3. Introduction to Artificial Intelligence
Part One
Definition and development history of artificial intelligence
4. What is AI?
Artificial intelligence, also known as intelligent machinery and machine intelligence, refers to the
intelligence shown by the machines made by people. Generally, artificial intelligence refers to the
technology that represents human intelligence through ordinary computer programs. Figuratively speaking,
artificial intelligence is a technology that enables machines to see, hear, think, speak, and move like
humans
5. History of Artificial Intelligence
1. Birth of artificial intelligence
In 1950, the computer scientist Alan Turing published a landmark paper predicting the possibility
of creating a machine with true intelligence. Given that "intelligence" is not easy to measure, he produced
the famous Turing Test, which measures a machine's intelligence by its ability to disguise a human
conversation.
In 1956, the Dartmouth Conference brought together leading scientists such as Marvin Minsky,
John McCarthy, Claude Shannon, Nathan Rochester, Allen Newell and Herbert Simon, Jointly determined
the name and task of artificial intelligence, marking the formal birth of the subject of artificial intelligence.
6. History of Artificial Intelligence
2.Two lows and three highs of artificial intelligence
⢠The First High Tide (1956 - 1974)
ď The decade after 1956 - golden age of artificial intelligence.
ď Computers were used to prove mathematical theorems, solve algebraic word problems and other fields.
⢠The First Low (1974-1980)
ď By the early 1970s, people gradually found that only the ability of logical reasoning was not enough to achieve AI,
and many problems were not solved over time.
ď The previous optimism made people expect too much and lack effective progress.
ď Many institutions gradually stopped funding AI research.
7. History of Artificial Intelligence
2.Two lows and three highs of artificial intelligence
⢠The Second High Tide (1980-1987)
ď 1980s - many companies began to develop and apply expert systems.
ď Therefore, knowledge engineering, which expert systems rely on, has become the focus of AI research.
ď John Hopfield invented the Hopfield network and solved the famous traveling salesman (TSP) problem.
ď David Rumelhart proposed Back Propagation, BP algorithm, which solved the learning problem of multi-layer
neural network.
⢠Second trough (1987 - 1993)
ď From the late 1980s to the early 1990s, the problems of expert systems, such as narrow application fields, difficult
knowledge acquisition, and high maintenance costs, gradually emerged.
ď AI has encountered a series of financial problems and entered the second low.
8. History of Artificial Intelligence
2.Two lows and three highs of artificial intelligence
⢠The third climax (since 1993)
ď Since the mid-1990s, with the rapid development of computer performance, the accumulation of massive data and
the unremitting efforts of AI researchers, AI has continuously made breakthrough achievements in many fields,
setting off a new round of climax.
11. Parallel Computing & Training & Inference
TrainingďźPipeline Parallelism
ďŹ Divide the model into different parts to run on different
GPUs in order, which can save a lot of GPU memories.
ďŹ Model layers are computing in order, so some GPU will
wait for othersâs result
12. Parallel Computing & Training & Inference
TrainingďźPipeline Parallelism
OptimizationďźPipeline Parallelism + Data Parallelism
14. CV
Auto
pilot
NLP
Audio
Computer Vision
⢠Obeject Detection
⢠Object Classification
⢠Face Recognition
⢠......
Autopilot
Natural Language Processing
Audio
The Application of AI
⢠Man-Machine Interaction
⢠Translating
⢠Emotional Analysis
⢠......
⢠Perceptual Algorithm
⢠Decision-Making
⢠......
⢠Speech Recognition
⢠Text Reading
⢠Audio Audit
⢠......
......
15. Impact of Artificial Intelligence
1. As the core force of a new round of scientific and technological revolution and industrial transformation, AI
promotes the overall leap of social productivity, promotes the upgrading of traditional industries, drives the rapid
development of "unmanned economy", and has a positive impact on the development of people's livelihood in the
fields of intelligent transportation, smart home, smart medical care, etc.
2. The application of artificial intelligence not only affects the human way of thinking and traditional concepts, but
also changes people's way of thinking and concepts. For example, traditional knowledge in the past was usually
published in magazines or books and newspapers, so the content of traditional knowledge cannot be changed.
3. The application of artificial intelligence will make the contradiction in labor employment more prominent. As AI
can take the place of human beings to do all kinds of mental work, the labor efficiency of the whole society will be
greatly improved, but at the same time, some people will have to change their types of work, or even cause their
unemployment.
The use of artificial intelligence will change the employment mode of the whole mental labor. These mental workers
will not be reused in the labor market. Once this situation occurs, it will be a big disaster for the society.
17. Conclusion
GPU high performance parallel computing drives the AI industry to start. The proposal of deep
learning has brought about a fundamental breakthrough in AI technology, which has greatly
improved the accuracy of complex classification tasks. However, deep learning requires complex
calculations on a large amount of data, so the significant improvement of GPU parallel computing
performance and big data have created conditions for the popularization of deep learning and the
launch of the AI industry.
Computing power is the basis of AI. The breakthrough of large-scale parallel computing
technology has brought about an upward turning point in the development of AI. Parallel
computing is also called parallel computing (compared with serial computing). Parallel
computing is an algorithm that can execute multiple instructions at the same time. The purpose of
parallel computing is to expand the scale of problem solving, improve the computing speed, and
solve large and complex computing problems. The so-called parallel computing can be divided
into time parallel and space parallel. Temporal parallelism refers to pipeline technology, while
spatial parallelism refers to parallel computing performed by multiple processors. Therefore, the
realization of large-scale parallel computing capability has made AI take a big step forward
Assalamualaikum and good morning to Dr normi and my fellow friends,
The topic that has been choose for today presentation is about Application of parallel computing in artificial intelligence.
before i start lets me start by introduce my groupmates first, me aishah and my other groupmates are kaimei and azreen
The content for our presentation will includeâŚ.
It seems that when we were finally understanding, implementing and getting used to industry 4.0, the term 5.0 came about.
Industry 5.0 adds a personal human touch to the two main pillars of Industry 4.0, automation and efficiency. It refers to people working alongside robots, smart machines, and technologies.
This is not to say that we should underestimate all that AI offers but rather move the conversation toward how we can make this work the best for us.
Artificial intelligence applications are all around us, but what does it really mean?
âŚâŚ
Artificial intelligence has been used in computer programs for years, but it is now applied to many other products and services. For example, some digital cameras can determine what objects are present in an image using artificial intelligence software.
To understand the idea behind Ai, we will continue with its history
First high tide 3rd point
These achievements make researchers confident about the future and believe that fully intelligent robots will emerge within 20 years.
First low 4th point
Artificial intelligence encounters the first downturn
Second high tide 5th point
- AI ushered in another round of climax.
Parallel processing for AI problems is of great current interest because of its potential for alleviating the computational demands of AI procedures. Hence, for this part I will pass to my teammates KaiMei for further explanation.
Parallel computing is very important to AI, but it is not that there are many people with great strength. The work efficiency of five people cannot be improved by five times. Therefore, it is necessary to use appropriate methods to assign one task to five people for two methods: data parallel and pipeline parallel
The principle of data parallelism: Different machines have multiple copies of the same model. Each machine is assigned to different data, and then the calculation results of all machines are combined in a certain way.
Communication mode: ring communication is the most effective way
Pipeline parallelism is to divide each layer of a model into different GPUs for calculation, which can save a lot of GPU storage space. Large parameters of the large model are stored on different GPUs separately.
However, it has a disadvantage: every time you have a GPU to calculate, it wastes hardware resources.
Data parallelism and pipeline parallelism are certain, so they are often optimized in AI: the combination of data parallelism and model parallelism can greatly reduce the waiting time.
Computer Vision
Target detection: detect whether the image contains target objects, and detect whether there are targets in the video in real time
Target classification: judge the category of objects in the picture, such as animals, plants, etc.
Face recognition: mobile face unlocking, real name authentication
Natural Language Processing
Human computer interaction: computer understands human natural language and can only interact with human
Machine translation: Google Translation, Baidu Translation
Emotional Analysis: Analyzing the Emotions in Human Natural Language
Autopilot
Automatic driving perception: detect pedestrians, obstacles, traffic lights, zebra crossings, etc. around vehicles
Automatic driving decision: process the detected information and decide the next action of the vehicle
Audio
Speech recognition: to recognize human speech into text
Text reading aloud: reading novels with emotion
Audio auditing: Automatically filter and audit sensitive audio