Do I need special hardware (video card) for the basics of machine learning?

Question

Machine learning

Hello! In the near future, I plan to start studying machine learning for linguistic tasks, possibly neural networks. At the same time, I know that it’s very good for neural networks to have a good video card, but I’m not sure that I’ll write large and long algorithms right away. Right now I'm choosing a laptop: the version without a graphics card is Asus VivoBook 17-X712JA-AU265, with a graphics card - Lenovo IdeaPad L340-17IRH Gaming 81LL00FCRU.

Dear experts, attention, question: what kind of laptop should I choose?

PS: if you would like to express an opinion, but you are too lazy to justify it, you can mark the option here https://forms.gle/jP8YfAJpXkwGtVPg6

Answer 1

You can just try it on Google's hardware in Colab and other alternatives. If you are going to train all sorts of GPT-3 from scratch, then no laptops will save you.

Answer 2

What are we comparing? Some computers are faster, newer, more powerful, with a video card but more expensive. The second one is correspondingly less powerful, slower, without a video card, but cheaper.
There is money - take a newer one, no money - for the first (and long) time you will completely manage with a cheaper product.
Success in the study of machine learning is primarily determined not by the power of the computer used, but by the power of the brains of its owner.
PS However, you already have one laptop:
Why can the touchpad sometimes ignore clicks on a new laptop?
why do you need another one?

Answer 3

You do not need your resources for the basics of ML at all. Clients need to build models on AWS, google, azure, kaggle, etc., etc. platforms, and each of these platforms has a free plan, where there are more than enough resources so that those who are learning can learn , and for commercial projects, clients will then pay for the plan.

Answer 4

Hmm, take this handkerchief. More than enough for tests
https://aliexpress.ru/item/1005001709727911.html

Answer 5

The second option with an NVidia video card is better, especially if you run MatLab (you can use an inexpensive Home license). There are options for running parallel computing using the CUDA library.