Time Travel: 2014

Strong artificial intelligence is the ultimate goal of artificial intelligence development.

But how can strong artificial intelligence be so easy to achieve?

The technical principles that strong artificial intelligence wants to achieve are not complicated, and the corresponding technical indicators are easy to explain.

It is nothing more than independent awareness, independent learning, and independent decision-making.

Or to put it more popularly, it is necessary for machines to learn from people.

Learn the way people understand the world and then try to use machine learning to perform bionics at the logical level to reproduce the way people understand the world.

Everyone understands the principle, but there are various difficulties in implementing it.

Artificial intelligence in the past life was very impressive.

Anyway, until the eve of Lin Hui's rebirth, no one had developed strong artificial intelligence at that time.

Having said that, strong artificial intelligence is risky and difficult to implement.

Why pursue strong artificial intelligence?

The first is that strong artificial intelligence will significantly reduce training costs compared to weak artificial intelligence.

In the past, artificial intelligence applications generally relied on big data for training.

It sounds very high-end with the words "big data".

But in fact, it is very difficult to operate, and with the help of big data, it is actually a last resort in many cases.

We all know that artificial intelligence is the pursuit of machines that can handle problems like humans.

But you must know that people learn things from small samples.

Take the small field of image classification as an example. To classify images, people actually don’t need big data at all.

Only a few samples are needed to achieve accurate classification.

When a child of two or three years old begins to understand the world.

In order to let their children learn about animals, parents do not show their children thousands of photos, but only show them some animal atlases.

These atlases may still have dozens of pages, so the sample size is actually very small.

To be more specific, if you want children to know what kind of animal is a cat.

At most, show him a few pictures of cats and tell him what characteristics cats have and how they are different from other animals like dogs or pigs.

Soon, children can quickly and accurately identify dogs.

These are small sample trainings.

But artificial intelligence in previous lives wanted to achieve the same function.

But it requires quite a large amount of data for training.

That is to say, in the usual sense, we need to rely on big data to feed us.

Taking the same example just mentioned, it is also about recognizing animals. The neural network needs to learn to reach the level of children's recognition of animals.

Not to mention the current neural network applications that span time and space.

It relies on the more advanced deep residual neural network in previous generations.

If you want to achieve the distinction between animal images that human children have.

It also requires tens of thousands of images to conduct relatively sufficient training and obtain more accurate results in actual applications.

The reason why neural network learning “learns slowly” and “learns hard” is that

The deep residual neural network that was once the mainstay of weak artificial intelligence does not yet have the ability to think logically, associate, and reason.

With the help of strong artificial intelligence, there is no need to worry about this.

Once strong artificial intelligence can achieve a breakthrough, it will not only be easier to learn and train traditional neural networks.

Even with the help of strong artificial intelligence, it may help to achieve breakthroughs in research in basic disciplines.

These are the reasons why people pursue strong artificial neural networks.

In addition, there are many reasons why researchers pursue strong artificial networks.

It is worth mentioning that.

Some researchers believe that humans are biologically flawed.

Human beings long for a more perfect self.

Many times, the desire for strong artificial intelligence or even super artificial intelligence is also the pursuit of a more perfect self by human beings.

This is not only the limitation of this single neural network, but in fact it is also the limitation of weak artificial neural networks.

Ideal is full, the reality is very skinny.

In previous lives, no one has developed strong artificial intelligence, let alone in this life, which is slightly lagging behind in all aspects of neural network learning.

The current scientific research situation is not to mention the realization of strong artificial intelligence. The fact is that even weak artificial intelligence can be understood.

The neural network learning algorithms developed by current scientific researchers still lack the advanced functions of human beings such as association, memory, learning, and reasoning.

It would be difficult to rely on such an algorithm to create weak artificial intelligence.

I really don’t know why anyone is still worried about the risks posed by artificial intelligence ethics and machine ethics under such circumstances.

In fact, even if Lin Hui makes use of existing information advantages to make a fuss, he can only make a fuss about weak artificial intelligence.

And we are equally at a loss when it comes to strong artificial intelligence.

Under such circumstances, Lin Hui feels that instead of worrying about these risks and other frivolous things, people nowadays should actually engage in concrete and practical research.

To take a step back, are you afraid of it even if there are risks?

Is there anything in the world that is completely risk-free?

Lin Hui feels that artificial intelligence has appropriate risks, which will help the correct development of artificial intelligence.

For example, the content of the Temple Escape game is very similar to that of most parkour games. You must overcome obstacles and traps and keep running forward.

However, in Temple Escape, the player controls an Indiana Jones-like character who escapes from an ancient temple in the tropical rainforest and is chased by a group of monkey-like demon guards in the temple.

The character automatically keeps flying forward, and the player needs to control him to avoid various dangers encountered on the way to escape.

The various dangers along the way make people focus more on parkour itself.

In the same way, when a certain technology has the possibility of incurring certain risks.

Appropriate known risks may instead assist research.

Because scientific researchers may consciously stay away from corresponding risks.

Thereby focusing more on algorithm logic and application scenarios.

…

Perhaps this statement is not serious enough.

But Lin Hui has always believed that development problems must be solved through development.

Development cannot be rejected just because it may have problems.

The same is true for developments in artificial intelligence.

Even though artificial intelligence may cause a certain level of controversy.

But Lin Hui also believes that many scientific researchers hold the same view as him.

At least Lin Hui knew that Eve Carly had the same view as him.

It would be nice to have someone who can travel with me on the road ahead.

To some extent, Lin Hui also needs someone to accompany him.

In previous lives, under the instigation of the media, the hype about artificial intelligence can be said to be swarming.

There are also mixed opinions about artificial intelligence.

However, these propaganda generally focus on the propaganda that artificial intelligence replaces humans.

But they haven’t figured out the real value of artificial intelligence.

The real value of artificial intelligence lies in the changes it brings to labor and production relations.

This is why its real impact is far-reaching.