Man invented tools at the dawn of history. For millennia these tools were improved and redesigned, until today, for the first time in human history, these tools might be smarter than us. Super-smart computers are scary and unpredictable. There are reasons both for concern and for optimism. You are welcome to read more here, here, here, here, and here.
Just how smart is artificial intelligence?
The current level of artificial intelligence is not very high. The machine can definitely win a match against the human champion in chess, go, poker and trivia. A human is still a better translator and driver. People and AI work together in call centers, supplementing each other strengths and weaknesses. In 2030 driving, translation, call center operation and many other jobs will be probably performed by artificial intelligence. Quite likely in 2040 your doctor, accountant, lawyer, and your child’s teacher will be some sort of combination of human and machine. Sometimes around 2060, human interaction might not be required and there will be a strange question: why do the machines need people?
In the automotive industry, there is a measure of cooperation between man and machine. The autonomous driving levels are as follows:
- Level 0: A human driver does everything. No AI operation.
- Level 1: Humanly driven, only some particular acceleration and decelleration tasks are performed by the car, like in hybrid vehicles.
- Level 2: Mostly humanly driven, with some automated lane centering and cruise control. This is the level of today’s technology.
- Level 3: Mostly machine driven, with driver intervention in complex and critical scenarios.
- Level 4: Fully autonomous in most scenarios.
- Level 5: Human is no longer required.
As the technology reaches level 4, the driver will no longer be steering. At this stage, quite possibly, there will be steering wheels only on off-roaders and emergency vehicles. Quite likely as autonomous driving reaches level 5, most professional drivers and construction workers will become irrelevant. In 2040, most professions will be in “Level 4” stage: computers will do the bulk of the work and people will overtake in extreme scenarios. A passenger aircraft has a computer capable of flying the full flight end-to-end, and two pilots which occasionally take control, mainly to remember how to fly in case of computer malfunction.
There is a universal belief that programmers will be the last people to be replaced by machines. This statement probably needs clarification.
- In early days of computing, most programmers were punch card operator. This job was eliminated long ago.
- In 2000, most programmers were designers of web applications. Today there is a handful of platforms everybody uses.
- Currently most of the programming jobs are for coders, a sort of the next generation blue-color jobs where people connect and adjust existing solutions based on given specifications. I think it can be easily replaced by AI in 2040 or earlier.
- Big-data engineers will likely be the last line of popular positions, where people will be essentially babysitting AI systems, feeding them information and watching them play well till they grow. Adversarial networks are quite capable of learning with very little human supervision, however advanced neural networks can become violent. People will probably be needed for safety measures.
And once the computers will learn beautiful and ugly, right and wrong, we will no longer be required to guide them.
Jobs people do better
There are some jobs people do better than computers, and will probably continue to do better in the visible future.
- Visionaries. Most of the great people are visionaries: finding patterns in very little information, generating a guiding vision in times of chaos, suggesting possible solutions that beat conventional logic and are non-the-less true. Visionaries are great, and we have very few of them. There is no educational program I know capable of generating large numbers of visionaries. Moreover, most visionaries take risks and end up being plain wrong, which makes a society governed by visionaries quite frightening.
- Politicians. Human conflicts are unpredictable and quite often defy logic. Politicians somehow represent large groups of people and allow division of resource and conflict resolution in highly ineffective but typically peaceful and sustainable form. The last thing anyone needs is a huge scale cyber war.
- Cleaners. Quite often things mess up in unpredictable ways. If the mess is simple to solve, people usually solve it. Our hands are very agile, and we are everywhere. This is easier to use a human instead of a machine in case of emergency. Quite often there will simply be no machine around to do the task a human can do because we are so versatile and ubiquitous.
- Artists. A computer may master drawing style much better than a human, but it will not be able to tell a story as well as we can. It will be likely guided by rules and less willing to break the rules than a great artist. Good photography gear is easy to buy, but where do we find a machine with a good eye for the situation?
People are enhancing themselves cybernetically in several ways. Our smartphones and search engines are just one of the steps in this evolution. Soon we will have smart glasses and smart earplugs, soft exoskeletons, and direct cyber links in our brain. The volume of human capabilities enhanced by technology will probably rise exponentially. Quite possibly we will also use some biological devices: designer viruses to fight cancer, 3D printable body parts, chemical compounds to raise our performance in critical times. The human being as a standalone creature will be no longer feasible, as our cybernetic enhancements rival our original capabilities.
There are already ideas of putting a financial taxation on computational capabilities. It is important to put a negative incentive on huge computers that may potentially be smarter than all of the humanity, even after we enhance ourselves cybernetically.
There were thoughts around that people are better because they are more creative, more moral, and more versatile than any computer. Recent progress shows that artificial intelligence may have all of these qualities, but they tend to behave erratically in some situations. The artificial intelligence tends to “overfit” its behavior to the data it observes. Once it is presented with information it has never seen before its behavior is unpredictable.
It is very hard to build a computer that is not biased. A typical AI algorithm tends to follow the patterns shown to it, without trying to understand the reasons underlying these trends. People with dark skin are notoriously difficult for certain computer vision algorithms. Researchers found that high-paying job ads were being shown disproportionately to men by AI. Microsoft’s chatbot Tay was turned into a racist xenophobe within hours. The human activity provides the data that trains the machine. But that means it inherits the biases of people and quite often increases them. In the case of Trending Topics, when enough people share a story — fake or odious or not — the algorithm deems it important and promotes it in popularity. In the case of Tay, human trolls turned her into one of them.
As computers gain power, security issues of a cyber attack are ever more frightening. Just as there are two pilots on an aircraft, there probably be two monitoring humans on most critical systems, in case the systems are maliciously attacked.
Economy of the future
Approximately 50 years from now, the world will probably change without recognition. Computers will do most of the work and pay most of the taxes, with humans taking very specific jobs.
Most of the things we need, will probably be produced on the spot by various 3D printing technologies and melted after being used.
Quite probably the main job for humans will be Policy setting and security monitoring, making sure nothing catastrophic happens in critical systems of the computers and taking over in a case of emergency.
There will be some artists and visionaries, heavily linked to very strong computers, working in “centaur mode” with computers, setting new goals and generating exciting new solutions and products.
And then there will be the rest of humanity, as foreign to the advanced cyber workers as the Bangladesh farmers are foreign to the American engineers, and possibly as poor in comparison. Their fate is unpredictable, and it is best not to be one of them. Maybe, just maybe, their ubiquitous presence and limited computational capability will make them cheaper than robots, and their life will be similar to the life of most people today.