Monday January 8, 2018 0 comments
By Thomas Frey
It was rare to find a tech columnist last year that didn’t make some reference to artificial intelligence. But it was also rare to find a writer who could clearly differentiate between the hype and true relevance of these accomplishments.
Keep in mind, even after decades of tech progress, we still get lost with our GPS, our digitally translated documents are often unreadable, and our smartphones still drop calls.
Even relatively simple computational tasks like scanning documents with optical character recognition is still not 100% accurate.
Still, many well-meaning thought leaders have issued impassioned warnings of the dangers of general AI, which is not anywhere in sight. This does not mean we are without danger, just not that kind of danger.
Perhaps the earliest dangers will come on the job front. The efficiencies that AI gives us will eliminate the need for many tasks, and even though many will be tedious jobs, working in undesirable conditions that few can point to as their dream position, they do serve as a current basis for employment, affecting countless lives.
At the same time, few jobs are truly secure with most continually morphing along with their industries. 70% of the tasks software engineers did in 2000 didn’t even exist in 1990.
Farmers, switchboard operators, and assembly-line workers in the 20th century were replaced by computer specialists, accountants, and dental hygienists. Over the coming decades we’ll see drone command center operators, data optimizers, experience designers, and other jobs that we can’t yet imagine.
Millions of unemployed workers will need to be retrained, but we don’t have a great track record here. With the vast majority of higher ed money going to colleges, the U.S. government spends a smaller share of resources on retraining than all but two other OECD (Organization for Economic Co-operation and Development) countries.
12 mind-blowing advances made by AI
It’s important to consider the dishwasher analogy. While dishwashers do offer significant efficiencies, their role in the average household is quite different than what was originally intended. In addition to being dishwashing machines, they provide an out-of-sight staging area for dirty dishes and a higher bar for all-around kitchen cleanliness.
With AI, rather than witnessing a mass elimination of jobs, we will likely see a higher bar – more thorough analysis, more control, and more certainty in the way jobs are performed.
With that in mind, here are some of the mind-blowing advances made by AI over the past year:
1.) Self-taught AI beats doctors at predicting heart attacks
As most doctors will tell you, our tools for predicting a patient’s health are no match for the complexity of the human body. Heart attacks are particularly hard to anticipate. Stephen Weng, an epidemiologist at the University of Nottingham in the United Kingdom showed that computers capable of teaching themselves perform far better than established medical processes, significantly increasing prediction rates. Once implemented, the new method will save thousands, perhaps even millions of lives a year.
2.) NASA uses A.I. to discover of two new planets
Scanning space is intensely boring work. For this reason, NASA tried a new approach and used machine learning to discover two new planets. Working with old data from the Kepler space telescope, it was able to locate two new additions to our galaxy. This wasn’t the first time researchers applied AI to sift through the massive amount of data NASA’s telescopes collect, but it is a promising example of how neural networks can leverage even some of the weakest signs of distant worlds. Thanks to AI, NASA discovered a whole new planetary system.
3.) AI is learning what makes you cry at the movies
AI is no expert at human emotions, but some visionary filmmakers have found a way to use it to gain insights on how to increase a story’s emotional pull. Through this process they were able to identify musical scores or visual images that help trigger the right feelings at the right time! In the storytelling industry, understanding the cause and effect relationships between stimuli triggers and human reactions is a powerful tool.
4.) Using AI to converts images of food into a list of ingredients
Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory have developed an AI algorithm to analyze food photos and match them with a list of ingredients and recipes. Starting with more than a million annotated recipes from online sites, their neural network sifted through each list of ingredients and a number of images associated with it. In the future, this kind of tool will help people learn to cook, count calories, and track our eating habits.
5.) Amazon develops an AI fashion designer
Amazon has developed an algorithm that can design clothing by analyzing a series of images, copying the style, and then apply it to new garments generated from scratch. In-house researchers are working on several machine-learning systems that will help provide an edge when it comes to spotting, reacting to, and perhaps even shaping new fashion trends. Since Amazon has stated, they want to become “the best place to buy fashion online,” this seems like a logical move.
6.) Resistbot, an ingenious AI chatbot will contact lawmakers for you
If you’re looking for a more efficient way for your voice to be heard, pay close attention to Resistbot. This AI-powered chatbot makes it easy to contact your political representatives. Users simply text the word “resist” to 50409. The automated bot will ask for a name and zip code to determine which public officials to contact. Since users create their own messages, creativity and clarity are critical for this service that pride’s itself on avoiding standard “form letters.”
7.) China is using AI to predict who will commit crime next
Taking a page out of the movie “Minority Report,” China is developing predictive analytics to help authorities stop suspects before a crime is committed. With their unchecked access to citizens’ histories, Chinese tech companies are helping police develop artificial intelligence they say will help them identify and apprehend suspects before criminal acts are committed. By tapping into facial recognition tech, and combining it with predictive intelligence, they hope to notify police of potential criminals based on their behavior patterns. Even though it sounds like promising tech, applications like this are getting tons of scrutiny.
8.) AI beats world’s top poker players
Humankind has just been beaten at yet another game, this time Heads-Up No-Limit Texas Hold’em poker. Since poker is a game of uncertainty, players don’t know what cards the other players have or what cards will be dealt in the future. In a game like chess or Go, all players can see the board, meaning that everyone has complete information. This makes chess and Go much easier to program than poker. Poker also requires understanding the psychology of the other players – are they bluffing, should I fold, or should I bluff? Poker also involves betting – when should I bet, how much should I bet? Does this mean the gambling industry is doomed?
9.) Researchers create a lip-reading A.I.
If you wonder why NFL coaches are now covering their mouths when they talk into their microphones, it’s because someone or something might be reading their lips. Working with Google’s Deep Mind neural network, researchers developed an elaborate training process using thousands of hours of subtitled BBC television videos. The videos showed a broad spectrum of people speaking in a wide variety of poses, activities, and lighting to simulate real life conditions. While still not perfect, their algorithm achieved promising results.
10.) ‘Mind reading’ AI is able to scan brains and guess what you’re thinking
Carnegie Mellon University scientists have developed a system that can read complex thoughts based on brain scans, even interpreting complete sentences. Using data from functional magnetic resonance imaging (fMRI) scans, the team was able to demonstrate different brain activations being triggered according to 240 complex events, ranging from individuals and settings to types of social interaction and physical actions. Using the smart algorithms they developed, the team was able to discern a person’s thoughts with 87% accuracy.
11.) AI learns to write its own code by stealing from other programs
A team of researchers at Microsoft and the University of Cambridge created a system called DeepCoder for solving coding challenges like those used in most programming competitions. Without teaching it how to code, DeepCoder uses a technique called program synthesis to piece together lines of code taken from existing software, similar to what most programmer do. Framing the outcome around a list of inputs and outputs for each section of code, DeepCoder learned which pieces of code were needed to achieve the desired results.
12.) Google’s AI was used to build it’s own AI, and it outperformed those made by humans
The creation of an AI capable of building its own AI does raise more than a few concerns. For instance, what’s to prevent the parent from passing down unwanted biases to its child? What if it creates systems so fast that society can’t keep up? It’s not very difficult to see how it could be employed in automated surveillance systems in the near future, perhaps even sooner than regulators could put something in place to control such systems.
12 critical takeaways to put AI in perspective
After scanning through each of these accomplishments, it’s easy to assume that AI is right around the corner. But a successful experiment does not a finished product make.
We are very much in the primitive early stages of AI. People in the future will often look back shaking their heads saying, “what were they thinking?”
Here are some important takeaways to help sort the reality from the hype:
A.) AI is based on algorithms
Even though today’s AI’s algorithms are very sophisticated, giving them the appearance of “humanness,” they’re still only fallible machines.
B.) AI skills will be developed in a fraction of the time of human skills
Go-master Lee Sedol began serious training when he was 8 years old for 12 hours a day. In just a few days, AlphaGo reviewed over 30 million human games and played an additional 30 million practice games with itself before taking on Lee Sedol. That means AlphaGo received at least 500 times as much practice as Lee to win the competition.
C.) There’s no such thing as a perfect AI solution
Researchers have had much more success tailoring individual AI systems to specific problems than building a logic machine capable of general intelligence. Just as AlphaGo could never be used to pilot a driverless car, AI algorithms are designed to work on specific problems.
D.) AI is forcing us to rethink what it is that makes us human
We live in a very human-centric world. Human need is what creates our global economy. There is generally no economy for things that do not benefit humankind. But what is it that sets us apart from AI and the machines that use it? We’re still a long ways from understanding where AI capabilities end and uniquely-human skills begin.
E.) As AI grows progressively ubiquitous, it’ll become increasingly invisible
AI will touch virtually every aspect of our lives, finding its way into our cars, TVs, phones, lighting, and music. With this level of ubiquity, we will quickly lose our reference points as to what life was like before AI.
F.) As we become more reliant on automation we will experience a degrading of skills and readiness when things go wrong
And yes, something will always go wrong, eventually!
G.) Blind faith in technology will cause blindness to danger as well
Once something works well, we begin to trust and rely on it. However, there is no perfect technology and the complexity of AI will make the true danger of hidden flaws nearly undetectable until it is too late.
H.) Artificial intelligence can never achieve 100% accuracy
It may indeed be quantum leaps better than anything we’re using today, even surpassing six-sigma reliability, but 100% is still not possible.
I.) The greatest dangers associated with AI will involve human failure
Sometimes the danger will stem from human ignorance, lack of oversight, or poor monitoring, but we must be constantly vigilant when it comes to spotting the purposeful failures that nefarious coders bury deep within a system.
J.) Businesses that become over-reliant on AI will fail
Admittedly there is a fine line between being over-reliant and not-reliant-enough, but hard lessons will be learned by those who fail to employ the proper checks and balance systems to oversee their AI operations.
K.) Human-based common sense will remain indispensable for the foreseeable future
Humans will continue to surpass machines for some time in areas like appreciating contextual nuances, weaving together disparate ideas, comprehending human motive and intent, integrating interdisciplinary conceptions of the world, and general intelligence. AI is simply not at our level yet.
L.) AI will soon prove to be just as good at job creation as it is at job destruction
There are currently 1 million truck drivers in the U.S., earning on average $21 an hour. It’s hard to imagine a future where those numbers don’t dwindle. But that’s only half of the story. Properly directed, AI will be able to tell us where humans are most needed in every system, process, and business operation. But beyond that, AI will be able to roadmap emerging technologies and identify the skills needed for new positions months if not years before the openings occur.
Every new technology brings its own set of dangers, and AI is no different. However, with this level of complexity, the types of danger become exponentially more difficult to understand.
It’s important to understand the symbiotic relationship, if the human economy collapses, so will the AI economy.
While we’re not going to let the bots take over just yet, it’s clear that bots are going to be meeting many of our needs, offering proactive advice, and serving us in favorable ways. Since the best possible interfaces come from the inside out, working with AI will be far less about us trying to understand the technology and far more about technology trying to understand us.
Over the past decade, the digital revolution was about us becoming accustomed to using computers all day, sending texts, connecting with each other over social media, and even learning to code.
In the AI era, technology will slide further behind the curtain into more of an assistive role, one that is not meant to be all about shiny new gadgets and operating system updates. Over time, the gadget craze will subside, as we shift our collective attention to rethinking the human experience.