Time Travel: 2014

Although Lin Hui has already thought of this.

But the actual reproduction of AlphaGo is still very difficult.

The biggest difficulty faced by Lin Hui does not come from the software level alone. The difficulties at the hardware level are obviously more difficult.

If these two technologies, AlphaGo and alphaZero, are to be launched, they will also need deep learning framework and hardware support.

Both AlphaGo and alphazZro involve the TensorFlow deep learning framework proposed by Google/Google in the previous life.

For this deep learning framework in the previous life, Google/Google also carried out related hardware development in 2016 specifically to enhance the work efficiency of the framework.

More worth mentioning is the tensor processor.

The tensor processor (TPU) is a dedicated chip customized by Google for machine learning.

(ps:...the first-generation TPU is an 8-bit matrix multiplication engine, using a complex instruction set, and driven by the host through the PCIe 3.0 bus. It is manufactured using a 28 nm process, with a die size of less than 331 mm and a clock speed of 700 MHz, and has a thermal design power of 28–40 W. It has 28 MiB of on-chip memory and 4 MiB of 32-bit accumulator, taking the results of a 256×256 systolic array of 8-bit multipliers. The TPU also packages 8 GiB of Dual-channel 2133 MHz DDR3 SDRAM, with a bandwidth of 34 GB/s. The TPU instructions send and receive data to the host, perform matrix multiplication and convolution operations, and apply activation functions.)

It’s so outrageous, it’s hard to describe domestic chips.

This is good in foreign countries, both people and ghosts are showing off.

People who seem to have nothing to do with chips are also working on chips.

Speaking of which, Lin Hui remembers some of the rough specifications of the first generation TPU.

Although I don't remember it all.

But Lin Hui still vaguely remembers that the first generation TPU was made using a 28nm process.

Although from Lin Hui’s perspective, the 28nm process is a bit rough.

But you must know that the chip process on iPhone5s is only 28nm at this time.

That sucks.

In other words, even if Lin Hui can produce TPU, the cost will most likely remain high.

Although Lin Hui can directly use the computer he brought from his previous life to run the model.

But personal computers will eventually limit the performance of some AI programs.

Even if the full performance of AI could be unleashed on a personal computer, Lin Hui would not be able to face the world with the technology from the next seven years.

Hey, it’s difficult.

Despite all the troubles.

But as the saying goes, “It’s more important to ask questions than to solve them.”

As long as your thinking doesn't slip, there are always more solutions than problems.

Compared to this era, people are still groping forward in the dark.

Lin Hui at least knows how to solve problems and the difficulties faced in solving them.

On this point, I have the definite advantage!

It involves the use of the most advanced technological processes of the era.

It seems inevitable that it will involve cooperation with Waiguo.

A certain Semiconductor Manufacturing Co., Ltd. seems to be looking at Wai Guoren’s face.

Why does Waiguo use this kind of technical capital to "enemy" it?

It sounds like it would be a good idea to set up a crooked endorsement company.

After all, there is a nest of crooked nuts, snakes and rats.

Faced with the same Waiguo company's chip manufacturers, it seems that they will not be too wary.

Waiguo Company?

Lin Hui suddenly thought of deepmind.

Since the time and space of deepmind is so miserable.

It seems good to acquire this company.

Thinking about a company that is about to file for bankruptcy protection shouldn’t be very expensive, right?

The more Lin Hui thought about it, the more he felt that it was very feasible.

The reason why Lin Hui had such an idea was besides foreign cooperation.

There is another level of consideration.

To some extent, Lin Hui needs a spokesperson.

The advent of new technologies often faces huge controversy.

Take the innovation of AlphaGo in its previous life as an example of Go.

It is difficult for people who have not learned Go in their previous lives to understand why Go has been so impacted by artificial intelligence compared to other chess sports.

Because of the complex changes in Go, each player has a very obvious style when playing chess.

Different styles will cause chess players to adopt different strategies in different stages, situations and chess shapes.

Compared with other chess games, this point is more obvious in Go.

The so-called "game" is not only a competition of brainpower between two chess players, but also a confrontation of the players' personal styles.

However, the emergence of artificial intelligence has led to the emergence of the so-called "optimal solution" in Go, which is the most complex, most varied, and best embodies the fun of "playing".

There are winners and losers in any game, and Go as a game is no exception.

In order to win, no one in the past Go world did not learn the so-called "optimal solution" of artificial intelligence.

After all, no matter how personal a chess player is, it will be difficult to defeat an artificial intelligence that has almost optimal solutions.

Go has been developed for thousands of years, and the different styles of chess displayed by countless masters have disappeared because of the emergence of artificial intelligence.

When artificial intelligence appears, many of the humanistic and emotional things that Go carries are destined to weaken a little bit.

The moment when Go is completely cracked by the powerful computing power of artificial intelligence in the future will come to a standstill.

And what is left to future generations is destined to be just some memories of the past by predecessors.

After the advent of AlphaGo artificial intelligence in the previous life.

It is extremely cruel to professional chess players.

Originally, humans thought that Go had no end.

Therefore, younger generations continue to challenge new limits that their predecessors have never been able to reach in their entire lives.

Chess players also regard this as a belief, just like climbing a mountain with no idea where the top is.

Although I don’t know what the top looks like.

But the desire for the top has become the belief that supports the chess players.

But with the appearance of AlphaGo, the originally mist-shrouded mountaintop appeared.

Everyone saw what the top looked like.

But it is a pity that humans have discovered that their understanding of Go is far behind artificial intelligence.

The emergence of AlphaGo made everyone realize that it would never be possible to reach the top of Go by relying on the physical body.

This may indeed be too cruel for professional chess players.

After AlphaGo came out, some people even thought pessimistically.

The emergence of artificial intelligence has directly destroyed the ancient game of Go.

From Lin Hui’s point of view, it’s hard to say whether the emergence of artificial intelligence will have more beneficial or harmful effects on chess.

As a general enthusiast, the emergence of AlphaGo is definitely beneficial.

The emergence of artificial intelligence has given ordinary people the ability to appreciate the games of masters.

The ability of ordinary people to match or even surpass the strongest humans in front of the screen.

…

There are various opinions.

It can be seen that the birth of a new technology is faced with great controversy.

Lin Hui is willing to take on greater responsibilities.

But I don’t want to involve myself in these meaningless disputes involving specific technical aspects.

Taking all things into consideration, it is necessary to have a crooked fruit company to deal with certain matters.

…

Having said that, in addition to using Waiguo Company to take advantage of the thunder.

It also seems great to use Waiguo to do some technical digging.

Even if Lin Hui does not deliberately use these companies to dig holes, these companies have also dug holes.

It is unjustifiable that such a professional team should not be fully utilized.

Moreover, Wai Guoren probably wouldn't care too much about the mine that was also planted by one of his own people.

Of course, in terms of technology, digging pits must be carefully arranged.

Technical pitfalls that are of little use value at first glance will not be stepped on by others.

It’s best to be the kind of pit that seems to have great commercial value but is actually very difficult to implement.

The first thing Lin Hui thought of was gait recognition technology.

Since entering the new century, a lot of biological information has been extracted as features as a means of biometric identification.

Gait recognition is no exception.

Gait recognition is an emerging biometric identification technology that aims to identify people through their walking posture.

Compared with other biometric technologies, gait recognition has the advantages of being non-contact and long-distance and not easy to camouflage.

In the field of intelligent video surveillance, it has more advantages than image recognition.

Gait refers to the way people walk, which is a complex behavioral characteristic.

Criminals may disguise themselves to prevent even a single hair from falling off their bodies at the scene of the crime, but there is one thing they have a hard time controlling, and that's the way they walk.

Everyone walks differently because of differences in muscle strength, tendon and bone length, bone density, visual acuity, coordination, experience, weight, center of gravity, degree of muscle or bone damage, physiological conditions, and There are subtle differences in individual walking "styles".

It's very difficult for a person to fake their walk, and whether a criminal is walking naturally towards a bank teller with a mask on or running away from a crime scene, their gait can give them away.

Humans themselves are very good at gait recognition, and have experience in identifying familiar people based on their gait from a certain distance.

The input of gait recognition is a walking video image sequence, so its data collection is similar to face recognition, non-invasive and acceptable.

Although gait recognition is a fairly new development direction, its working principle is still very clear.

Gait recognition aims to find and extract the changing features between individuals from the same walking behavior to achieve automatic identity recognition.

Objectively speaking, gait recognition is a technology that integrates computer vision, pattern recognition and video/image sequence processing.

When gait recognition technology is used, a person's gait is first collected by a surveillance camera, a video sequence of the gait is obtained through detection and tracking, and the person's gait characteristics are extracted through preprocessing analysis. That is, key processing in the early stage of gait recognition such as motion detection, motion segmentation, and feature extraction are performed on the gait motion in the image sequence.

Secondly, after further processing, it becomes the same pattern as the gait stored in the database;

Finally, the newly collected gait features are compared and identified with the gait features in the gait database, and if there is a match, a warning/alarm is performed. If there is no match, the surveillance camera will continue to collect gait.

An automatic gait recognition system for intelligent video surveillance is often composed of a surveillance camera, a computer and a good set of gait video sequence processing and recognition software.

But the most critical thing in this system is the software algorithm for gait recognition.

Therefore, the research on automatic gait recognition of intelligent video surveillance systems is mainly the research on software algorithms for gait recognition.

Gait recognition is a biometric technology for identification based on a person's walking posture. In many cases, gait recognition has the ability to uniquely identify people.

Theoretically, even twins whose face recognition is difficult to distinguish can be easily distinguished using gait recognition technology.

Gait recognition has broad application prospects and can be used in various security scenarios, criminal investigation, suspect retrieval and other scenarios.

Although it has broad prospects, it also has a strong uniqueness.

However, the advancement of gait recognition algorithms is not that easy.

Because the data volume of the sequence images input by gait recognition is large, the computational complexity of gait recognition is relatively high, and it is also difficult to process.

(The reason why gait recognition is difficult to implement is mainly because of the complexity of building the model.

In addition, the feature points identified by gait recognition are easy to change.

For a person, walking has different postures at different ages.

When you are in different moods, you may walk in different postures...

Putting this aside, under fairly controlled variables.

Gait recognition also encounters many challenges.

Walking with one's hands empty may be a gesture.

It may be another posture when walking with a heavy load.

It's the same load, but different load conditions are different.

After all, even with the same load, different load points mean different centers of gravity.

The existence of these problems increases the complexity of modeling)

In previous lives, although a lot of research work has been done on gait in biomechanics.

But even on the eve of Lin Hui's rebirth, the research work on gait-based identification can only be said to have just begun.

At least it is miles away from large-scale commercial use like facial recognition.

Precisely because of the difficulty of research, many companies/entrepreneurial teams that engaged in gait recognition in the past ended up failing.

In the previous life, research on gait recognition began around 2014.

However, Lin Hui’s impression is that there is no unified standard in the world until 2020.

What does it mean to have no standards? Means no commercial use.

A distinctive feature of the arrival of the third wave of artificial intelligence is:

The speed of technological maturity has been greatly accelerated.

Many technologies in this area of ​​machine recognition can be implemented in a few months but in a year or two.

And a technology like gait recognition that has not been put into practice for five or six years can be said to be a sinkhole.

It can be said that there were quite a lot of people who stepped on the trap in the past life.

Under this circumstance, Lin Hui believes that once the gait recognition thing is developed, many people will step into its trap.

Such technology seems to be a good choice to confuse competitors and distract competitors.

Moreover, Lin Hui had heard from Huang Jing before that there had been large-scale protests in this time and space regarding facial recognition.

Some countries believe that facial recognition will invade privacy.

In other words, in the direction of face recognition, there will be breakthroughs in face recognition in the future.

There is no guarantee that this country, which is a mess in time and space, will not raise privacy questions about facial recognition again.

But gait recognition is different.

No one would treat walking as privacy. After all, people’s walking must be seen by others.

Recognition through gait characteristics as a biometric should not be objectionable.

There is also a very small chance that there will be interference from factors other than technology.

In this way, wouldn't it become the best choice to attract people into the trap?