The technology behind ChatGPT holds immense potential that may exceed all our expectations. What is coming next in the world of artificial intelligence will make chatbots look like just a simple beginning.
The AI revolution goes beyond conversation
The world is set for massive changes across multiple fields thanks to the advancement of artificial intelligence. From developing plastic-eating bacteria to save the environment and producing new drugs to treat diseases like cancer, to robots that assist us in our daily lives and fully self-driving cars, all of these innovations rely on the technology that serves as the foundation for applications like ChatGPT, yet they go far beyond mere chatbots.
What is transformer technology?
Transformer technology was first introduced in a research paper by Google researchers in 2017. This technology allowed computers to understand the context and deep structure of any dataset, enabling them to predict upcoming information. The technique relies on understanding the relationship between each part of data and other parts within the same set.
The evolution of artificial intelligence has helped launch amazing applications. For instance, Waymo is using transformer technology to improve its self-driving cars, while Evolutionary Scale uses the same technology to design new protein molecules that could open unprecedented horizons in medicine and chemistry.
Wide-ranging impact
Since the launch of the original research paper, transformer technology has been cited in over 140,000 research papers, highlighting its significance. Its impact spans multiple sectors, such as the tech industry, where Nvidia has become among the most valuable globally due to its adoption of this technology in computer chips. It has also contributed to the creation of new data centers, leading to increased energy consumption and higher costs.
From translation to general understanding
Transformer-based AI evolution is a significant advance because it moves beyond the traditional capabilities of earlier systems that relied on linear data analysis without the ability to grasp full context. For example, a transformer system can accurately translate a sentence like “I reached the bank after crossing the river,” understanding that the word “bank” refers to the riverbank rather than the financial institution, based on context.
This deep understanding of relationships between data parts allows these systems to predict next steps more accurately, opening the door for entirely new use cases.
Applications in scientific research
In scientific research, companies like Evolutionary Scale use transformer technology to study protein structures. For example, the company succeeded in developing a new protein that performs functions similar to those that make jellyfish glow, but with a completely different structure. This type of innovation could lead to designing new molecules that help develop targeted treatments for specific diseases or produce enzymes capable of breaking down plastics.
Transformer technology was also the foundation for developing AlphaFold, the Google-developed system for understanding protein structures whose development team was awarded the 2024 Nobel Prize in Chemistry. These achievements open the door to new uses, such as inventing unprecedented materials or designing drugs that target cancer with precision.
Robotics: From conversation to action
In robotics, startups are taking transformer technology to a new level. Physical Intelligence aims to develop a universal AI model that can control any type of robot to perform any task. For example, the company managed to train a robotic arm to fold laundry without human intervention other than providing initial data.
Autonomous vehicles
In the field of self-driving cars, transformer-based models are used to equip systems with broader cognitive capabilities, allowing cars to recognize objects they have not been trained on before, such as animals crossing the road. Companies like Waymo and Nuro are working to improve this technology to develop safer and more efficient vehicles.
Limits of artificial intelligence
Despite tremendous progress, artificial intelligence systems still face challenges. These systems rely heavily on available training data, meaning they can be limited by the quantity and quality of such data. Furthermore, full reliance on AI for functions like autonomous driving remains far off due to the need for more testing and safeguards.
The beginning of a long journey
While AI evolution has shown great potential, the road ahead is still long. Developing systems to suit real-world environments requires massive investments and a great deal of time. Nevertheless, continuous progress in this technology indicates that we are on the cusp of a true revolution.
