Creativity and imagination are inherent characteristics of human beings. At least, until the consolidation of the technology known as generative artificial intelligence (AI) which uses learning methods that gather information about a specific field and combines all the captured data to generate new and realistic ideas from scratch.
AI tends to break a new barrier in terms of advancement. It will no longer be limited to using huge sets of input data to process it into an algorithm and generate a result, but will also be able to develop original content for a website, write a meaningful text, compose a melody, design a character for a video game, create a painting or set an application.
According to Gartner, by 2025, generative artificial intelligence will account for 10% of all data produced in the world. Today, it’s less than 1%. It also claims that companies that adopt these practices will triple in value relative to those that neglect them. But the future is coming at full speed: the same consulting firm included generative AI in its ranking of the 12 key technologies for this year.
Between deep learning and GANs
Generative AI takes the multi-layered learning model that seeks to mimic the behavior of the human brain using deep learning. Let’s remember that when learning takes place in a higher sphere, it goes backwards to incorporate the new data into the previous ones to improve the learning process.
It also uses GAN (generative antagonistic networks), which consists of pitting two neural networks against each other without intervention or supervision, either human or from another algorithm, so that they compete in an endless game. The first of the networks is called the “generator” and is responsible for “generating” ideas. The second, known as the “discriminator,” verifies whether the content emitted by the other is truthful. Each of the algorithms improves with each interaction in order to beat the other.
Suppose the challenge is to create from scratch a picture of a rhinoceros. The first network will show images related or unrelated to that animal that will be challenged or accepted by the second network until the desired figure is formed. At the end of the process there will be a drawing of a rhinoceros that did not exist previously, created from scratch by these two networks.
Use cases
Generative AI is expected to drive and accelerate some activities, such as research and development cycles in medicine or speeding up the launch of new products.
In IT, it can be very useful to create synthetic data sets to train AI systems. In general, when a company creates these sets based on real-world data, the task can be very cost and time-consuming. Even in some specific fields of knowledge, the total amount of available data may be below the amount necessary for the success of the project. This is the case, returning to medicine, of rare diseases: images from real studies can be added to those created by this type of intelligence to increase the efficiency of the AI model.
Between opportunities and risks
From the point of view of computer security, AI may be able to build dynamic protection models that adapt to shape-changing virtual threats. The opportunity is also the risk: cybercriminals can use it to create new attack models that are not covered by existing defense mechanisms.
Even governments can use this technology to alter photographs and identify fugitives who may have changed their hair color or even had cosmetic surgery. Here again is the risk: generative AI can create photographs or videos of non-existent people and encourage identity theft on networks.
As with any disruptive technology, there is still a long way to go. Governments, companies and end users will have to walk this path with great responsibility, so this technology with almost infinite potential does not generate problems but only solutions.