Generative artificial intelligence is one of the most interesting news of 2023. But how can it be used best? And how does it apply to software?
We hear more and more often about generative artificial intelligence, a branch of AI that is able, starting from an input, to provide different types of output (vocal, written, in video format, etc.), using machine learning algorithms.
Currently, the most widespread models are those that allow the creation of software capable of transposing instructions in written or vocal format. This technology, however, is in constant transformation, so it is possible that new types of models will soon see the light.
One of the best-known software at the moment is ChatGPT, the chatbot created by OpenAI, a company chaired by Sam Altman. ChatGPT can -among other things- provide an answer to many of the questions that the user asks in a short time thanks to its artificial intelligence. To become so "prepared" and quick, the model on which the software is based has been "trained" with a lot of data, which has made it exceptionally effective.
Generative artificial intelligence: usage patterns
There are many models of generative artificial intelligence. Let’s see which are the most popular and which software uses them.
- Text-to-Image (T2I). The T2I is a generative artificial intelligence model that processes an image starting from a textual input. Among the most famous software that exploits this model is Dall-E 2, created by OpenAI.
- Text-to-Video (T2V). The T2V template can be used to create video production software. One of the most popular platforms that exploit this model is Synthesia. The platform allows users to create videos with avatars delivering sentences that they choose through artificial intelligence. The result is a "movie" with an avatar that speaks like a real person. It is great for tutorials and presentation videos.
- Text-to-Audio (T2A). The T2A model is perhaps one of the simplest since it requires the software that uses it to read "with its own voice" what a user writes. Some platforms already offer this possibility, including Google Translate, but thanks to innovation, other high-performing ones have also recently been born. One of the most popular models based on this technology is Resemble.AI, which allows you to choose between different available voices to read a speech. Once you have chosen the one you like best, you can download the audio.
- Text-to-Text (T2T). The T2T is one of the most famous models. ChatGPT, the software that can seemingly answer every question, uses T2T technology. In general, this type of model is capable of receiving a textual input and providing an output that can be of creative nature, but also for example a correction or a translation.
- Text-to-Code (T2C). T2C will certainly be one of the most popular models in the near future, also because of the widespread lack of computer programmers in the labor market. Using this model, a software can receive a descriptive input and, in exchange, return an output code that can be used to program. Obviously, to be effective, the input must be well structured: it must contain fundamental information such as, for example, the programming language in which the output code is desired. One of the software capable of performing this task will certainly be Google Bard, whose capabilities were presented during Google I/O 2023.
- Audio-to-Text (A2T). A software that processes voice inputs or video format (provided the video has audio) and transforms them into written format is based on the A2T model. One such tool is Speechtext.AI, available in both web app and API versions.
- Image-to-Text (I2T). The I2T model is based on a software that allows you to insert an image as input that contains a textual part, such as a photograph of a poster. It also allows you to obtain a file in return (in various formats) with the text visible in the image inside. An example of software of this type is OCR.best, which allows you to insert images with text as input and then be able to download the textual content.
Original article published on Money.it Italy 2023-06-05 10:00:00. Original title: Intelligenza artificiale generativa: 7 modelli di utilizzo
