- Email: [email protected]
A vote against deepfakes
COMMENT
A vote against deepfakes AI-generated hoax videos won’t wreck democracy this year as some are predicting, says Tim Hwang THERE is no doubt that recent demonstrations of believable fakes generated by machine learning are striking. From uncanny simulated voices to fabricated videos of political leaders, it is natural to worry that these techniques will make it easier to manipulate political discourse and public opinion. But it is worth taking a step back. Just because a capability exists does not mean that it will be widely used, or make a serious impact. There are good reasons to believe that – for the near future – “deepfakes” enabled by artificial intelligence will see limited use, and have limited impact. To that end, I’m taking part in a friendly public wager and betting that by the end of 2018, we will not have seen a political hoax generated by machine learning get more than 2 million views before being discovered. Given
that this is a year in which US voters go to the polls again for their mid-term elections – when we might expect disinformation campaigns to be likely – why do I think I’m backing the right horse? The economics are important: state and non-state actors using online propaganda want to achieve the most influence at the lowest cost. We need to recognise that rudimentary techniques can already have a big influence on public discourse. Numerous incidents attest to the fact that simply reusing an old video and asserting that it is something that it is not can be sufficient to fool many people. A crude Photoshop job can spread disinformation. At the same time, AI remains a relatively costly tool for generating hoaxes. The current breakthroughs in AI rely on two major inputs – large datasets and computational power – both of
What’s the beef? The term “lab grown meat” is already riling cattle farmers, says Sasha Chapman A DUTCH start-up announces it has raised $8.8 million to commercialise its lab-grown burger. Meanwhile, the US meat industry gears up for a scuffle with the new kids on the block. Mark Post, CEO of Mosa Meat, predicts he can get the company’s burger into restaurants at $10 a pop by 2021. This is substantially 22 | NewScientist | 11 August 2018
what this will mean for them. While Post was trumpeting his plans, meat industry lobbyists were in Washington DC to prepare for battle over the labelling and regulation of lab-grown meat, particularly what it calls itself. While the term “lab-grown meat” has taken hold, some advocates prefer “clean meat” because they see it as a way of dispensing with slaughter and the degradation of our environment to produce meat. Detractors say
cheaper than the world’s first cultured beef patty, which cost $300,000 when Post unveiled it in 2013. This bodes well for the economic viability and rapid development of cellular agriculture: the culturing of meat, “When people first used the word meat, they fish and dairy products in vats. simply meant food as Which is why conventional opposed to drink” farmers are now worrying about
lab-grown meat is “fake”. The US Cattlemen’s Association wants the US Department of Agriculture to define meat as “tissue or flesh of animals that have been harvested in the traditional manner” – as if an animal must die to make a burger authentic. Those who insist on limiting the definition of meat are hoping to continue a trend. Historian Benjamin Wurgaft, writing a book about cultured meat, points out the definition has narrowed over time. When people first used the word meat, they simply meant food (as opposed to drink). It wasn’t until the 1300s that it began to refer to animal flesh.
For more opinion articles, visit newscientist.com/opinion
Tim Hwang is a writer and researcher, and director of the Harvard-MIT Ethics and Governance of AI Initiative
And the definition has arguably narrowed again – today for a lot of people it equates to the animal proteins that dominate the food system: poultry, beef and pork. Maybe the debate over what to call flesh grown in a vat will upend that trend. Maybe it will narrow it further. What is certain is that a war of words lies ahead. Let’s hope that conversation also allows a broader discussion of a growing reliance on meat – and whether any technology can sustain that on a planet of limited resources. ■ Sasha Chapman is a Canadian writer focused on environmental and health implications of the global food industry
ANALYSIS 3D-printed firearms
ROBERT CLARK/NATIONAL GEOGRAPHIC/GETTY
which can be expensive to acquire. Moreover, the techniques behind the more impressive demonstrations of AI-driven fabrication, known as generative adversarial networks, are famously temperamental. This all suggests that, in the near term, AI-constructed fakes will be inferior to the existing arsenal of propaganda tools. As a result, researchers have more time to create effective methods for detecting this type of fakery. It also enables society to adjust to the knowledge that everyday video can be faked in this way, and to be on guard before widespread deployment. Ultimately, it is possible that deepfakes are a diversion from more crucial questions. Or we may eventually see AI-driven hoaxes become an attractive option for technology-minded propagandists. But it will be the draw of the underlying narratives pushed by these fakes that determine their influence. A better understanding of the behavioural factors driving belief in hoaxes will be the only way to avoid playing an unwinnable game of technological whack-a-mole. ■
The real US gun crisis is being overlooked Frank Swain
and quietly gave permission for the files to go online again from 1 August. Eight US states in turn sued the Department of State and got a temporary restraining order issued to block the release of the files. A hearing next week will decide what happens next. Amid all this legal wrangling, a key question has gone unanswered: are these weapons actually a threat? The prospect of roving gangs armed with 3D-printed guns is slim. The Liberator is a clunky and ineffective weapon that can only be fired a few times before the plastic barrel splits. Equally,
LAW-MAKERS in the US are fighting to keep blueprints for 3D-printed guns off the internet – but how worried should we be about untraceable plastic firearms? In 2013, law student Cody Wilson unveiled the Liberator, a plastic handgun produced on a 3D printer that could fire conventional ammunition. He posted the files online so that anybody could download them and, in theory, print their own pistol. He later added a second design that helps people to mill a rifle part called a lower receiver from a block of aluminium. These parts are controlled “The fate of 3D-printed firearms is a distraction under US gun regulations and carry from the US’s much larger identifying serial numbers. gun public health crisis” By releasing plans for untraceable weapons that could be produced at home, Wilson was making effective building your own gun is not illegal in gun regulation impossible. The US the US, and anyone with the required Department of State told Wilson to skills is free to create their own lower remove the files from the internet. receiver on a milling machine. They had already been downloaded Although untraceable firearms more than 100,000 times. sound particularly scary, the US is Wilson sued the Department of already full of them. And regulations State in response. It relented last designed to enforce US gun rights month, following a lengthy legal battle, prevent the government from creating
a searchable, computerised database of gun store sales. Instead, the records are kept on paper and microfilm, which must be painstakingly searched by hand. In many states, private sales are unmonitored. Behind the hysterics over new technology, the crux of the debate lies in whether the government should control access to firearms. As a risk to citizens, the Liberator is underwhelming; it poses a much greater threat to the authority of the US government. An age of easily shareable digital files and reliable fabrication offers resistance to what Wilson called “the collectivisation of manufacture”, putting it outside the control of authorities. Ultimately, the fate of 3D-printed firearms is a distraction from the US’s much larger gun problem: the public health crisis of the tens of thousands of people who die each year as a result of gun violence. Since 2013, when Wilson first released his plans for the Liberator, no one has been killed by a 3D-printed gun. According to the Gun Violence Archive, an organisation that launched in the same year to collate statistics on gun violence, 65,310 people in the US have been killed by guns since 2014. What’s more, that doesn’t include the roughly 22,000 gun suicides that take place in the US each year. The US has a gun problem – 3D printing is a distraction. ■ 11 August 2018 | NewScientist | 23
A vote against deepfakes AI-generated hoax videos won’t wreck democracy this year as some are predicting, says Tim Hwang THERE is no doubt that recent demonstrations of believable fakes generated by machine learning are striking. From uncanny simulated voices to fabricated videos of political leaders, it is natural to worry that these techniques will make it easier to manipulate political discourse and public opinion. But it is worth taking a step back. Just because a capability exists does not mean that it will be widely used, or make a serious impact. There are good reasons to believe that – for the near future – “deepfakes” enabled by artificial intelligence will see limited use, and have limited impact. To that end, I’m taking part in a friendly public wager and betting that by the end of 2018, we will not have seen a political hoax generated by machine learning get more than 2 million views before being discovered. Given
that this is a year in which US voters go to the polls again for their mid-term elections – when we might expect disinformation campaigns to be likely – why do I think I’m backing the right horse? The economics are important: state and non-state actors using online propaganda want to achieve the most influence at the lowest cost. We need to recognise that rudimentary techniques can already have a big influence on public discourse. Numerous incidents attest to the fact that simply reusing an old video and asserting that it is something that it is not can be sufficient to fool many people. A crude Photoshop job can spread disinformation. At the same time, AI remains a relatively costly tool for generating hoaxes. The current breakthroughs in AI rely on two major inputs – large datasets and computational power – both of
What’s the beef? The term “lab grown meat” is already riling cattle farmers, says Sasha Chapman A DUTCH start-up announces it has raised $8.8 million to commercialise its lab-grown burger. Meanwhile, the US meat industry gears up for a scuffle with the new kids on the block. Mark Post, CEO of Mosa Meat, predicts he can get the company’s burger into restaurants at $10 a pop by 2021. This is substantially 22 | NewScientist | 11 August 2018
what this will mean for them. While Post was trumpeting his plans, meat industry lobbyists were in Washington DC to prepare for battle over the labelling and regulation of lab-grown meat, particularly what it calls itself. While the term “lab-grown meat” has taken hold, some advocates prefer “clean meat” because they see it as a way of dispensing with slaughter and the degradation of our environment to produce meat. Detractors say
cheaper than the world’s first cultured beef patty, which cost $300,000 when Post unveiled it in 2013. This bodes well for the economic viability and rapid development of cellular agriculture: the culturing of meat, “When people first used the word meat, they fish and dairy products in vats. simply meant food as Which is why conventional opposed to drink” farmers are now worrying about
lab-grown meat is “fake”. The US Cattlemen’s Association wants the US Department of Agriculture to define meat as “tissue or flesh of animals that have been harvested in the traditional manner” – as if an animal must die to make a burger authentic. Those who insist on limiting the definition of meat are hoping to continue a trend. Historian Benjamin Wurgaft, writing a book about cultured meat, points out the definition has narrowed over time. When people first used the word meat, they simply meant food (as opposed to drink). It wasn’t until the 1300s that it began to refer to animal flesh.
For more opinion articles, visit newscientist.com/opinion
Tim Hwang is a writer and researcher, and director of the Harvard-MIT Ethics and Governance of AI Initiative
And the definition has arguably narrowed again – today for a lot of people it equates to the animal proteins that dominate the food system: poultry, beef and pork. Maybe the debate over what to call flesh grown in a vat will upend that trend. Maybe it will narrow it further. What is certain is that a war of words lies ahead. Let’s hope that conversation also allows a broader discussion of a growing reliance on meat – and whether any technology can sustain that on a planet of limited resources. ■ Sasha Chapman is a Canadian writer focused on environmental and health implications of the global food industry
ANALYSIS 3D-printed firearms
ROBERT CLARK/NATIONAL GEOGRAPHIC/GETTY
which can be expensive to acquire. Moreover, the techniques behind the more impressive demonstrations of AI-driven fabrication, known as generative adversarial networks, are famously temperamental. This all suggests that, in the near term, AI-constructed fakes will be inferior to the existing arsenal of propaganda tools. As a result, researchers have more time to create effective methods for detecting this type of fakery. It also enables society to adjust to the knowledge that everyday video can be faked in this way, and to be on guard before widespread deployment. Ultimately, it is possible that deepfakes are a diversion from more crucial questions. Or we may eventually see AI-driven hoaxes become an attractive option for technology-minded propagandists. But it will be the draw of the underlying narratives pushed by these fakes that determine their influence. A better understanding of the behavioural factors driving belief in hoaxes will be the only way to avoid playing an unwinnable game of technological whack-a-mole. ■
The real US gun crisis is being overlooked Frank Swain
and quietly gave permission for the files to go online again from 1 August. Eight US states in turn sued the Department of State and got a temporary restraining order issued to block the release of the files. A hearing next week will decide what happens next. Amid all this legal wrangling, a key question has gone unanswered: are these weapons actually a threat? The prospect of roving gangs armed with 3D-printed guns is slim. The Liberator is a clunky and ineffective weapon that can only be fired a few times before the plastic barrel splits. Equally,
LAW-MAKERS in the US are fighting to keep blueprints for 3D-printed guns off the internet – but how worried should we be about untraceable plastic firearms? In 2013, law student Cody Wilson unveiled the Liberator, a plastic handgun produced on a 3D printer that could fire conventional ammunition. He posted the files online so that anybody could download them and, in theory, print their own pistol. He later added a second design that helps people to mill a rifle part called a lower receiver from a block of aluminium. These parts are controlled “The fate of 3D-printed firearms is a distraction under US gun regulations and carry from the US’s much larger identifying serial numbers. gun public health crisis” By releasing plans for untraceable weapons that could be produced at home, Wilson was making effective building your own gun is not illegal in gun regulation impossible. The US the US, and anyone with the required Department of State told Wilson to skills is free to create their own lower remove the files from the internet. receiver on a milling machine. They had already been downloaded Although untraceable firearms more than 100,000 times. sound particularly scary, the US is Wilson sued the Department of already full of them. And regulations State in response. It relented last designed to enforce US gun rights month, following a lengthy legal battle, prevent the government from creating
a searchable, computerised database of gun store sales. Instead, the records are kept on paper and microfilm, which must be painstakingly searched by hand. In many states, private sales are unmonitored. Behind the hysterics over new technology, the crux of the debate lies in whether the government should control access to firearms. As a risk to citizens, the Liberator is underwhelming; it poses a much greater threat to the authority of the US government. An age of easily shareable digital files and reliable fabrication offers resistance to what Wilson called “the collectivisation of manufacture”, putting it outside the control of authorities. Ultimately, the fate of 3D-printed firearms is a distraction from the US’s much larger gun problem: the public health crisis of the tens of thousands of people who die each year as a result of gun violence. Since 2013, when Wilson first released his plans for the Liberator, no one has been killed by a 3D-printed gun. According to the Gun Violence Archive, an organisation that launched in the same year to collate statistics on gun violence, 65,310 people in the US have been killed by guns since 2014. What’s more, that doesn’t include the roughly 22,000 gun suicides that take place in the US each year. The US has a gun problem – 3D printing is a distraction. ■ 11 August 2018 | NewScientist | 23