Skip to main content

DeepMind Details All the Ways AGI Could Wreck the World

1 hour 53 minutes ago
An anonymous reader quotes a report from Ars Technica, written by Ryan Whitwam: Researchers at DeepMind have ... released a new technical paper (PDF) that explains how to develop AGI safely, which you can download at your convenience. It contains a huge amount of detail, clocking in at 108 pages before references. While some in the AI field believe AGI is a pipe dream, the authors of the DeepMind paper project that it could happen by 2030. With that in mind, they aimed to understand the risks of a human-like synthetic intelligence, which they acknowledge could lead to "severe harm." This work has identified four possible types of AGI risk, along with suggestions on how we might ameliorate said risks. The DeepMind team, led by company co-founder Shane Legg, categorized the negative AGI outcomes as misuse, misalignment, mistakes, and structural risks. The first possible issue, misuse, is fundamentally similar to current AI risks. However, because AGI will be more powerful by definition, the damage it could do is much greater. A ne'er-do-well with access to AGI could misuse the system to do harm, for example, by asking the system to identify and exploit zero-day vulnerabilities or create a designer virus that could be used as a bioweapon. DeepMind says companies developing AGI will have to conduct extensive testing and create robust post-training safety protocols. Essentially, AI guardrails on steroids. They also suggest devising a method to suppress dangerous capabilities entirely, sometimes called "unlearning," but it's unclear if this is possible without substantially limiting models. Misalignment is largely not something we have to worry about with generative AI as it currently exists. This type of AGI harm is envisioned as a rogue machine that has shaken off the limits imposed by its designers. Terminators, anyone? More specifically, the AI takes actions it knows the developer did not intend. DeepMind says its standard for misalignment here is more advanced than simple deception or scheming as seen in the current literature. To avoid that, DeepMind suggests developers use techniques like amplified oversight, in which two copies of an AI check each other's output, to create robust systems that aren't likely to go rogue. If that fails, DeepMind suggests intensive stress testing and monitoring to watch for any hint that an AI might be turning against us. Keeping AGIs in virtual sandboxes with strict security and direct human oversight could help mitigate issues arising from misalignment. Basically, make sure there's an "off" switch. If, on the other hand, an AI didn't know that its output would be harmful and the human operator didn't intend for it to be, that's a mistake. We get plenty of those with current AI systems -- remember when Google said to put glue on pizza? The "glue" for AGI could be much stickier, though. DeepMind notes that militaries may deploy AGI due to "competitive pressure," but such systems could make serious mistakes as they will be tasked with much more elaborate functions than today's AI. The paper doesn't have a great solution for mitigating mistakes. It boils down to not letting AGI get too powerful in the first place. DeepMind calls for deploying slowly and limiting AGI authority. The study also suggests passing AGI commands through a "shield" system that ensures they are safe before implementation. Lastly, there are structural risks, which DeepMind defines as the unintended but real consequences of multi-agent systems contributing to our already complex human existence. For example, AGI could create false information that is so believable that we no longer know who or what to trust. The paper also raises the possibility that AGI could accumulate more and more control over economic and political systems, perhaps by devising heavy-handed tariff schemes. Then one day, we look up and realize the machines are in charge instead of us. This category of risk is also the hardest to guard against because it would depend on how people, infrastructure, and institutions operate in the future.

Read more of this story at Slashdot.

BeauHD

Air Conditioning, Not Data Centers, Driving Global Energy Demand Growth

2 hours 33 minutes ago
Air conditioning will contribute more to rising global energy demand than data centers through 2030, according to an International Energy Agency. While attention has focused on computing power consumption, the IEA projects data centers will account for less than 10% of increased energy demand by 2030, significantly less than space cooling requirements. Global cooling degree days, a measure of air conditioning need, were 6% higher in 2024 than 2023 and 20% above the long-term average for the first two decades of the century. China, India and the United States saw particularly sharp increases. Air conditioning represented 7% of global electricity consumption in 2022, with some U.S. regions reporting that cooling can comprise over 70% of residential energy use during peak periods. The number of air conditioning units worldwide could nearly triple from fewer than 2 billion in 2016 to approximately 6 billion by 2050, creating a growing challenge for power grids.

Read more of this story at Slashdot.

msmash

Flux off: CISA, annexable allies warn of hot DNS threat

2 hours 39 minutes ago
Shape shifting technique described as menace to national security

The US govt's Cybersecurity Infrastructure Agency, aka CISA, on Thursday urged organizations, internet service providers, and security firms to strengthen defenses against so-called fast flux attacks.…

Thomas Claburn

US Stock Markets See Worst Day Since Covid Pandemic

2 hours 51 minutes ago
U.S. stock markets suffered their worst day since the Covid pandemic after Donald Trump announced sweeping new tariffs, triggering a global selloff and wiping out $470 billion in value from tech giants Apple and Nvidia. From a report: The tech-heavy Nasdaq fell 6%, while the S&P 500 and the Dow dropped 4.8% and 3.9%, respectively. [...] Meanwhile, the US dollar hit a six-month low, going down at least 2.2% on Thursday morning compared with other major currencies and oil prices sank on fears of a global slowdown. Though the US stock market has been used to tumultuous mornings over the last few weeks, US stock futures -- an indication of the market's likely direction -- had plummeted after the announcement. Hours later, Japan's Nikkei index slumped to an eight-month low and was followed by falls in stock markets in London and across Europe. Multiple major American business groups have spoken out against the tariffs, including the Business Roundtable, a consortium of leaders of major US companies including JP Morgan, Apple and IBM, which called on the White House to "swiftly reach agreements" and remove the tariffs. "Universal tariffs ranging from 10-50% run the risk of causing major harm to American manufacturers, workers, families and exporters," the Business Roundtable said in a statement. "Damage to the US economy will increase the longer the tariffs are in place and may be exacerbated by retaliatory measures."

Read more of this story at Slashdot.

BeauHD

Intel, TSMC Tentatively Agree To Form Chipmaking Joint Venture

3 hours 13 minutes ago
Intel and Taiwan Semiconductor Manufacturing Co. have reached a preliminary agreement to form a joint venture operating Intel's chipmaking facilities, with TSMC taking a 20% stake, The Information reports [non-paywalled source]. Intel and other U.S. semiconductor companies would hold the majority of shares in the proposed venture. Instead of capital investment, TSMC has discussed sharing chipmaking methods and training Intel personnel. The talks face internal opposition from some Intel executives concerned about widespread layoffs and the abandonment of Intel's own technology, according to the report. The deal could help TSMC neutralize a struggling competitor while potentially giving Taiwan more leverage with the U.S. administration, which recently imposed tariffs on Taiwanese goods excluding chips.

Read more of this story at Slashdot.

msmash

Microsoft's Miniature Windows 365 Link PC Goes On Sale

3 hours 50 minutes ago
An anonymous reader shares a report: Microsoft's business-oriented "Link" mini-desktop PC, which connects directly to the company's Windows 365 cloud service, is now available to buy for $349.99 in the US and in several other countries. Windows 365 Link, which was announced last November, is a device that is more easily manageable by IT departments than a typical computer while also reducing the needs of hands on support.

Read more of this story at Slashdot.

msmash