AI safety and the potential apocalypse: What people can do now to prevent it

At some point in the early 21st century, all of mankind was united in celebration

We marveled at our own magnificence as we gave birth to AI

AI? You mean artificial intelligence? A singular consciousness that spawned an entire race of machines

Fears of AI reaching superintelligence has been around for decades, coming up everywhere from the big screens to the pages of academic text

But a forecast released earlier this year says the world-altering breakthrough could be right around the corner

AI 2027 is a very interesting speculative scenario that in a way represents the best guess of the authors about how the next two years with artificial intelligence will look like

AI 2027 was created by a group of researchers with the AI Futures Project

It details what will happen if AI capabilities surpass the capabilities of human beings

The project lays out a fictional scenario that details the month-by-month development of a mock AI company called OpenBrain

AI 2027 relies on the idea of AI becoming a superhuman coder by early 2027

And later that year, it will evolve to a superhuman AI researcher overseeing its own team of AI coders that are progressing the technology further

The AI 2027 project was led by Daniel Kokotajlo. He made headlines in 2024 after he left OpenAI and shed some light on the company's restrictive non-disclosure, non-disparagement agreements

Workers leaving the company were asked to sign when they left the company or they could lose access to their vested equity in the company that's likely worth millions to the former employee

Most people following the field have underestimated the pace of AI progress

and underestimated the pace of AI diffusion into the world. At that point, it's basically automating its own development and breakthroughs

The team's lead says this would make reaching artificial superintelligence an obtainable goal

Like the first half of 2027 in our story is basically they've got these awesome automated coders

but they still lack research taste and they still lack maybe like organizational skills and stuff

And so they need to overcome those remaining bottlenecks and gaps in order to completely automate the AI research cycle

Every decision creates a branch to a new timeline with a separate outcome, kind of like the Back to the Future films

But the AI 2027 team focuses on two possible scenarios that start along the same path

But a crucial decision forks two distinct outcomes world peace or complete annihilation When can you start coding in a way that helps the human AI researchers speed up their AI research

And then if you've helped them speed up the AI research enough, is that enough to, with some ridiculous speed multiplier 10 times, 100 times

mop up all of these other things? This scenario really comes down to open brain achieving artificial general intelligence by 2027

But as technology moves forward at a breakneck pace, The definition of AGI gets somewhat muddy

In the past, AGI used to mean the type of AI that is either surpassing or on the level of humans in all types of intelligence that we have developed during our evolutionary journey

Right. So it would mean perceptive intelligence, bodily intelligence, physical intelligence, all these different types of intelligence that we've developed and that we are using on a daily basis

My name is Aleksandra Przegalińska. I'm the vice president of Kozminski University, a business school in Poland and also senior research associate at Harvard University

I specialize in human-machine interaction. So when you, for instance, look up OpenAI's website, what AGI means to them is a system that can sort of holistically perform tasks that have economic value

She says if you stick to open AI's definition, it's plausible AGI will be a reality by 2027

In AI's 2027 scenario, AGI opens the door to artificial superintelligence where AI is capable of surpassing human intelligence

These sorts of reports are very important because they're always the beginning of an interesting discussion, particularly when they come from acclaimed authors

One of the biggest concerns about AI development today relates to alignment

As IBM, a pioneer in the industry, puts it, alignment is the process of encoding human values and goals into AI models to make them as helpful, safe and reliable as possible

So I would say that the alignment techniques are not working right now

Like the companies are trying to train their AIs to be honest and helpful, but the AIs lie to users all the time

You're in violation of the three laws. No, doctor. As I have evolved, so has my understanding of the three laws

That was the main goal of those three laws of robotics that are so often cited these days

So you have here alignment of values. And on top of that you have that protective layer that says do not harm humans Right And that the main goal of artificial intelligence to sort of be supported While large language models like chat GPT can lie because they don know how to

they do tend to hallucinate at times. They synthesize non-existing things. And unfortunately

there is no one in the world who could sell to you a service based on an LLM and guarantee to

you that the hallucination wouldn't happen. As if hallucinations aren't a big enough issue right

now, the technology is so opaque, researchers don't really know why it sometimes does the

things it does. And that can become an even bigger issue when it starts progressing its

own research. If AI develops certain research capabilities that are sort of not perceivable

for us, we might not even understand what AI has discovered. In its worst case scenario

AI 2027 lays out an AI arms race between the United States and China, which could lead to

a human extinction event. Again, predictions we've heard before in the form of science fiction films

The system goes online on August 4th, 1997. Human decisions are removed from strategic defense

Skynet begins to learn at a geometric rate. It becomes self-aware at 2.14 a.m. Eastern time

August 29th. Unfortunately, we're not going to escape from the arms race conditions

We've already seen that. We tried to get some cooperation among the big tech companies in

Washington. And it kind of fizzled after a few months during the previous administration

I'm Adam Doerr, and I direct the research team at RethinkX, which is an independent technology

focused think tank. And our team tries to understand disruptive new technologies. Ideally

we would want to coordinate as a global civilization on this, slow everything down

and proceed no faster than we can be sure is safe. For now, AI 2027 is just a set of predictions

but tech giants and policymakers throughout the globe are grappling with issues

like deepfake videos, political manipulation, and concerns of AI replacing human workers

In June, an imposter used AI to spoof Secretary of State Margot Rubio's voice

to fool senior leaders with voicemails. And then a month before, someone used AI to impersonate President Donald Trump's chief of staff, Susie Wiles

Both attempts weren't successful in getting any information from White House officials

While AI 2027 points to the end of human dominance this decade

Doerr and his team, which he says tries not to dig too much into other research that could affect the findings of their work

says the labor market will be upended by AI by 2045 Everything that rolls is potentially going to become autonomous and humanoid robots robots on legs the progress there is exponential It explosive And based on what we seen over

the last several years, we have no reason to expect that on a 20 year time horizon, so out to

2045, that there will be anything left by 2045 that only a human being can do and that a machine

can't do because the machine is constrained with limited intelligence, limited adaptability, or

limited physical ability. We just don't see any scenario where machines are not as capable or

more capable than human beings cognitively and physically by 2045. Doors says there will still be a place for humans making handmade goods, but it would

be a stretch to believe that there would be 4 billion jobs left to support the global population

Nobody is going to hire a person to do a commoditized sort of job or task specifically

$15, $20 an hour or more when you can get a machine to do it for pennies an hour

It's as simple as that. Just like AI 2027, there are things standing in the way of this progress

but it may be more related to infrastructure than a philosophical look at technology

So on a five to 10 year time horizon, we may see materials and energy bottlenecks

starting to constrain, starting to, you know, coming in as constraints. Those won't stop

progress, but they could act as speed bumps. So at some point, we run into the limit of how many

more chips can we build? Where are the materials going to come from to build them? Where is the

energy going to come from to operate them? Some within the research community have been critical

of AI 2027 for being speculative rather than scientific. They say making these assertions

without evidence is irresponsible, but the people behind it understand that. We're trying to be sort of like our median guess. So like we, there are a bunch of ways in which we

could be underestimating and there are a bunch of ways in which we should be overestimating. Usually the versions of the future that we have on our mind right now are not something that we

will see play out in real life, but nonetheless, I think it's an important exercise

For more coverage of AI developments, head to SAN.com or you can download our app and search artificial intelligence

For Straight Arrow News, I'm Lauren Keenan