New Trend Analysis: 2016 Will Be the Year of (Artificial) Intelligence
First, please enter the information requested below. The form must be filled out completely. Then, simply click on the "Download" button to download the file directly from our website. We guarantee that the personal data you submit will be used for internal purposes only, and will not be shared with any third parties.
Perhaps you've noticed: The trend analysts at the 2b AHEAD ThinkTank have held back in giving our ranking of the TOP trends for the coming year. These kinds of rankings can be found everywhere, and are often nothing more than meaningless, unfounded theories given a forum by the routine annual change to a new year. I would rather take some time at a couple weeks' distance to talk about what is probably the most important future trend facing us now: the appearance of truly intelligent computers.
It is in fact likely that computers will achieve the intelligence of humans in our lifetime. And it is even more probably that the increasingly fast progress of artificial intelligence will not stop at the stage of human cognitive ability, but will simply pass it on its way into the stratosphere. From there we will quickly have computers more intelligent than their human masters. And then things will truly get exciting, for "intelligence" does not simultaneously mean "reason," or at least reason according to human values.
You probably read the populistic headlines when Bill Gates and Elon Musk recently warned that this development represents the greatest danger facing humanity. On the other side of the debate, experts like Ray Kurzweil and Kevin Kelly (Wired Magazine) are full of excitement about the idea of the Singularity, or the point in time when computers exceed human intelligence. Then, fortunately, there are still some moderate voices in this debate, such as Stephen Hawking and Nick Bostrom.
I would like to use the beginning of the year 2016 not only to speak to you about the opportunities and dangers inherent in artificial intelligence, but also the many possible ways to steer that future, and even perhaps what the best way is ... that is, the question of what you and I can do to make sure that the anxieties shared by Gates and Musk turn out to be unfounded. This also means that we will have to strain our own intelligence a bit more than the the typical "Top 10" trend lists and one-off features would have us believe. Over the next couple of weeks, I would like to present two trend analyses by the 2b AHEAD ThinkTank:
Part 1: How advanced are intelligent computers today?
Part 2: What progress can be predicted, and how will this change humanity?
Why are we discussing this in 2016?
We futurists at the 2b AHEAD ThinkTank are confident that the continuing development of applications for artificial intelligence will be the factor that has the largest impact on changes in our living environments and business the models of our companies in the coming decades.
All digital corporations of global significance have concentrated their research and development on this area - from Google to IBM, from Microsoft to Facebook. The consequence here will be great progress in "machine learning." We normal people will first come to see this with our smartphones. We will come to talk to our phones more often, and they will answer us naturally and intelligently. We will come to see today's Siris and Cortanas become truly intelligent assistants that users can talk to practically like human assistants.
These intelligent assistants on our smartphones will automatically understand what we are working on at the moment and will support us with genuinely helpful tips without needing prompts. They will listen to our conversations with others and will automatically display background information or will take care of the tasks that arise during the conversation (calendar entries, travel booking, etc.). Even in our everyday work, they will understand - and attempt to influence - our emotions. They will recognize when we are down and will find ways to cheer us up and keep us in a good mood.
We will also come to see that user interfaces for devices and websites will become increasingly simple, because the next steps in the development of artificial intelligence will allow user interfaces to be minimized so that they are practically invisible. These will become replaced by speech control and by the automated recognition of activities, objects, emotions, and the like through the use of sensors. "Zero UI" is the magic word for Microsoft and company in 2016.
Whether we like it or not, these things will determine the year 2016. But this will only be child's play in comparison to what comes next.
Are we "summoning the demon"? (Elon Musk)
Bill Gates, Elon Musk, Stephen Hawking ... obviously technological progress in the last few months has made some of the most innovative business leaders of today very nervous. Gates appears "upset" in interviews, Musk already sees the devil at the door, and Hawking believes that this development "could spell the end of the human race." And they are all correct. Just like chief Google trend analyst Ray Kurzweil is correct when he believes that the trend could mean an advantage for the human race.
As different as these positions are, how can they both be right?
The answer is that this discussion focuses not on one individual question, but rather on several. Our protagonists give the same answers to some questions, but differing ones to others. Every person named so far will be right in some areas, but possibly wrong in others ...
The key questions in the debate
In today's trend analysis and in the one to follow, I would like to discuss the most important questions in this debate.
Please form your own opinion here. Don't worry - you won't end up being all wrong. Whether you're name is Musk or Miller, no one really has all the answers.
So far no one has interacted with a computer that was more intelligent than a human being. We are all moving in the realm of the educated guess.
The key questions:
1. Will computers become as intelligent as humans?
2. If so, when?
3. Is that a bad thing?
4. Will computers then exceed human intelligence?
5. If so, who will have control over superintelligent computers: Humanity in general? A single person or corporation? Or the computers themselves?
6. Can we give computers that cannot be controlled human reason so that they will always act for the good of humankind?
7. If so, how?
8. If not, what does this mean for the human race?
The story so far
The discussion about artificial intelligence (AI) is really nothing new. It has been going on for decades. AI has had high and low points, has been highly praised and demonized. Currently, however, it is experiencing a rapid upturn. This upward trend began about 20 years ago. In 1997, IBM's supercomputer DEEP BLUE made headlines when it defeated world chess champion Gary Kasparov. For centuries, world chess champions were considered likely to be the most intelligent individuals of all. After 1997, we knew that computers are more intelligent in this area.
By the way: Three years before that, the computer CHINOOK defeated the reigning human world champion in women's chess. In Backgammon and Scrabble, humans also no longer stand a chance against computers.
In 2004, the first competition for self-driving cars took place: a 240-mile rally. Unfortunately, however, not a single vehicle crossed the finish line. The best car of that time fell off a cliff eleven kilometers into the race. Today, however, 12 years later, Google's autonomous cars alone have covered more than 3.2 million kilometers.
That's rapid development! And these vehicles only caused a total of 17 minor accidents. Is that better than humans can do? Statistically, yes. But judge for yourself ...
In 2011, IBM put the successor to its chess computer DEEP BLUE, named WATSON, on American TV. And the supercomputer won the notoriously difficult game show Jeopardy! against the two human players who had previously beat all human competition. It was the most impressive demonstration of the power of artificial intelligence to date.
In other areas, however, computers are somewhat behind: With crossword puzzles they have not exceeded beyond expert level; at bridge they are world-class, but not world champions; when it comes to poker they have not quite reached the level of the best players in major Texas Hold'em competitions, but can beat all human opposition at other variants of the game.
And at the royal game of go, computers have reached merely the league of skilled amateurs, the 6th Dan. Each year they improve roughly one Dan. At this rate, it will take roughly ten years before we see the first computer world champion1.
The logic behind this development seems obvious. And it's nothing new. During the dawn of steam machines, human beings held countless tug-of-war competitions with their mechanical counterparts to see who was the strongest.
"The first time a machine won, it was over. Human beings never had a chance after that.."2 It was the same story with chess and co.
Today, artificial intelligence is already doing positive things for humanity in many areas: It diagnoses illnesses better than human beings, it can find effective therapies, it can develop regenerative energies, it helps to clean the environment, it supports worldwide educations, it helps the disabled. It can be found in hearing aids, in navigation systems, in recommendation systems at Amazon & co., and in countless robots: vacuum-cleaner robots, lawnmower robots, robots for surgery, rescue robots, industrial robots, etc. In all, we now share our planet with more than ten million robots.
And the development continues to grow rapidly. It is being advanced by the gigantic investments major tech corporations are making in AI systems. Nearly every world-class internet and computer company is driving development programs of its own rapidly forward. Even investors like Elon Musk and Peter Thiel are outdoing each other with announcements for billion-dollar investments in AI research.
These developments, in the meantime, have passed the stage of corporate secret labs. On the contrary: Nearly all major drivers have gone public about their AI systems. Their reason for doing so is revealing: The more users an AI system has, the more training it receives, hence the better it gets.
For just as well will soon see, speed is possibly the decisive factor here. It could well be that first place is the only winner in this competition. But we will turn to this question in the second part of our trend analysis.
Intelligence doesn't mean reason!
A familiar argument in discussions about artificial intelligence is the thesis that computers will never achieve true human intelligence because they are merely calculating machines and are thus incapable of having feelings, emotions, or consciousness.
This is true on the one hand, but also false on the other. It is correct that intelligence has nothing to do with humanistic reason. Intelligence, in this discussion, means merely the capability of making predictions, of planning, and of end-and-means thinking in general.
What is wrong about this argument is the assumption that artificial intelligences lacking human reason must necessarily be inferior to humans. We should be wary of this self-exaltation. We humans are the result of simple evolutionary processes. It would be foolish to believe that the human being is already the best possible cognitive system. We are far removed from being the most intelligent species imaginable. It is more probable that we are, in fact, the "dumbest species capable of founding a technological civilization."3
We should therefore consider the theoretical possibility that powerful, intelligent computers could come to exist that are in themselves neither moral nor possess human reason, but are merely intelligent. When AIs, however, in the areas of planning and strategy alone achieve and exceed human levels, then it is entirely possible that these intelligences will come to manipulate and dominate human beings.
Under what values and goals this will happen, however, will become the decisive question. But we're getting ahead of ourselves.
Will computers become as intelligent as humans?
Yes! In all probability! Most experts agree on this. But not all!
Let's start with the skeptics. Spiegel Online, for example, cites Google boss Larry Page's view that worries about artificial intelligence are exaggerated:
"Yes, we are making progress towards artificial intelligence, but we are still far from the goal." It is very important, according to Page, that we make progress here: "Because the potential for improving everyone's quality of life, of making the world a better place, seems enormous to me."4
The lead developer of IBM's WATSON supercomputer takes a similar position. Eric Brown is the Director of Watson Algorithms at the IBM Watson group. He was a keynote speaker at the 2b AHEAD ThinkTank Future Congress in the summer of 2015, where he shared his own prediction about the development of artificial intelligence.
Eric Brown's magic moment came in 2011 when his computer competed against human opponents in the game show Jeopardy! There were two human candidates who had previously won against all human competition, Ken Jennings and Brad Rutter ... and someone named WATSON dared to face them both. The only problem was that WATSON is neither a he nor a she. It is an it. It is a computer. Eric Brown and his colleagues in the IBM Watson Group refuse to this day to speak of the computer as a person. WATSON should remain what it is: a tool, an assistant.
But of course, things didn't end with Jeopardy!. WATSON was then taken to a university hospital, where doctors today say that he is probably a better cancer diagnostician then the best human specialists. Next Watson will revolutionize the call center industry, and not long after will make headlines as an intelligent executive assistant.
Despite the rapid progress, however, Brown doubts - just like Larry Page - that computers will could ever become more intelligent than humans, that they could ever step out of their role as assistants. Why is that?
The answer given here is relatively clear: From today's perspective, no one has ever seen a computer with human intelligence. Today's supercomputers are specialists for individual areas of expertise. And in these areas they can beat humans hands-down. But they have no true general intelligence.
The intelligence of today's supercomputers is based solely on high calculating power and advanced pattern recognition.
In all areas of life that are not covered by that skillset, humans are better: in creativity, in business strategy, in innovation, in unconventional thinking. Even in the case of chess today, the best computers can still be beaten - by a human being assisted by another computer.
Which suggests, therefore, that it is better to spread optimism here and to call for humankind to stop the race of the human brain against computers when it comes to calculating speed, and instead to concentrate on more "human" abilities.
The optimistic forecast follows: In the future, human beings will be evaluated (and paid) according to how well they can work together with computers. Because in this team, humans are unbeatable.
... it's just not going to be that easy!
If you speak with the world's leading researchers in the field of artificial intelligence, you will realize that most of them assume that the intelligence of computers in the future will not remain limited to specific areas. The theses that humans will merely need to concentrate of the areas where they have a competitive advantage sounds comforting, but is rather unlikely.
On the contrary: Even in our lifetime, we are very likely to see the day when computers achieve the same general intelligence as humans. This so-called HLMI (human-level machine intelligence) would be considered achieved by scientists when artificial intelligence can perform roughly 80% of human jobs at least equally well as the average human.
Obviously no one can name exact dates here. We're not fortune-tellers. We can, however, get an idea about the basic time frame, when we conduct surveys among the world's leading research teams in the area of artificial intelligence. If we assume that these experts know what they are doing, at what speed they are doing it, and among themselves have a consensual feeling about which results they will achieve on what timeline, then the following picture appears:
The probability that human-level machine intelligence will be achieved by the year 2022 lies at 10%.
The probability that human-level machine intelligence will be achieved by the year 2040 lies at 50%.
The probability that human-level machine intelligence will be achieved by the year 2075 lies at 90%.5
Or to put it a bit more populistically: From 2045 on the human race might no longer be the top species on earth!
This could lead us to face bigger problems than we have had to solve so far.
Is this a bad thing?
The world's top experts have different answers to this question. Which is no wonder, for the answer to this question is based on their individual worldviews.
Those who assume that the human race in existence today is the best-developed and most perfect species natural evolution can produce are certain to find the development of greater intelligence to be a pretty bad thing.
To put it another way: Those who hold today's human to be the pinnacle of natural evolution will aim to preserve him in this form. Any changes or any development of higher intelligence would then be a dangerous and threatening attack, because the end result could be the end of the human race as we know it today. This view seems to shape the arguments of Bill Gates and Elon Musk.
Those who believe, however, that humanity at it exists today is only the interim result of an everprogressing evolutionary process will need to think objectively about how tomorrow's human being swill differ from us. One possible answer here: They will have optimized their bodies. They will be healthier, will live longer, be more physically powerful. And they will have higher cognitive ability.
How would that work? The answer of AI-optimists like Ray Kurzweil is just as simple as it is disturbing: The human brain will directly profit from the intelligence of computers and will be connected with them. This is the way that evolution will make humanity more intelligent.
A third opinion
If you would like to know my opinion, I consider the second basic conception of evolution described here to be more realistic and more probable than the first. In the years to come, we will experience the increase of intelligence in the world - in leaps and bounds!
Even if, from today's perspective, we might consider this to be "inhuman," our descendants will have another view a hundred years down the line. For them it will be normality.
However, I feel that Ray Kurzweil's vision has a heavy portion of optimism that should not be taken for granted. This view assumes that it will always be the human being who holds the power and thus dominates improved intelligence systems. Kurzweil argues in the New York Times that the future will not see a single artificial intelligence in a single hand, but rather two billion artificial intelligences in two billion hands.
I wouldn't make that assumption, though I do share his optimism that there will not be a human dictator, a powerful business magnate, or a world-dominating corporation that controls tomorrow's intelligent technology.
Kurzweil, however, fails to consider another possibility in his argumentation: the probability that a superhuman intelligence will be developed in the coming decades that will free itself from human control simply because it is more intelligent than us humans. This superhuman intelligence would probably be capable of manipulating, and thus dominating, human beings with their comparatively simple cognitive ability. This is a familiar lesson from the course of evolution. In all of our optimism, we need to consider a few things:
What are the possible routes to this superhuman intelligence?
Who will control these superintelligent computers?
Can we give "uncontrollable" computers a humanistic reason that causes them to act for the good of humanity?
... These are the questions for the second part of our trend analysis. The trend researchers of the 2b AHEAD ThinkTank will deliver the answers soon. Stay tuned for more!
We wish you a wonderful future!
1 Figures taken from Nick Bostrom: Superintelligence, p.27ff
2 IT professor Geoffrey Hinton, University of Toronto. In Spiegel Online: "Künstliche Intelligenz: Dieser Herr macht bald Ihren Job", www.spiegel.de/netzwelt/web/google-will-maschinen-denken-beibringen-a-1069072.html
3 Nick Bostrom, Superintelligence, p. 69 f
4 Cited in Spiegel Online: "Artificial Intelligence: Dieser Herr macht bald Ihren Job",www.spiegel.de/netzwelt/web/google-will-maschinen-denken-beibringen-a- 1069072.html
5 These statements are based on the average values of several surveys among the top 100 experts for artificial intelligence compiled by Nick Bostrom. See Nick Bostrom, Superintelligence, p. 37f.