Even if you do not write code, meeting you will take you one step further, and will help you understand artificial intelligence more properly.

Here are the use of artificial intelligence, which is not so ringing of more than one person:

Artificial intelligence terms:

• Federative learning
• Don’t forget a disaster
• Flush
• Gradient loss
• Reinforcement learning
• Attention mechanism
• Mode collapse
• Zero -firing learning
• A few firing learning
• Hallucination
• Neuroievrim
• Herd intelligence
• Transfer learning

Federative learning

If the data closure is valuable, it is necessary to know this term. Federated Learning, ie federative learning, provides learning on devices without collecting information in the center.

So the information remains on the device, the model is trained there. Even Google’s keyboard suggestions use it. Both the user data is protected and the model is developing.

Don’t forget a disaster

Have you ever heard of forgetting the old ones while trying to teach something new to an artificial intelligence model? It is called “caattrophic forgetting .. In particular, it is the trouble of always learning systems! We can also say that artificial intelligence is “fish memory ..

Flush

How to convert a word, photo or user to number? Embedding transforms complex things into vectors in a way that the machine understands.

For example, it is represented by numbers “cat” and “dog” because they are close to mana. Unknown weapon of content proposal systems.

Gradient loss

A common problem during training in deep border networks. The things that the model should learn does not reach “damping” in the layers, ie the system cannot learn. If this problem had not been solved, deep learning could not have progressed so much today.

Reinforcement learning

We can also say “award” based learning. Artificial intelligence takes an action, the award or punishment compared to its conclusion. He learns the truth with time with this cycle. Artificial intelligence playing games and controlling robots is trained in this way. Patience is literally.

Attention mechanism

Which part of a text is precious? Artificial intelligence makes this decision with attention system. This structure teaches the model “what to consider what”. Chatgpt plays a major role in the success of language models. You learn as much as you pay attention, right?

Mode collapse

It is a common condition in Generative AI (YZ) models. The model always starts to produce just like or for example. So the diversity ends, production becomes uniform. In particular, a problem with headaches (productive networks).

Zero -firing learning

We can also say that the model has never seen a mission. Yes, without any example! This is a valuable step on the road to general artificial intelligence. The model adapts to a new mission based on old information. This is exactly that to produce analysis from scratch.

A few firing learning

Instead of giving thousands of examples to train the model, you want him to learn the job with a few examples? This is exactly that’s Few-shot Learning. One of the cornerstones of artificial intelligence learning for human beings. It comes into play here to achieve a lot of work with less data.

Hallucination

The fact that the language models are unrealic, but the fact that it sounds wrong is explained in this way. For example, artificial intelligence, a non -resource or information “can make up.” We frequently encounter models for chatgpt. It is so close to realism that sometimes it is difficult to distinguish the difference.

Neuroievrim

A name given to the process of optimization of neural networks with evolutionary algorithms. In other words, artificial intelligence models are developed with a process of natural selection. In particular, it is used in cases where classical learning systems are insufficient.

Herd intelligence

As you can imagine, many easy units are inspired by the herd behavior in nature, so that complex problems are solved. Ants use this procedure in finding food while artificial intelligence, optimization problems.

Transfer learning

Our last term is transfer learning. The information learned in one area is transferred to the other area. An artificial intelligence model can use this information to recognize dogs after learning to recognize cats. On the one hand, time and resource savings are provided in this way.

You knew how many of these phrases? We do not include, but if you have to add, we are waiting for comments.

For more about artificial intelligence: