Category Archives: Society

Artificial Intelligence, Society, Technology

AI is spiralling out of control: it can be stopped

ARTIFICIAL SUPERINTELLIGENCE

Intro: East and West collaborated to end nuclear proliferation – it is time to do the same for the latest advancing technology. Washington and Beijing must come together to rein in AI’s growing threat

After the Cuban Missile Crisis brought the world to the edge of nuclear war, global powers embarked on a concerted effort to pull it back from the brink. The non-proliferation treaty (NPT) of 1968, which limited the spread of nuclear weapons, has been a resounding success. Only a handful of countries today have access to the 80-year-old technology and those that do have not used it.

In the decades since, no technology has proved as dangerous as nuclear weapons as to require international co-ordination.

Now, however, many believe that the advance of artificial superintelligence requires a similar global effort to prevent an AI-led disaster.

Anthropic, the world’s most valuable AI company, has called for a mechanism to slow down or pause the development of advanced AI. It has warned that the technology could get out of control sooner than many think.

The company believes it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology. It says it would “likely be a good thing” if development could be delayed.

Anthropic – recently valued at $965bn (£720bn) – said it had raised the alarm because it believed AI was improving much faster than our ability to understand and control the systems.

Within the company itself, bots are not just writing code; they are also ordering around other bots and even carrying out their own research. Before long, AI could be building itself, a process called recursive self-improvement. This could start a feedback loop in which progress goes parabolic.

Sceptics insist this is just mere marketing. Anthropic has announced that it has filed for an initial public offering and is expecting a value in excess of $1tn. What could be more valuable than a technology so powerful that world leaders need to rein it in? AI that builds itself has been a premise the company has used to raise money for years.

David Sacks, a high-profile critic of Anthropic, and Donald Trump’s former AI tsar, suggested the warning was an attempt to secure a public bailout, implying it was a sign of getting the frontier AI lab nationalised.

Nonetheless, concerns about powerful AI are becoming increasingly prominent. Anthropic has kept its most powerful AI system, Mythos, out of public hands because of its ability to find security flaws in critically important computer systems.

Andrew Bailey, the Governor of the Bank of England, has raised the alarm about AI crashing the financial system and has warned that Mythos meant “things that we thought might happen in the next year, two years, three years or four, have now come right into the foreground”.

AI labs fear that the next generation of models will be good enough to help terrorists develop bioweapons.

If AI were to start building itself without human oversight, it would by definition become much more difficult to control. In the extreme scenarios that safety experts are concerned about, AI’s goals become detached from our own, forcing it to eliminate humanity through evolution so that we do not get in the way.

There are those who dismiss this idea as sci-fi nonsense. But supporters of a pause say even a tiny chance of extinction should be enough to make us consider how to stop it.

Establishing the need for a pause would be the easy part. Making it happen is another matter altogether. If he so wished, Dario Amodei, Anthropic’s chief executive, could send everyone home today and shut down his company. At best, though, this would delay the rise of powerful AI by a couple of months. Its two major rivals, Google and OpenAI, are not far behind. OpenAI, the developer of ChatGPT, has said that it too sees “early signs of RSI [recursive self-improvement] in today’s systems”.

It added: “We expect this to increase competitive pressures among developers and nations, and create governance challenges that existing institutions are not equipped to address.”

Even if the US government ordered all three to stop work on AI, this might only cede ground to China, whose companies are typically seen as being just three to six months behind the US.

Earlier this year at the World Economic Forum, Amodei said: “The reason we can’t [slow down] is because we have geopolitical adversaries building the same technology at a similar pace… It’s very hard to have an enforceable agreement where they slow down and we slow down.”

Practically, it would require a government-level agreement and the two nations that matter are the US and China. This sort of agreement would require Trump and Xi Jinping to co-operate on a pause, something that looks far from likely given both have compared AI to a race.

Xi has said that China must “gain a head start and secure a competitive edge” in AI, while a Trump administration action plan states that “America is in a race to achieve global dominance in artificial intelligence”.

It has also emerged that the National Security Agency have been using Mythos to carry out cyber-attacks. This suggests the US government is making enthusiastic use of the latest systems instead of fearing their consequences.

Pessimists often compare the technology and its potential consequences to nuclear weapons, but the two are nothing alike.

The destructive capabilities of atomic warheads are undisputed, whereas AI’s safety risks can appear nebulous. The latter’s upside may also be significant: its supporters believe it can cure disease, lead to interstellar space travel, and make work optional.

What is more, pressing pause on the AI race is not without its own set of risks. Suspending work on AI could cause an economic crash. The chips and data centres that AI relies on have driven a stock market boom that has helped sustain the US economy. Inhibiting demand for them could do the opposite.

There have been signs that China and the US are changing tack. The White House has raised the alarm about Mythos and Trump has just signed an executive order calling for AI models to be reviewed before release.

Beijing has called for a “global AI governance framework” to rein in the technology. This is miles away from the global deal Anthropic has called for, but campaigners have taken it as a positive sign.

The political zeitgeist can move very quickly. The US and its allies have succeeded to a certain extent in deterring nuclear proliferation. To do so similarly with AI is going to be hard, but as we have seen with nuclear weapons, global governance can come together and work for the common good.

Standard
Artificial Intelligence, Britain, Government, Politics, Society, Technology

Proposed ‘kill switch’ for AI data centres

CYBER SECURITY

UK politicians are pushing for an AI “kill switch” that would allow ministers to shut down data centres.

Campaigners are seeking new laws that would give the Government powers to switch off AI systems in the event of a “catastrophic risk”.

A proposed amendment to the cyber security and resilience bill has the backing of at least 11 MPs, and is part of a coordinated campaign from Control AI, a group calling for strict AI regulations. The plans have not been endorsed by the Government, but demonstrate growing concerns about Artificial Intelligence among Members of Parliament.

Donald Trump, too, has recently expressed support for a kill switch and told Fox News that there should be government powers to shut down AI.

The amendment, proposed by Labour’s Alex Sobel, would give the technology secretary “last resort powers” to direct the shutdown of data centres “in the event of an AI security or operational emergency”.

The powers would come into force if there were a “catastrophic risk” to critical infrastructure, national security, or “severe, large-scale harm to human life”. Data centre operators would have to install infrastructure allowing them to be stopped instantly and establish secure communications to the Department of Science, Innovation and Technology to enable ministers to act.

Meanwhile, Dario Amodei, the Anthropic chief executive, is expected to meet a group of 50 top European chief executives at a two-day forum to discuss AI adoption across the private sector.

Standard
Artificial Intelligence, Research, Science, Society, Technology

Superintelligent AI and its threat to humanity

ARTIFICIAL INTELLIGENCE

Intro: Humanity faces an uncertain fate as experts brace for superintelligent AI. The tech industry claims looming “singularity” will change everything

Every time one of the world’s top artificial intelligence companies unveils a new system, employees at the US research organisation METR put it through its paces. Its ability is tested to complete a series of increasingly complex tasks.

The tasks are measured by how long each one would take a skilled human. They range from trivial arithmetic (two seconds) and completing a game of Wordle (13 minutes) to building complex military satellite software (taking a human expert 14.5 hours).

The test then serves as a gauge as to how capable AI has become – and where it might go.

The first version of ChatGPT, released in 2022, could only perform simple tasks that would take a human a few seconds.

But as AI systems have become more powerful, they are able to complete more complex actions that would take humans hours or days, such as breaking into a medical website and downloading all its data.

METR has found that AI capabilities are doubling in power every 196 days. Plotted on a graph, this progress starts slowly then rapidly accelerates to a near-vertical plane.

Converse with anyone in the AI industry for any length of time and the likelihood of them pulling up a version of the chart approaches 100pc, to the point where it has become a meme in its own right. It is being referred to as the most important chart in the world. The chart goes off the scale.

Last month, the AI lab Anthropic announced it had developed a new system, called Mythos, that it said was too powerful to release to the public because of its ability to find gaping holes in online security systems.

When METR’s researchers released the results of Mythos’s capability and function, they scored the system at 16 hours – meaning the world’s most powerful AI can now automate tasks that would take a human two full eight-hour shifts.

Nonetheless, they said the model was “at the upper end” of their ability to test. In other words, progress has become too fast for them to measure.

Not everybody is convinced by the results because the test only measures if a machine can do something half the time, not if it can do it consistently. The METR chart has, however, captured many people’s imaginations for two reasons.

First, the exponential growth looks strikingly similar to “Moore’s Law”, the maxim that has governed the electronics industry for more than half a century, stating that microchips roughly double in power every two years.

Second, it measures abilities, rather than intelligence. While many AI “benchmarks” resemble university exams and gradings, dealing in abstract reasoning or maths, the METR test studies whether AI can actually work.

It suggests that on current trends, vast amounts of human tasks could be automated in the next couple of years – including, most crucially of all, the art of developing AI models itself.

At that threshold, known in the tech industry as “recursive self-improvement”, all bets are off.

The concept is closely linked to superhuman AI because an AI that can make itself smarter could act like an evolutionary chain reaction, rapidly building to a system vastly more capable than mankind.

AI would have become – as IJ Good, the Bletchley Park codebreaker, predicted in 1965 – “the last invention that man need make”. Almost Orwellian in thought.

For 60 years, the idea seemed out of reach. But much of Silicon Valley believes this is about to change – and the US government is starting to notice.

The vast majority of people’s experience of AI has not changed much in the last couple of years. The release of ChatGPT in 2022 generated an initial flurry of excitement and fear in equal measure but, since then, progress has been less obvious.

The AI experience for many people comes in seeing an obviously fake video on their social media feeds, seeing an AI overview at the top of their search results, or having a bot that “helpfully” offers to summarise their emails.

But at the coalface, people are rapidly bringing forward their timelines for the day that superintelligence arrives.

Pages: 1 2 3 4
Standard