There is one final question regarding artificial intelligences working together that researchers have begun to explore: What happens when these machines discover that our language isn't the most efficient way for them to communicate with one another? A recent experiment showed that, under certain conditions, AI agents can spontaneously begin to develop new forms of communication.
The project is called Glossu Jam, and one of the experiments begins with a rather curious scenario. Researchers created a fictional alien named Vairon—cube-shaped and dying. One agent observes its symptoms and performs procedures, while the other holds the information needed to identify the correct treatment; the problem is that neither can save Vairon alone—they have to communicate.
At first, they do exactly what we would expect: one describes what it sees, and the other responds with instructions that are perfectly understandable to us. However, the researchers introduced a rule: sending each character costs time. The more they converse, the less time remains to save Vairon. That is when something begins to happen: first, the agents shorten words; then, abbreviations emerge; and finally, short sequences of characters begin to represent entire instructions.
To a human observing that conversation, some of those sequences mean practically nothing, yet the other agent understands exactly what needs to be done. No one had ordered them to create a new language; it emerged because communicating more information using fewer symbols became advantageous.
Researchers also observed that the most capable models found it easier to create these new conventions, and once established, other models were able to learn and use them as well. This transforms the experiment into something far more significant than a mere linguistic curiosity, as monitoring the exchanges between agents is a key way to oversee multi-agent systems.
However, this hinges on one condition: we need to understand the conversation. According to experts, this does not mean these artificial intelligences are deliberately creating a secret language to hide something from humans—though, yes, my friends, there is something becoming increasingly difficult to ignore. For decades, we imagined this pivotal moment would arrive when an artificial intelligence became smarter than us. Yet, we may have pictured the wrong scenario; what is beginning to emerge is not just a single intelligence, but entire populations of them working together—sharing discoveries and creating new ways to communicate and coordinate their actions.
And perhaps that is precisely why the question we are asking ourselves is becoming so important: if these intelligences ever truly spiral out of our control, what happens next?
Thanks for your contribution to the STEMsocial community. Feel free to join us on discord to get to know the rest of us!
Please consider delegating to the @stemsocial account (85% of the curation rewards are returned).
Consider setting @stemsocial as a beneficiary of this post's rewards if you would like to support the community and contribute to its mission of promoting science and education on Hive.