Are We Ready For AGI? According to the AI Safety Clock, No.

-

The new AI Safety Clock launched by IMD to assess AGI safety shows that we’re 29 minutes to midnight. What does this mean for us?

The AI Safety Clock measures the risk of an uncontrolled AGI; Source: IMD

Remember the Doomsday Clock? Yeah, researchers at IMD have now developed an AI Safety Clock, set to 29 minutes before midnight; 29 minutes to the point where Uncontrolled Artificial General Intelligence (UAGI) could pose a significant threat to humanity. 

But what does all this even mean? 

Generative AI already significantly impacts our everyday lives—from large-language models (LLMs) to social media algorithms to robots being developed to relieve the human workforce in critical tasks. So far, AI is manageable. But AI is also becoming more powerful by the day. 

Governments and AI developers across the globe are now preparing for a future of AI in the form of Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI). An AGI differs from the GenAI we have at the moment in that it would be able to be as good or even better than humans in most tasks. 

AGI is the next great milestone expected to be achieved in AI research. Meanwhile, ASI—better than any human—is still far away. And nobody knows for sure when AGI will be reached. A decade ago most AI researchers agreed that AGI would be achieved by 2075. 

With every year, this estimate has come closer to the present time. 

By now, the general consensus for the arrival of AGI is before 2050. And some believe AGI might be closer than expected, with estimates ranging from as far as the mid-2040s to within this decade. 

Whether or not researchers are being too optimistic, it’s clear that AGI lies ahead of us. However, as the AI Safety Clock shows, we are not ready for it. 

There are a lot of problems that our current, narrow AI faces—discrimination, bias, privacy. But the problems an Uncontrolled AGI (UAGI) could present are worse than that. 

As one of the researchers who developed the AI Safety Clock, Michael Wade, wrote for Time: an AGI that works independently without human interference could decide to deploy military weapons or start large-scale misinformation campaigns. 

We already saw the power of AI during this year’s presidential election in the US. Deepfake photos and audio have been used to paint wrong pictures of elections and drive opinions to a certain direction. 

Feels familiar? The same techniques used during the 2016 US Presidential Election are exploited all over again—and it has become much more difficult to distinguish between what’s real and what’s fake; the line between reality and cyberspace keeps shrinking. 

But all these misinformation campaigns have been instiged by humans. A UAGI would do this without first having to ask for human permission; it simply acts. 

AGI offers a lot of opportunities to improve the world we live in. Controlled AGI would benefit us all, but a UAGI will make AI indeed humanity’s last invention. 

But let’s not be too pessimistic. 

As the past has shown, researchers have long been over-optimistic about the development of AI, and if we manage to raise an AGI that is kept within a fence, we will win once more. 

Thank you for reading my article! If you would like to support a young writer, check out my Ko-Fi!

Comments

Post a Comment

Popular posts from this blog

Europa Clipper is On Its Way to Jupiter: This is How the Mission Will Change Our Understanding of Life Forever

-
Image
As Europa Clipper moves farther away from Earth on its way to the Jupiter moon Europa, let’s take a look at some of the NASA flagship mission’s ways it’ll change our understanding of life forever.  Falcon Heavy taking off from Kennedy Space Center with Europa Clipper on board; Source: Spaceflight Now Jupiter has 95 known moons, with the four Galilean moons being some of the most interesting worlds of the Solar System. But scientists are particularly interested in Europa, a moon slightly smaller than the Earth’s Moon.  What makes Europa so interesting isn’t its surface, but what lies below, as scientists believe that it harbors a global subsurface ocean. And what Earth has taught us best is that where there’s water, there’s life.  This brings us to the possibility that life exists in the oceans of Europa. Naturally, scientists want to take a closer look to establish whether or not Europa could be habitable—and answering this question will forever alter our understanding...
Read more

Billions of Stars Travel Through the Intergalactic Medium. Why Were They Expelled From Their Galaxies?

-
Image
  All stars are born within galaxies. However, there are also lone travelling stars in the emptiness between galaxies. No one knows for sure why they are there, but one thing is certain—only violence can tear stars from their homes.  The Antennae Galaxies, two merging galaxies. It's possible for stars to be ejected during mergers; Source: Wikipedia In outer space, stars and galaxies like to stay as close as possible due to their gravity. Stars accumulate to galaxies like the Milky Way, and galaxies group to clusters. The closest larger galaxy cluster is the Virgo Cluster, prominent among amateur astronomers for its relatively bright galaxies such as Messier 87.  But galaxy clusters aren’t the ultimate; clusters are just small structures on the very large scale, where they grow to superclusters. For example, the Milky Way is part of the Local Group, which in turn is part of the Laniakea Supercluster.  Superclusters reside in galaxy filaments — the largest known str...
Read more