Blog

What are the two types of AI?

Written by Marketing Lenet | Jan 4, 2022 3:43:56 PM

Artificial intelligence, the new power

Why is this important?


We are now in the 3rd wave of the Internet, the Artificial Intelligence Era (or AI for Artificial Intelligence). Mastering artificial intelligence means taking advantage of the extraordinary opportunities it represents.

What is IA in 1 sentence?

"Artificial intelligence, (…) is the decryption of the human learning process, the quantification of thought, the explanation of human behavior, and the understanding of what makes intelligence possible."

Kai-Fu Lee

 

The two types of Artificial Intelligence

  • Rule-based intelligence: expert humans "teach" machines the rules via a programming language.
  • The intelligence of neural networks (machine learning): machines "learn" autonomously by mimicking how the human brain works and its neural connections.

Machine learning is no longer science fiction, thanks to the proliferation of digital data (Big Data) and the increased power of computers. Whether or not you know of it, you are already in contact with this type of intelligence …

 

With deep learning, machine learning becomes a reality.

A subset of machine learning, deep learning, applies to a given field for a concrete application. Some examples :

  • Automatic piloting of autonomous vehicles
  • The evaluation of a loan file at the bank
  • Interpretation of MRI-type medical images
  • Online advertising targeting to increase clicks

Who controls Artificial Intelligence?

AI distributors: GAFAM and BATX (the Chinese giants Baidu, Alibaba, Tencent, Xiaomi) sell data hosting platforms (AWS for Amazon, etc.) and AI languages (TensorFlow from Google, Microsoft Cognitive Toolkit …). Other languages, resulting from an Open Source contribution, are also available: Pytorch, Caffe, Keras …The specialists: startups build specialized AI solutions for specific use cases. Their goal is to supplant existing players whom IA has overtaken.

 

GPT-3: the world's most potent language model

The GPT-3 language model (Generative Pre-trained Transformer – 3rd generation) uses deep learning to, among other things, produce complex, high-quality texts that humans could have written.

It can analyze 175 billion language parameters, ten times more than the most advanced previous prototypes.

GPT-3 was created by the company OpenAI, one of the flagships of Silicon Valley, with the support of Reid Hoffman (Linkedin), Peter Thiel (Palantir and one of the first investors of Facebook), and Microsoft, who has entered into an agreement with OpenAI including the exclusive license of GPT-3.


The lesson to be learned

We are no longer in the era of experimental AI but rather in the age of its implementation, which puts an end to the supremacy of expertise. Our analytical intelligence is always necessary but is no longer sufficient since the machine now does better than us …