New Week Same Humans #39
DeepMind researchers propose a new route to general AI. Jeff Bezos is going out of this world. Plus more news and analysis from this week.
Welcome to the mid-week update from New World Same Humans, a newsletter on trends, technology, and society by David Mattin.
If you’re reading this and you haven’t yet subscribed, then join 17,000+ curious souls on a journey to build a better shared future 🚀🔮
💡 On Sunday I wrote about a new supersonic passenger jet, flying cars, and the Great Stagnation. Go here to read The Future that Never Came.💡
This week’s NWSH is brought to you by Ground News.
Now more than ever we need our news to be optimised for balance, usefulness, and truth. But in 2021, all too often it’s optimised only for clicks.
I’ve written in this newsletter about the consequences: rising polarisation, and a public square riven with misinformation.
Ground News is a startup that wants to fix that problem. It’s on an exciting mission to put readers, and not algorithms, in the driver’s seat.
Ground News is a news source comparison platform that empowers readers to form a balanced opinion of the day’s most important stories. It lets you to compare how sources from the left, centre, and right are covering any story, so you can see the full picture, instead of just one side.
The result is a game-changing platform, which gives you the tools you need to make your own informed judgements on what’s really happening.
NWSH readers are on an endless quest to better understand the world we live in. So I know this will be right up your street. Download the free app, and let me know what you think!
This week, could a machine trial and error its way to true general intelligence? Researchers at DeepMind say it may be possible.
Also, Jeff Bezos announces big plans for July. And Twitter wonders whether we should tax his $187 billion fortune.
Plus, cloud services provider Fastly goes down, and then up.
Let’s go!
🤖 The path to an artificial mind
In a new paper published in the journal Artificial Intelligence this month, researchers at DeepMind quietly float a new path towards general artificial intelligence.
The authors of ‘Reward is Enough’ say that straightforward trial and error learning processes, which seek to optimise reward exposure, may lead us to highly evolved artificial minds. Minds, that is, that display the kind of varied and complex cognitive abilities that we do: perception, language, memory, and social skills.
Look at it this way: squirrel intelligence – including the cunning needed to make fake stores of nuts to fool competitors – evolved via trial and error food reward maximisation. Similarly, a kitchen robot tasked with cleanliness maximisation will eventually evolve complex cognitive abilities, including the social skills needed to order itself a pizza and so avoid using the kitchen altogether.
This all sounds a little obscure. But it means a well-established technique, reinforcement learning, may be enough to lead us to human-like general AI. While deep learning is all about training an AI on pre-existing data sets, reinforcement learning sees AIs train themselves via trial and error. Both techniques were at work in recent high-profile DeepMind achievements, including AlphaFold’s solution to the iconic protein-folding problem.
⚡ NWSH Take: The DeepMind researchers behind this paper are clear that it’s only an early hypothesis. And they give no details on the kind of advanced trial and error processes that might lead to general AI. Still, it’s an intriguing prospect: reinforcement learning as a path towards a true silicon mind. // This all taps into one of the key emerging themes inside this newsletter. Human nature will shape the decades ahead just as it has shaped out past. But a new kind of mind, a new way of seeing the world, is emerging, and that too will help shape our shared future in ways that we’re only beginning to understand.
🧑🚀 Spaceman, always wanted you to go
Jeff Bezos will fly to space next month, courtesy of his own startup.
In a slick announcement video published this week, Bezos says he’ll travel in the first crewed space flight of Blue Origin’s New Shepherd rocket. And he’s taking his younger brother, Mark, with him.
The perfect family day trip for the neo-feudal overlord who has everything: an 11-minute flight to the edge of space.
Meanwhile, news site ProPublica claimed that iconic American billionaires pay a ‘true tax rate’ of just a few per cent. The platform say they accessed a cache of leaked data to gain a unique insight into the finances of the ultra-rich.
Cue much outrage on Twitter; much of it aimed at the billionaires, but a good deal at ProPublica.
Critics say their ‘true tax rate’ is intentionally misleading, because it looks at the tax paid against increase in wealth, rather than income. That is, they say, like asking an ordinary citizen to pay more tax because the value of her house increased: the tax system just doesn’t work like that. Divide Elon Musk’s tax contribution ($455 million) by his reported income ($1.52 billion) and you get a tax rate of 29.9%.
I have some thoughts.
⚡ NWSH Take: The critics of ProPublica are right; they’re also serving pretty thin gruel. The US doesn’t tax unrealised capital gains, and it’s almost impossible to devise a good system that does. So sure, this ‘true tax rate’ is disingenuous. // Still, faced with a world in which Jeff Bezos increased his wealth by $99 billion in 2020, is that all we have to say? Do we reserve any ire for the likes of Elon Musk, who reportedly paid no income tax in 2019? // It’s not that hard to devise a tax system that does access these vast accumulations of wealth; the easiest way is probably to increase inheritance tax. // Next month will see an event that symbolises neo-feudal capitalism so perfectly it could be a scene from a novel. The world’s second richest man will fly to space with his brother, in a rocket created by the company he owns, and look down on us all from a height of 60 miles. A good day, also, to get serious about a global wealth tax?
👋 Notes on a launch
I’ll be back on Clubhouse this Friday, talking to the CEO of Faves, Tyler Maloney.
This week Faves launched on Product Hunt, and won a place on the coveted Top Five Products of the Day list.
I’ll be talking to Tyler and other Faves curators about why Product Hunt is such a powerful showcase, how to launch successfully, the future of social, and more.
The room opens this Friday 11 June at 10am California time, 1pm NYC, 7pm Berlin, and you can add it to your calendar here.
☁️ Down, then up
Cloud computing services provider Fastly went down on Tuesday, and half the internet went dark. The outage temporarily took out some of the world’s largest websites, including Amazon, UK government sites, and the Guardian.
The company say a bug was triggered when a single customer updated their settings, taking out 85% of the organisation’s network.
Fastly’s share price rose 12% on the day. Why? Most likely because the market got to see just how dependent key web players are on the company.
⚡ NWSH Take: This was a quick reminder – though it probably didn’t feel quick to executives at Fastly – of a key truth. The internet isn’t only about ethereal virtual space; it runs on infrastructure, and that infrastructure can fail. // There’s something analogous, here, to the dawning realisation that crypto places huge energy demands on national electricity grids. Our digital lives depend on IRL resources. Twitter founder Jack Dorsey has just pledged that his payment company, Square, will invest $5 million in solar-powered bitcoin mining.
🗓️ Also this week
🛰️ A startup created by the CCP will send 13,000 satellites into orbit to provide broadband across China. In Do We Need a World Government? I wrote about the threat posed by satellites to our view of the night sky.
🛥️ Amsterdam is testing autonomous electric boats on its canals. The boats could carry passengers and pick up rubbish, say project partners MIT and The Amsterdam Institute for Advanced Metropolitan Solutions.
📥 Apple is killing the ability to track email opens inside its Mail app. There goes my obsession with NWSH open rates.
👁️ Chinese tech companies are using surveillance software called Third Eye to ensure their employees work 12-hour shifts. The company behind Third Eye say it is ‘all-powerful, controlling, and stable’.
⚖️ The state of Ohio wants Google to be recognised as a public utility under Ohio law. A lawsuit filed by the state compares the tech giant to ‘a railroad or electricity company’.
🤣 The FBI created a fake encrypted messaging app and tricked thousands of criminals into using it. The ANOM app was used by drug gangs and members of the mafia; over 800 people have now been arrested.
🚪 Amazon’s Ring smart doorbell company won’t say how many users have had their footage shared with police. The company has partnerships with at least 1,800 US police departments.
🌱 A US research team created a self-sustaining electronic microsystem. The system can respond to information inputs without requiring any external energy input, in the way a living organism can.
🌍 Humans of Earth
Key metrics to help you keep track of Project Human.
🙋 Global population: 7,871,545,262
🌊 Earths currently needed: 1.7859628177
💉 Global population vaccinated: 6.1%
🗓️ 2021 progress bar: 44% complete
📖 On this day: On 9 June AD 68 the Roman emperor Nero commits suicide, and a civil war begins.
Trick of the mind
Thanks for reading this week.
I’m intrigued by the idea floated by DeepMind researchers this month: that reinforcement learning – the process you and I call trial and error – might just lead us to a true artificial mind.
We’re a long way from that right now. But NWSH will be watching closely as this idea develops.
And there’s one thing you can do to help with that mission: share!
Now that you’ve made it to the end of this week’s instalment, why not forward the email to someone who’d also enjoy it? Or share it across one of your social networks, with a note on why you found it valuable. Remember: the larger and more diverse the NWSH community becomes, the better for all of us.
I’ll be back on Sunday. Until then, be well,
David.
P.S Thanks to Monique van Dusseldorp for additional research and analysis.
Hi David,
On "The path to an artificial mind", DeepMind is probably right, but there's a huge shortcut we could take, albeit not an easy one. Mimic the human hardware.
As Jeff Hawkins states in "A thousand brains", the neocortex functions as a kind of “memory prediction framework”, building models about how the world works and then constantly making predictions based on those models. Each one of our 150,000 cortical columns is a tiny learning machine ceaselessly adapting its reference frame in response to the external stimuli they receive such as the saccadic movements of our eyes, which send fresh inputs to the brain three times a second. Our brain then counts the “votes” from all of the related cortical columns to form a unified understanding of what is going on in the world at any particular time and any particular subject, compares new inputs against the predictions and makes decisions based on that.
Iif reference frames can be built into AI systems then machines could far exceed human capabilities. Whereas human neurons take at least five milliseconds to do anything useful, silicon transistors can operate a million times faster and could therefore, potentially think and learn a million times faster.
DeepMind (Google) should try this approach too.
Cheers,