A Star Trek-style future, in which
Artificial Intelligence systems will be connected to each other
in a similar way to the Borg hive mind: this is how a group of researchers describes a possible future development of AI, based on the
most recent
research trends
in this sector.
Described in the journal Nature Machine Intelligence, the scenario is the result of research coordinated by the Italian Andrea Soltoggio, who works at the British University of Loughborough.
"The idea is that
the knowledge acquired by Artificial Intelligence systems can be reused and shared at a global level
, in a
network of many independent units
. This - Soltoggio told ANSA - has not yet occurred", but
the management in which many research groups in the world are working is precisely this
.
"The convergence of these scientific and technological advances - we read in the article - will lead to the birth of new types of scalable, resilient and sustainable artificial intelligence systems".
The
future of AI
would therefore seem
very different
from what one might imagine today
by looking at current large systems, such as ChatGPT, which according to researchers are "expensive, not shareable and not long-lasting".
There is also "a lot of concern about
the environmental impact
of these systems, which consume a lot of energy", observes Soltoggio.
The challenge is therefore to make AI even more sustainable
.
According to the perspective proposed by the researchers, the future will instead belong to
collective and sustainable AI units
,
capable of evolving and sharing information and data
.
In this hive AI, each unit would still retain its individuality and this feature, according to the researchers, could be a
defense against attempts to spread dangerous or unethical information
.
"If something illicit or unethical were communicated by a system, it could be redistributed at a global level almost instantaneously. The solution - says the researcher - is therefore to create
entities that have the ability to distinguish legitimate information from harmful information and to decide on the matter at an individual level
. There will have to be control. The rules do not necessarily have to be global:
when an entity has its own list of rules, a democracy can be created
: I would prefer hundreds of thousands of systems with different rules, compared to a single system with good rules".
The idea, he continues, is "to create a system in which there is no concentration of power in a few companies".
The new direction in which we are starting to work is therefore that of a
collective network of AI units capable of continuously learning and adapting, with learning that continues over time and is updated on the basis of new data
.
It is what the authors of the research define as
Collective Artificial Intelligence
, in which each system communicates its experiences and knowledge to the other.
The challenge
now is "to
demonstrate the possibility of this cooperative system with facts
. Many research groups, including mine, are working in this direction. We hope - Soltoggio concluded - that these examples will be expanded and used".
Diagram of the future organization of Artificial Intelligence according to the vision of the authors of the research led by Andrea Soltoggio, of the University of Loughborough (source: A. Soltoggio et al, Nature Machine Intelligence, 2024)
Reproduction reserved © Copyright ANSA