Due to the much larger number of parameters in GPT-4

[suchona.kani.z]

In spring 2022, Google released the Pathways Language Model (PaLM), an application that achieved very good results in solving hundreds of tasks without being explicitly trained for it. Shortly afterwards, Gato was introduced. This is an application from DeepMind, a subsidiary of Google known for having built an AI that beat the best Go player in the world. Gato is an AI that can solve 600 different tasks, including playing computer games, stacking building blocks and Lego bricks, or creating images.

The application possibilities for such general models are unlimited. They fall precisely into the AI ​​Act planned by the European Union, because it is necessary that they are registered with the authorities and regularly checked. The problem is that such applications have capabilities for image and speech recognition, audio and video generation, pattern recognition, question and answer recognition and translation.

A forecast could be as follows: These applications may be able to complement iran consumer email list humans perfectly in the future and will probably be indispensable in the future. For example, an AI could be used as a digital assistant to a doctor and carry out all administrative tasks automatically. This would give medical staff more time for patients.


The year 2022 was full of interesting and new AI developments and the field is developing at an incredible speed. And it remains exciting: OpenAI GPT-4 is scheduled to be released this year - a huge step forward, as the visualization of the number of parameters shows:


Difference between GPT-3 and GPT-4, Source: tinykiwi

it will probably outperform ChatGPT and GPT-3 and have significantly better performance. Of course, this also comes at a much higher cost to the environment and users. The problem with these networks is that they require a lot of energy and time to train.

The AI ​​language model Bloom produced 50 tons of CO2 during its training. For comparison: one ton of CO2 is equivalent to 3,300 kilometers in a gasoline car, a flight from Frankfurt to New York or 8,800 cups of coffee. It was trained for four months on 384 A100 graphics processors (approx.