Good morning! Today: how close are we to artificial general intelligence? Also: room-temperature superconductivity has been achieved for the first time, and a NASA spacecraft is about to scoop up some asteroid rubble. Get your friends to sign up here to get The Download every day. | Artificial general intelligence: Are we close?
What is AGI? Back at the dawn of AI, artificial general intelligence was conceived as the ability to “make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves.” Half a century on, we’re still nowhere near making an AI with the multitasking abilities of a human—or even an insect. AGI has become one of the most divisive ideas in AI. But the goalposts of the search for AGI constantly shift. And even if we do build an AGI, we may not even realize we’ve done so. How do we make an AGI? AI is incredibly good at achieving particular goals in particular environments, from games to protein folding. The tricky part comes next: yoking multiple abilities together. Moving from one-algorithm to one-brain is one of the biggest open challenges in AI. Deep learning is the most general approach we have, but some researchers say we shouldn’t fixate too much on it, instead combining multiple techniques and approaches to achieve AGI. Why is AGI controversial? Part of the reason nobody knows how to build an AGI is that few agree on what it is. Without evidence on either side about whether AGI is achievable or not, the issue becomes a matter of faith. Why does it matter? A few decades ago, when AI failed to live up to the hype, funding disappeared and researchers moved on. It took many years for the technology to emerge from what were known as “AI winters” and reassert itself. That hype, though, is still there. And while it’s needed to get investors and policymakers excited, unrealistic expectations can infect decision-making, leading people to ignore very real problems—such as the way racial bias can get encoded into AI by skewed training data, the lack of transparency about how algorithms work, or questions of who is liable when an AI makes a bad decision—in favor of more fantastical concerns about things like a robot takeover. Want to read the whole story? It’s for MIT Technology Review subscribers only. But, trust me, it’s well worth becoming a subscriber. For just $50 per year you can get unlimited access to all our award-winning journalism online. And for $80 you can get the magazine delivered to your door every two months too. —Will Douglas Heaven
| | Room-temperature superconductivity has been achieved for the first time What’s happened: Room-temperature superconductors—materials that conduct electricity with zero resistance without needing special cooling— could revolutionize the electric grid, enable levitating trains, and boost quantum computing. But up to now, superconductors have had to be cooled to extremely low temperatures, which has restricted them to use as a niche technology. Now, researchers say they’ve made a breakthrough, achieving room-temperature superconductivity in a compound containing hydrogen, sulfur, and carbon at temperatures as high as 58 °F (13.3 °C). The previous record had been 8 °F. Like the previous records, the new record was attained under extremely high pressures—roughly two and a half million times greater than that of the air we breathe. Woah there: Superconductors could change our world, but they still need to be cooled to low temperatures in order to work. This makes them more complicated, expensive, and prone to failure. It remains to be seen whether scientists can devise stable compounds that are superconducting not only at ambient temperature, but also at ambient pressure. Read the full story. —Konstantin Kakaes | A NASA spacecraft is about to scoop up some asteroid rubble Since December 2018, the OSIRIS-REx spacecraft has been orbiting the asteroid Bennu and trying to find out more about its chemistry and geology. And for good reason: “Bennu is a time capsule,” says Thomas Zurbuchen, the head of NASA’s Science Mission Directorate. “It has been out there for 4.5 billion years, and carries the history of that environment with it.” If we study Bennu and other asteroids like it in deep detail, we can potentially unravel how the ingredients that lead to the formation of planets like Earth—and, eventually, life—come together. Now comes the hard bit: getting a sample. That’s precisely the goal come October 20, when OSIRIS-REx will plunge down toward the surface of the asteroid and attempt to scoop up some rubble and dust from the surface. It will be one of the hardest things NASA has ever attempted, taking place over a fraught 4.5 hours more than 200 million miles away from Earth. And then it has to bring the sample safely back over the course of the next three years. There are more than a few ways things can go wrong. Read the full story. —Neel V. Patel | | The top ten must-reads I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 How a White House appointee undermined the CDC from within The agency is in crisis after months of unforced errors. ( Science) + Central Europe is starting to experience its first really big covid-19 surge. ( NYT $) + Daily infections have reached record levels in Germany. ( Reuters) + It’s a good thing that three coronavirus trials have been paused. ( NYT $) + A powerful new way we could track covid-19: its genetic code. ( WP $) 2 We all need a fix of fresh air 💨 This article focuses on schools and children, but really it applies to all shared indoor spaces. ( Wired $) 3 Facebook and Twitter are limiting sharing of an unsubstantiated Biden story It looks like a classic “hack and leak” operation, similar to Hilary’s emails. ( NPR) + Trump is going to claim the election was rigged, and a giant online network is poised to amplify him. ( NBC) + Misinformation is stoking up calls for violence on election day. ( NYT $) + Why we’re not voting on our phones. ( NBC) 4 Dreading the winter? Here’s what to do ❄️ Try to find activities that stop you overly focusing on yourself. ( Vox) 5 Bluetooth bugs are a big problem for contact tracing apps We need more and better data to know how well they’re working. ( Wired UK) 6 YouTube has banned coronavirus vaccine misinformation Big Tech is starting to step up on this issue, ahead of any potential approval of a vaccine. ( Reuters) + The elderly and numerically illiterate are more likely to share fake news. ( The Guardian) + Looking back over four years of Trump’s tweets. (WP $) + Twitter is banning Holocaust denial posts. ( The Verge) + New York’s financial watchdog says we need a dedicated regulator for social media companies. ( WSJ $) 7 There are online forums dedicated to unmasking chronic illness “fakers” What crushingly sad lives these people must lead. ( OneZero) 8 Amazon is avoiding a tax in the UK by passing it onto small traders Lovely company. But also—how did lawmakers not foresee this?! ( The Guardian) + Amazon workers say it reinstated dangerous productivity quotas for Prime Day. ( Bloomberg $) 9 How to cope with nonstop Zoom Sometimes, you do just need to say no. ( Wired $) 10 The company known as WeWork is going back to being called WeWork But it gave us all a good laugh when it dedicated its original name change to “the power of We.” ( Reuters) | | “It’s horrific. It doesn’t work. Stop trying to use it.” —Hannah Smethurst, a legal tutor at the University of Edinburgh, has some advice for universities considering buying software to remotely surveil students, Wired UK reports. | | | | | |
No comments:
Post a Comment