Why do people use AI / ML interchangeably when the really cool stuff is DL? ML, DL and Accelerated Computing are all part of AI..
Discussion
DL?
Deep Learning. That is the one that imitates the neural network learning style of the human brain.
Machine Learning learns differently, basically by trial and error, usage.
Trial and error is Reinforcement Learning. Machine Learning includes Supervised Learning, Unsupervised Learning, Self-Supervised Learning and Reinforcement Learning (these are types of problems). And any of these can be solved using Deep Learning or other methods.
Artificial intelligence is a umbrella term for all neural networks including DL.
Yes, exactly, but since this last AL craze, I see a lot of ‘AI/ML’ even when referring to their AI in general..
It's even more than Neural nets.
Agree and the most important component is data. It's a new oil.
I did a PhD in the field. And now I'm bored by it and want to focus on Bitcoin 🫠
That's great to hear! I'm also working in this field mainly recommendation systems and image recognition. I'm sure I'll have a lot to learn from you.
Good examples!
Most recommendation systems will use ML, most image recognition will use DL, it’s a blast, enjoy!
Ex. Supercomputers can now train on millions of images of cancer, and can tell you if a tumor or lesion is cancerous or not, with >90% accuracy 🤯
Greatful 🫡
Correct. Deep Learning is the neural net one, I think that is some of the coolest tech.
Deep Learning is training a computer, likely using GPUs, with millions of points of data, great artists, what plants are, or cancer looks like, etc.
Machine Learning is the machine learning patterns over time, ads targeting people, Google search results, etc.
Mathematically, DL is a subset of ML. ML is a subset of AI.
Also any ML model that involves neural nets can be termed as DL if the number of neural nets present in the hidden layer is really big that such model is now capable of predicting the underlying deep and intricate features present in the given test datasets
Generally AI and DL models are separated when describing them. Normies don’t even know what ML DL even is, much less technicalities if a subset.
When talking AI, if you say it was trained using ML, it suggests it wasn’t using DL.
Wikipedia also agrees with what I said
I agree with you also, technically correct, but it is not how people talk about them, or necessarily the best marketing to always say AI/ML.
We might as well say; AI / ML / DL..
Theoretically, but in practice they are different things when describing types of AI. When someone says, “This was trained using ML” means it didn’t use DL.
Hey, #[2]. 👆🏼
They’re all subsets of AI. I think Deep Learning has been the “cool” and popular one for many, many years. Since the “Attention is All You Need revolution”, there’s been an apparent need to differentiate. Transformers are DL, but they work much better than the now classic CNNs, RNNs, etc. Transformers is probably too overwhelming a term to be the popular label, so AI/ML it is.
Most don’t even know the difference between ML/DL, even less that one is a subset of another. Still seems adding AI/ML leaves out DL. Why not just call it ‘AI’ since that encompasses everything anyway? AI/ML if their discipline is ML I get it, but for general marketing?.. morning rathole