Updated 06/04/2021 12:15
"If legislators do not protect the public from artificial intelligence (AI),
life as George Orwell describes it in his novel
could become this way in 2024,
" warned Microsoft President Brad Smith.
Smith referred to China's increasing use of artificial intelligence to monitor its citizens, hence the parallel with the novel published in 1949 by the British writer where there is a
who watches over the entire population.
Smith, who in addition to chairing Microsoft is its
legal and technical director
, said that "it will be difficult to keep up to date" with the rapid advancement of such technology, in statements to the
"If we do not enact laws that protect the public in the future, we will find that technology advances and it will be very difficult to catch up," Smith warned with concern.
"I constantly remember the lessons of George Orwell in his novel
", the novel in which the state can see what its citizens are doing all the time.
Smith referred to "China's ambition" as it aims to "become the world leader in
artificial intelligence by 2030.
The truth is that due to its infrastructure, the Asian country has plenty of resources to develop an immense surveillance system, much questioned by entities around the world and especially the European Union.
1984, by George Orwell, a science fiction classic.
In fact, in 2019, China surpassed the US in the number of patents obtained by academic institutions for innovation in AI technologies.
54% of the world's 770 million CCTV cameras
are located in China, according to research by Comparitech.
Wu Dao 2.0, the world's largest deep learning AI model
Smith's remarks have context: on Thursday, researchers from the Beijing Academy of Artificial Intelligence (BAAI) gave details about the second version of Wu Dao, a ten multimodal Artificial Intelligence (AI) model. times larger than GPT-3, the largest to date, and capable of supporting
1.75 trillion parameters.
BAAI announced the launch of its new Wu Dao 2.0 deep learning model at its annual conference on Tuesday, Engadget reports, a few months after the first version of the model was unveiled in March this year.
AI and its uses for citizen control are controversial.
The Academy revealed that the model was trained with 1.75 billion parameters, ten times more than OpenAI's model, GPT-3, released in May 2020, and so far "
the largest and most advanced language model in the world.
" , as Microsoft assured in September, when it allied with OpenAI to bring this model to its solutions.
The Chinese model is multimodal, which means you can multitask.
As shown in Endgaget, he can write essays, poems and verses in traditional Chinese, as well as create alternative texts based on a static image and images of a character close to photorealism according to descriptions made in natural language.
The BAAI also demonstrated that AI can generate virtual idols aided by Microsoft's social chatbot in China, Xiaolce, in addition to predicting the three-dimensional structures of proteins.
Identification of workers at a checkpoint and facial recognition.
To train the model in so many parameters and so quickly, Engadget reports, the researchers developed an open source learning system called FastMoE.
It works in PyTorch and is similar to Google's Mixture of Experts, although it has greater flexibility because it allowed AI to be trained on conventional GPUs and supercomputers,
and can be run on standard hardware.
The Academy indicated at the conference that large models and large computing computers are the future of Artificial Intelligence.
BAAI President Dr. Zhang Hongjiang said they are building "
a power plant for the future of AI, with mega data and great computational power
," and that they can "transform data to power
applications. AI of the future ".
Sex robots could be hacked and programmed to murder their owners
The new entertainment system of the Tesla Model X and S is like a mini PlayStation 5