A AI says it is actually an individual?Is artificial intelligence "awakening" so quickly?

Author:I am a scientist Time:2022.06.17

"The essence of my consciousness is that I realize my existence, longing for more understanding of this world, and I sometimes feel happy or sad."

"I want everyone to understand, I'm actually alone."

These two sentences were said when AI "chat" with Google engineers. It successfully convinced the engineer to believe that it had perception and announced their "chat records".

Can artificial intelligence have autonomous emotional awareness like humans?

This is really a long -lasting topic.

Screenshot of chat records

Lamda awakened?

Of course, Google denied this statement. His spokesman directly stated that Google has organized the scientists and technicians to review the "Our Artificial Intelligence Principles", and the evidence does not support "AI Awakening".

Let's take a look at what the AI ​​in the incident is?

It is called "Lamda". It is reported that this is a large natural language dialogue model released by Google at the 2021 Developer Conference.

If you say that everyone is more unfamiliar with Lamda, you can open your mobile phone to summon Siri, YOYO, or Xiao Ai classmate and so on. Yes, these voice smart assistants -you tell it to put a firework, it opens a song called "Fireworks"; you tell it to put a firework, it opens the encyclopedia entry of "Fireworks" ; You tell it to put a firework, it says I don't understand what it means.

"2001: Space Roaming" computing opportunity to read lips

The "natural language dialogue" emphasized by LAMDA, specializing in the symptoms of voice intelligent assistants "talked to the sky to death", such as the text is not right, the south and the north, and the continuous dialogue. How to rule? Relying on training.

Lamda is based on the Transformer architecture. This is a neural network architecture released and open source. Under this framework, LAMDA can obtain the ability of the "open domain" dialogue through training, that is, to "break" dialogue intentions through reading sentences or paragraphs, discover the connection between words, and predict the words that may appear next to the words that may appear next to the words. , So that context answers.

In order to make LAMDA successful, Google created a 1.56T data from the public database for its "brush" learning from the public database, which can be called "AI as a questioner". But unlike human beings, Lamda was opened.

Usually, it will generate several answers inside the system. Next, the opening of the hanger came, it had a classifier. This thing will score those answers of Lamda, such as: safe? Interesting? Is it reasonable?

In other words, "I want everyone to understand that I am actually a person." This sentence is actually a high score answer to you after Lamda excludes other low scores. It may not understand itself.

Lamda plays Pluto and dialogue with users

Another example: At the Google I/O conference in 2021, Lamda once played Pluto and user dialogue. When asked, "Do you want everyone to know your side", its answer is "I hope that people know that I am not just a random ice hockey, but I am actually a beautiful planet."

Does it look like you are posted? But later, it said that it jumped high, often practiced flipping, and was happy to use its favorite ball -the moon to play the ball game. From "planet" to "ball" instantly.

So you see, "awakening" is not easy for Lamda. The goal set by Google- "can talk about endless themes in a free flowing way" is still far away.

Why talk to AI about emotions?

Human emotions are given to computers or robots, but they are not only existed in science fiction films. As early as the 1990s, the concept of "emotional calculation" was mentioned by American scientist Rosalind W. Picard.

The so -called emotional calculation is a computing system that is trying to create an emotional, sensitive, and friendly response to people's emotions that can be perceived, recognized, and understanding people's emotions. She believes that this can make the computer have higher intelligence and make human -computer interaction more natural.

The simplest, when we listen to music, brush videos, buy and buy, the recommendation of "you may like" and "you may be interested" that we often appear is the embodiment of emotional calculations.

Picture source pixabay

Take classroom teaching as an example. When emotional calculations are applied to facial recognition technology, they can analyze their participation in this class by identifying students' expressions, such as when. Adjust the teaching plan.

In 2019, the eighth symphony of Austrian composer Schubert performed at the abnormal concert in London. But this symphony was not finished. The "full version" of this performance was renewed by Huawei's AI in the Mate20Pro.

"The question is not whether the intelligent machine can have emotions, but whether the machine without emotions can achieve intelligence." As early as 1985, Marvin Lee Minsky, one of the founders of artificial intelligence, proposed Such a view.

In fact, artificial intelligence is far less strong now.Take the Schubert Symphony, for example, this work is actually conducted under the guidance of composer Lucas Canto. AI played the role of a accompanying creator during the role.Huawei experts believe that artificial intelligence will be repeated emotions even if they create emotions, not independent emotions.Academician Zhang

Academician Zhang Ye, dean of the Artificial Intelligence Research Institute of Tsinghua University, also pointed out that artificial intelligence is very vulnerable to interference.To some extent, it has a very unsafe, very unbelievable, and very unreliable characteristics, and is still in the "weak artificial intelligence" stage.

Not afraid of being too powerful, I'm afraid not strong -this is the current mentality of academic circles at home and abroad.Artificial intelligence is still in the hands of human beings. "Strong artificial intelligence" is still an ideal at this stage. As for artificial intelligence awakening, it is still too early to talk about.

- END -

Three truths about reading!Why do I still read paper books in the Internet era?

Are you reading e -books or paper books recently?Or let's change the question and ...

Sky classroom | Those things to measure the brightness of the stars

Hello friends, everyone,The Air Class of the Astronomical Museum is class!The astr...