligence algorithms and techniques, despite decades of unfulfilled promises, are finally bringing fundamental changes to our society. Some of the results achieved so far are ahead of what we have ever dreamed of even a decade ago but there is still a long way to go in order to replace humans in most of our daily tasks.
Frequently Asked Questions (FAQ)
Are AIs going to become smarter than humans in most, if not, all tasks?
Yes, of course. Artificial General Intelligence (AGI) will happen and, very likely, will be our last invention.
When is it going to happen?
No one really knows for sure.
Who is ahead in the race for AGI?
Silicon Valley, closely followed by China.
When do you expect to see an AGI?
In my view, Artificial General Intelligence may be developed in the 2040s or 2050s. But as I said before, no one knows for sure.
Should our society be worried now about the quick development of AIs in many sectors?
Yes, because if we don’t, it can be too late.