How is python used in the AI (Artificial Intelligence) field?
Hi! I am a python beginner. I have heard from many people saying that python programming language is used in AI field. But what exactly are they meaning by saying so? Is it like, creating robots and stuff (like how we see in movies) or making 'softwares' that can interact with humans. Could anyone explain how python is used in AI field in a detailed manner? What are the things that I can do in AI field if I master python (like a real pro)?
8/1/2021 1:05:13 PMSanin NBCS
15 AnswersNew Answer
The number of gross oversimplifications here is incredible. If Python is so great at and simple to use for AI then perhaps some of you experts can get Tesla's fully autonomous driving working. It seems to be beyond the capabilities of all of the engineers at Tesla. While Python can be useful in the field of artificial intelligence and machine learning so are many other languages. The Python fan boys are always shouting about how amazing Python is for AI, which is why you will hear it often. However, I have yet to read one paper on a single scientific breakthrough in AI using a program written in Python. In fact I am unaware of any commercially successful Python programs, let alone AI solutions. Many of the major libraries used by Python are written in C/C++ to compensate for Python's poor performance. Tensor Flow, one of the major Python AI libraries, is written in C++. Numpy is written in C, and the TKinter library is nothing but a wrapper around the tk (toolkit) C code library originally written for the tcl/tk scripting language. I do wish the Python fan boys would learn something about their language. If you want to learn about AI and ML using Python then that's great. The important thing is to learn about AI and ML, not just regurgitate some Python script. Since both AI and ML are the subject of considerable ongoing research a university degree would be a good place to start.
Yes, artificial intelligence and machine learning is just a bunch of algorithms and a few connected arrays of data [facepalm]. Brain surgery is just poking around inside someone's head, and flying an aircraft is just a case of moving the controls. Every subject is "just ..." when grossly oversimplified. Please watch "The danger of AI is wierder than you think" by Janelle Shane... https://www.ted.com/talks/janelle_shane_the_danger_of_ai_is_weirder_than_you_think Janelle Shane is an optics research scientist and artificial intelligence researcher, so is highly qualified to talk on the subject. I'm sure some of you Python experts can tell her where she's going wrong though. I mean she has only designed modules for the International Space Station so what does she know?
Aleksei Radchenkov, I wasn't accusing you of using pseudo science or scamming. I was trying to demonstrate the lack of critical thinking with those examples. Statements such as "just use Python" are what I expect from an 8 year old. It's like saying "just build a nuclear reactor", it leaves out a lot of details. There is an old saying "The devil is in the details" which means that the hard part is getting the details right. I have been programming for over 40 years and want to encourage young minds to enter the field. This involves teaching them to think critically and logically. Telling students it's easy is just setting people up for frustration and failure when they discover a lot of details were omitted. To achieve progress requres scientific method and demonstrable results that can be verified. Not vague generalisations.
The amount of Pythonistas drinking the KoolAid in this thread is astounding. As well as some classic demonstrations of the Dunning Kruger Effect. Yes, you just sit at home and become the next Elon Musk by writing Python AIs on your phone. You don't need to study maths or go to university so long as you believe in yourself. I feel sad for the future of STEM due to the lack of critical thinking demonstrated these days. You expect low brow buffoons on twiter and tik-tok. Unfortunately, the circle jerk mentality is creeping into the hard sciences now... Theranos Hyperloop Solar Roadways ADE-651, GT200, Alpha 6, and others. The billions of dollars wasted on this pseudo science, bad engineering, and outright scams shows that not only are snake oil salesmen alive and well but so are the rubes that buy it. Idiocracy was meant to be a comedy not a prophecy. It's also happening 500 years earlier than in the movie.
Python is quite useful when operating with data, which AI basically is. And you definitely do not to be pro in python to make AI. The complexity of an AI only depends on how large dataset is so yeah... https://en.m.wikipedia.org/wiki/Machine_learning
Martin Taylor, making an "AI" is quite easy. The problem lays in training it. The most simplistic AI could be made without use of any libraries in a few minutes with python. As I said there are array or arrays with different random values in it. And what you do is you pass data "through" arrays using specific algorithm and round values using function similar to sigmoid to get desired output. But there are two main problems with making complex AIs: 1. Training complex systems will take way too much time, so you can't just use randomised weights. 2. Making somewhat intellegent is quite impossible at this level. Machine learning is an algorithm that simulates how brain works, however it is a lot slower and it will require a lot more space of hardware to create. And obviously Tesla is not entirely AI, that would be impossible at this stage. Tesla only uses ai at some circumstances.
I never said anything about "pseudo science" and "scams". Of course there tone of things to research. What I was trying to say was that making AI's is not as complected as many say. Of course it will take a lot of work and time for complected systems, but you don't need months of study to make AI to recognise a dog on an image or play Tetris. Most researches nowdays are aimed to make a hardware AI rather than software simulation, as it will greatly increase its speed tho.
AI is basically based on some complex calculations including random functions.
Aleksei Radchenkov Quantum So, is AI a subset of machine learning? Like, creating a model with an algorithm and feeding it lots of data?
No machine learning is a subset of AI
Here is how machine learning works: you have a few connected arrays with specific different values.
You then run your dataset through those trays using specific functions, and changing values in the arrays to get on average desired output.
Here is "simple" ;) explanation of how that works: https://www.explainthatstuff.com/introduction-to-neural-networks.html
Python is used in AI with modules like tensorflow, which are written in python. You read datasets with panda & numpy which are also python modules. This is the point, where you need Python. To connect and use the moduleswith your data. To use this modules in your AI project. There is a course on SL. I got it on half way, you need some skills in Python. Try it for your self. Data: If you want to do AI things, think about data. Much data. I needed for a project at least 50 mb of text spoken of a spezific person. This was the time I start to realize what 50mb text means. Think about datasources, or a concept, where do you can get data. The neural network need them to learn & test, what it has learned. Get data to try things is a own serious topic.
You know making a nuclear reactor is not as hard either if u get all the materials. ;) Just get lots of plutonium and other nuclear stuff with long lifetime and transfer heat from it into energy. Some kid in USA did it many years ago, but then he died in jail :(