In a recent blog post, Meta founder Mark Zuckerberg shared his read on where artificial intelligence is headed. He calls it the superintelligence era, and he’s confident it’s only a few years out.

He also took the time to spell out what a ‘Positive AI’ future would look like.

Getting there comes with plenty of caveats. Still, the fundamental principles and elements Zuckerberg lays out line up closely with the strategies other leading AI firms have put forward recently.

Invention will be the greatest contribution of superintelligence

Zuckerberg believes people will be able to use superintelligence to create and discover extraordinary things. With easier and affordable access to more AI tools, every person will become more capable of shaping our collective future. As a result, superintelligence will be synonymous with invention and not automation; especially considering the incredible progress being made by new, more capable AI models in different areas of research. Rather than centralizing superintelligence, Zuckerberg is confident that the next era of personal empowerment will be brought about by AI tools that help individuals to pursue their interests, improve the quality of their lives and support them in reaching their full potential. In fact, he underscores the need to distribute superintelligence rather than enabling extreme concentration of power; alluding to several instances in the past when the latter has led humankind to negative outcomes.

Empowering individuals with superintelligence is vital

Be it a personal AI agent that works 24/7 to improve the life of its user or an AI coding assistant that can convert ideas into deployable code, Zuckerberg is adamant that personal superintelligence will serve to improve all of our lives in the long run. He also remains confident that with the help of powerful new tools, people will be able to create new businesses and generate employment as a result. In terms of learning and development, both students and adults alike will be able to learn new job skills, languages, hobbies from their own personalized AI tutor; delivering knowledge with unlimited patience.

While talking about how Biohub has already shipped open-source biological models that are helping scientists make new discoveries and design new drugs, Zuckerberg asserted that superintelligence will help unlock personalized therapies and expedite trials in the near future. Beyond biology, the Meta founder is confident that AI will advance research in every field and contribute immensely to society.

Even though he admits to new technologies bringing more risks, his approach remains grounded in achieving the right balance of power; similar to how western countries have achieved through democratic governing institutions. People and institutions will naturally check each other out; together building a positive AI future by delivering superintelligence to everyone.

Achieving a balance of power that favors individuals

In order to create a positive future for everyone, it is vital that we strike a balance of power by ensuring that superintelligence is broadly distributed to empower people and always favors individuals.

While Zuckerberg outlines this central thought process several times in his blog post, he also admits that law enforcement and military will have more compute to generate more intelligence from the same models. This path allows individuals to freely use AI tools while affording defenders the edge to protect against cybersecurity, bioterrorism, and other threats. Thus, scientific and cybersecurity capabilities of AI models and tools should be enhanced rather than being restricted in the future.

In the superintelligence era, he argues that by giving defenders more compute to generate significantly more intelligence from the same models, they will be better equipped to keep bad actors in check. Zuckerberg also believes that as AI models are fortified with strong cybersecurity capabilities, AI systems and the entire ecosystem will move towards being more secure. Towards this end, he also implores other AI firms that are developing frontier AI models to commit significant technical resources towards helping the government harden critical infrastructure. Likewise, he proposes that leading AI labs should share intermediate training checkpoints of new models for government use and review rather than waiting until training has completed.

Building sustainable AI infrastructure with community participation

Recognizing the fact that the future of AI is largely dependent on the supporting infrastructure necessary to drive user adoption, the Meta founder also highlighted how important it is to pursue sustainable infrastructure development.

While referencing a key data center project being built in Louisiana, Zuckerberg highlights how Meta has inadvertently created demand for high-paying local jobs even as it makes sizeable investments towards enhancing public services. On the environment front too, the company is helping keep electricity prices low by incorporating energy-generating infrastructure and has also committed to being water-positive by 2030.

This approach of integrating sustainability in every aspect of the AI infrastructure being built is what Zuckerberg believes can help America quickly bring online the energy capacity and physical infrastructure needed to sustain USA’s AI competitiveness. Ultimately, he does also advocate for some contentious policy action to lessen USA’s disadvantage when compared to countries like China, making a part of his blog quite political in nature. Nevertheless, Zuckerberg’s call for greater collaboration between all players in the AI ecosystem has also been voiced out by Sam Altman, especially to buttress AI safety before AI agents run amok with their advanced cyber capabilities as seen in the Hugging Face incident.