We can now use cells from dead people to create new life. But who gets to decide?

This article is from The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, sign up here.

Peter Zhu was just 19 years old when he died following a skiing accident in West Point, New York. His donor card made clear he had wanted to donate his organs. But his parents wanted to collect his sperm, too.

His parents told a court that they wanted to keep the possibility of using the sperm to eventually have children that would be genetically related to Peter. The court approved their wishes, and Peter’s sperm was retrieved from his body and stored in a local sperm bank. 

We have the technology to use sperm, and potentially eggs, from dead people to make embryos, and eventually babies. And there are millions of eggs and embryos—and even more sperm—in storage and ready to be used. When the person who provided those cells dies, like Peter, who gets to decide what to do with them?

That was the question raised at an online event held by the Progress Educational Trust, a UK charity for people with infertility and genetic conditions, that I attended on Wednesday. The panel included a clinician and two lawyers, who addressed plenty of tricky questions, but provided few concrete answers. 

In theory, the decision should be made by the person who provided the eggs, sperm or embryos. In some cases, the person’s wishes might be quite clear. Someone who might be trying for a baby with their partner may store their sex cells or embryos and sign a form stating that they are happy for their partner to use these cells if they die, for example. 

But in other cases, it’s less clear. Partners and family members who want to use the cells might have to collect evidence to convince a court the deceased person really did want to have children. And not only that, but that they wanted to continue their family line without necessarily becoming a parent themselves.

Sex cells and embryos aren’t property—they don’t fall under property law and can’t be inherited by family members. But there is some degree of legal ownership for the people who provided the cells. It is complicated to define that ownership, however, Robert Gilmour, a family law specialist based in Scotland, said at the event. “The law in this area makes my head hurt,” he said.

The law varies depending on where you are, too. Posthumous reproduction is not allowed in some countries, and is unregulated in many others. In the US, laws vary by state. Some states won’t legally recognize a child conceived after a person’s death as that person’s offspring, according to the American Society for Reproductive Medicine (ASRM). “We do not have any national rules or policies,” Gwendolyn Quinn, a bioethicist at New York University, tells me.

Societies like ASRM have put together guidance for clinics in the meantime. But this can also vary slightly between regions. Guidance by the European Society for Human Reproduction and Embryology, for example, recommends that parents and other relatives should not be able to request the sex cells or embryos of the person who died. That would apply to Peter Zhu’s parents. The concern is that these relatives might be hoping for a “commemorative child” or as “a symbolic replacement of the deceased.”

The people who want to use the eggs, sperm or embryos of dead partners or family members are often painted as “selfish”, but, in James Lawford Davies’s experience, that just isn’t the case. Lawford Davies, a UK-based solicitor specializing in reproductive and genetic technologies, has been involved in several similar cases. “All of these cases have involved incredibly brave people who have been faced with tragedy,” he said. The people involved all wanted to deliver on the wishes of those who had died, he added.

Posthumous reproduction is undoubtedly a murky area, but there’s one thing that everyone I’ve heard and spoken to agrees on: each case is unique, and should be treated individually. “It’s hard to generalize,” says Shelly Simana, a bioethicist and legal scholar at Stanford University in California.

Simana wants more people to start thinking about the possibility of their own eggs, sperm and embryos being used after their death. In the same way that we’re encouraged to consider organ donation, we should all be writing down whether we’d be happy to have our sex cells retrieved and used, she says. “Ideally we’d have people write a biological will,” she says.

Quinn agrees. “We often tell people that Thanksgiving… when you’re sitting with your family… is a good time to express your wishes,” she says. “They are very hard conversations to have…Talking about death isn’t comfortable for a lot of people, but how else will you make your wishes known?” 

To read more about the latest in reproductive technology, check out these stories from Tech Review’s archive:

There’s a race to create egg and sperm cells in the lab—a technique that could potentially solve infertility problems and allow more people to have children that are genetically-related to them.  

I’ve also written about how advances in reproductive technology could lead to babies with four or more biological parents—forcing us to reconsider parenthood.

Earlier this year, scientists found a way to mature eggs from transgender men in the labwhich could offer them new ways to start a family.

Another team have used similar techniques to help create embryos from immature eggs of transgender men, taking the technology a step further

And then there’s the technology that uses material from three genetic “parents” to create an embryo. Here’s Emily Mullin’s 2017 profile of the startup trying to commercialize “three-parent” babies. 

From around the web:

How much would you pay to see a woolly mammoth? My colleague Antonio Regalado spoke to Sara Ord, director of species restoration at Colossal, who plans to revive extinct animal species and showcase them in zoos. (MIT Technology Review)

Offering freedom to people with Alzheimer’s disease might improve their symptoms, as well as providing them with a better quality of life, according to staff at this French dementia village. (The New Yorker)

Are you still taking precautions against covid-19? Here’s what the experts are doing, nearly three years into the pandemic. (STAT)

We’re better prepared for covid this winter compared to last year, say Anthony Fauci and Ashish Jha—as long as we keep up with vaccinations. (The New York Times)

Scientists have made fat-free “whipped cream” from bacteria. Yum. (Ars Technica)

A bot that watched 70,000 hours of Minecraft could unlock AI’s next big thing

OpenAI has built the best Minecraft-playing bot yet by making it watch 70,000 hours of video of people playing the popular computer game. It showcases a powerful new technique that could be used to train machines to carry out a wide range of tasks by binging on sites like YouTube, a vast and untapped source of training data.

The Minecraft AI learned to perform complicated sequences of keyboard and mouse clicks to complete tasks in the game, such as chopping down trees and crafting tools. It’s the first bot that can craft so-called diamond tools, a task that typically takes good human players 20 minutes of high-speed clicking—or around 24,000 actions.

The result is a breakthrough for a technique known as imitation learning, in which neural networks are trained how to perform tasks by watching humans do them. Imitation learning can be used to train AI to control robot arms, drive cars or navigate webpages.  

There is a vast amount of video online showing people doing different tasks. By tapping into this resource, the researchers hope to do for imitation learning what GPT-3 did for large language models. “In the last few years we’ve seen the rise of this GPT-3 paradigm where we see amazing capabilities come from big models trained on enormous swathes of the internet,” says Bowen Baker at OpenAI, one of the team behind the new Minecraft bot. “A large part of that is because we’re modeling what humans do when they go online.”

The problem with existing approaches to imitation learning is that video demonstrations need to be labeled at each step: doing this action makes this happen, doing that action makes that happen, and so on. Annotating by hand in this way is a lot of work, and so such datasets tend to be small. Baker and his colleagues wanted to find a way to turn the millions of videos that are available online into a new dataset.

The team’s approach, called Video Pre-Training (VPT), gets around the bottleneck in imitation learning by training another neural network to label videos automatically. They first hired crowdworkers to play Minecraft, and recorded their keyboard and mouse clicks alongside the video from their screens. This gave the researchers 2000 hours of annotated Minecraft play, which they used to train a model to match actions to onscreen outcome. Clicking a mouse button in a certain situation makes the character swing its axe, for example.  

The next step was to use this model to generate action labels for 70,000 hours of unlabelled video taken from the internet and then train the Minecraft bot on this larger dataset.

“Video is a training resource with a lot of potential,” says Peter Stone, executive director of Sony AI America, who has previously worked on imitation learning. 

Imitation learning is an alternative to reinforcement learning, in which a neural network learns to perform a task from scratch via trial and error. This is the technique behind many of the biggest AI breakthroughs in the last few years. It has been used to train models that can beat humans at games, control a fusion reactor, and discover a faster way to do fundamental math.

The problem is that reinforcement learning works best for tasks that have a clear goal, where random actions can lead to accidental success. Reinforcement learning algorithms reward those accidental successes to make them more likely to happen again.

But Minecraft is a game with no clear goal. Players are free to do what they like, wandering a computer-generated world, mining different materials and combining them to make different objects. 

Minecraft’s open-endedness makes it a good environment for training AI. Baker was one of the researchers behind Hide & Seek, a project in which bots were let loose in a virtual playground where they used reinforcement learning to figure out how to cooperate and use tools to win simple games. But the bots soon outgrew their surroundings. “The agents kind of took over the universe, there was nothing else for them to do” says Baker. “We wanted to expand it and we thought Minecraft was a great domain to work in.”

They’re not alone. Minecraft is becoming an important testbed for new AI techniques. MineDojo, a Minecraft environment with dozens of predesigned challenges, won an award at this year’s NeurIPS, one of the biggest AI conferences. 

Using VPT, OpenAI’s bot was able to carry out tasks that would have been impossible using reinforcement learning alone, such as crafting planks and turning them into a table, which involves around 970 consecutive actions. Even so, they found that the best results came from using imitation learning and reinforcement learning together. Taking a bot trained with VPT and fine-tuning it with reinforcement learning allowed it to carry out tasks involving more than 20,000 consecutive actions.  

The researchers claim that their approach could be used to train AI to carry out other tasks. To begin with, it could be used to for bots that use a keyboard and mouse to navigate websites, book flights or buy groceries online. But in theory it could be used to train robots to carry out physical, real-world tasks by copying first-person video of people doing those things. “It’s plausible,” says Stone.

Matthew Gudzial at the University of Alberta, Canada, who has used videos to teach AI the rules of games like Super Mario Bros, does not think it will happen any time soon, however. Actions in games like Minecraft and Super Mario Bros. are performed by pressing buttons. Actions in the physical world are far more complicated and harder for a machine to learn. “It unlocks a whole mess of new research problems,” says Gudzial.

“This work is another testament to the power of scaling up models and training on massive datasets to get good performance,” says Natasha Jaques, who works on multi-agent reinforcement learning at Google and the University of California, Berkeley. 

Large internet-sized data sets will certainly unlock new capabilities for AI, says Jaques. “We’ve seen that over and over again, and it’s a great approach.” But OpenAI places a lot of faith in the power of large data sets alone, she says: “Personally, I’m a little more skeptical that data can solve any problem.”

Still, Baker and his colleagues think that collecting more than a million hours of Minecraft videos will make their AI even better. It’s probably the best Minecraft-playing bot yet, says Baker: “But with more data and bigger models I would expect it to feel like you’re watching a human playing the game, as opposed to a baby AI trying to mimic a human.”

The Download: AI conquers Minecraft, and babies after death

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.

A bot that watched 70,000 hours of Minecraft could unlock AI’s next big thing

The news: An AI that binged on 70,000 hours of people playing Minecraft has learned how to play the game better than any AI before. The bot, created by OpenAI, showcases a powerful new technique that could be used to train machines to carry out a wide range of tasks by binging on websites like YouTube, a vast and untapped source of training data.

How they did it: The Minecraft AI learned to perform complicated sequences of keyboard and mouse clicks to complete tasks in the game. It’s the first bot that can craft so-called diamond tools, a task that typically takes good human players 20 minutes of high-speed clicking.

Why it matters: The result is a breakthrough for a technique known as imitation learning, in which neural networks are trained how to perform tasks by watching humans do them. Imitation learning can be used to train AI to control robot arms, drive cars or navigate websites. Read the full story.

—Will Douglas Heaven

BLACK FRIDAY DEAL: Read our innovative reporting for 50% less 

If you’re not already an MIT Technology Review subscriber, there’s never been a better time to sign up. Between today and Monday 28 November, you can save a whopping 50% off a Digital + Print subscription. That means you can access our world-leading insights from just $40 a year. But the offer won’t last for long—claim yours now.

We can now use cells from dead people to create new life. But who gets to decide?

Peter Zhu was just 19 years old when he died following a skiing accident. His donor card made clear he had wanted to donate his organs. But his parents wanted to collect his sperm, too.

His parents told a court that they wanted to keep the possibility of using the sperm to eventually have children that would be genetically related to Peter. The court approved their wishes, and Peter’s sperm was retrieved from his body and stored in a local sperm bank.

We have the technology to use sperm, and potentially eggs, from dead people to make embryos, and eventually babies. And there are millions of eggs and embryos—and even more sperm—in storage and ready to be used. When the person who provided those cells dies, like Peter, who gets to decide what to do with them? Read the full story.

—Jessica Hamzelou

This story is from The Checkup, our weekly newsletter giving you the inside track on all things biotech. Sign up to receive it in your inbox every Thursday.

Read more of our fascinating reproductive technology stories:

+ Inside the race to create egg and sperm cells in the lab. It uses a technique that could potentially solve infertility problems and allow more people to have children that are genetically-related to them. Read the full story

+ Scientists have found a way to mature eggs from transgender men in the lab. It could offer them new ways to start a family—without the need for distressing IVF procedures. Read the full story.  + How reproductive technology is changing what it means to be a parent. Advances could lead to babies with four or more biological parents—forcing us to reconsider parenthood. Read the full story.

The must-reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 Elon Musk wants to reinstate banned Twitter accounts
It’s an incredibly dangerous decision with widespread repercussions. (WP $) 
+ Recent departures have hit Twitter’s policy and safety divisions hard. (WSJ $)
+ It looks like Musk’s promise of no further layoffs was premature. (Insider $)
+ Meanwhile, Twitter Blue is still reportedly launching next week. (Reuters)
+ Imagine simply transferring your followers to another platform. (FT $)
+ Twitter’s potential collapse could wipe out vast records of recent human history. (MIT Technology Review)

2 Russia’s energy withdrawal could kill tens of thousands in Europe 
High fuel costs could result in more deaths this winter than the war in Ukraine. (Economist $)
+ Higher gas prices will also hit Americans as the weather worsens. (Vox)
+ Ukraine’s invasion underscores Europe’s deep reliance on Russian fossil fuels. (MIT Technology Review)

3 FTX is unable to honor the grants it promised various organizations 
Many of them are having to seek emergency funding to plug the gaps. (WSJ $)
+ Bahamians aren’t thrilled about what its collapse could mean for them. (WP $)

4 It’s a quieter Black Friday than usual
Shopping isn’t much of a priority right now. (Bloomberg $)
+ If you do decide to shop, make sure you don’t get scammed. (Wired $)

5 The UK is curbing its use of Chinese surveillance systems 
But only on “sensitive” government sites. (FT $)
+ The world’s biggest surveillance company you’ve never heard of. (MIT Technology Review)

6 Long covid is still incredibly hard to treat 
Its symptoms vary wildy, which can make it hard to track, too. (Undark)
+ A universal flu vaccine is looking promising. (New Scientist $)

7 San Francisco’s police is considering letting robots use deadly force
The force has 12 remotely piloted robots that could, in theory, kill someone. (The Verge)

8 Human hibernation could be the key to getting us to Mars 
It could be the closest we can get to time travel. (Wired $)

9 Why TikTok is suddenly so obsessed with dabloons 
It’s a form of choose-your-own-adventure fun. (The Guardian)

10 We can’t stop trying to reinvent mousetraps 🧀
There are thousands of versions out there, yet we keep coming up with new designs. (New Yorker $)

Quote of the day

“I get compared to a car crash or a train wreck a lot, where people can’t look away, which I enjoy hearing very much.”—Eli Betchik, one of TikTok’s most popular rage-bait chefs, explains their joy at concocting—and consuming—the most revolting recipes possible in front of a horrified audience to The Verge.

The big story

Toronto wants to kill the smart city forever

June 2022

In February, the city of Toronto announced plans for a new development along its waterfront. They read like a wish list for any passionate urbanist: 800 affordable apartments, a two-acre forest, a rooftop farm, a new arts venue focused on indigenous culture, and a pledge to be zero-carbon.

The idea of an affordable, off-the-grid Eden in the heart of the city sounds great. But there was an entirely different urban utopia planned for the same 12-acre plot, known as Quayside, just a few years ago.

But controversy ensued almost from the moment the project, run by Sidewalk Labs, the urban innovation arm of Alphabet, was announced in 2017. It struggled to build a neighborhood “from the internet up,” and the company pulled the plug in May 2020. With its very top-down approach, Sidewalk failed to comprehend Toronto’s civic culture—and doomed its vision for a smart city in the process. Read the full story.

—Karrie Jacobs

We can still have nice things

A place for comfort, fun and distraction in these weird times. (Got any ideas? Drop me a line or tweet ’em at me.)

+ This one’s for all of Duolingo’s victims across the world.
+ David Lynch is the master of musical suspense. Here’s just a couple of his best moments.
+ These fisherman sweaters, or jumpers, as we call them in the UK, look both beautiful and incredibly cozy.
+ Lao Gan Ma Spicy Chilli Crisp, a spicy chilli sauce and national Chinese treasure, sounds right up my street.
+ Happy-almost birthday to Flossie, who at 26 is officially the world’s oldest living cat!

Winc CMO on Complexities of Online Wine Sales

The story of Winc reflects the challenges of selling alcohol online in the U.S. The company produces and sells wine via direct-to-consumer memberships and wholesale to physical retailers. It launched in 2011 as Club W, rebranded to Winc in 2016, adopted in-house-only products, and went public in 2021. The stock (NYSE: WBEV) sells at 32 cents per share.

Jai Dolwani is Winc’s chief marketing officer, responsible for DTC sales, ecommerce, and engineering and technology — among other roles.

He and I recently discussed Winc’s journey and his role in the company. Our entire audio conversation is embedded below. The transcript is edited for clarity and length.

Eric Bandholz: Tell us about what you do.

Jai Dolwani: I’m a chief marketing officer at Winc, a wine-club membership company. We sell direct-to-consumer and wholesale at Trader Joe’s, Whole Foods Market, Target, restaurants, and bars. We have a few dozen in-house brands on our site, and we’re building a portfolio of wines focused on the next generation of consumers.

We sell only our own products and have a team of incredible winemakers. We launched in 2011 as Club W. In 2016, before I arrived, we re-branded to Winc. That’s when we shifted from selling third-party wines to creating in-house products and brands.

We don’t own vineyards or production facilities. We buy grapes directly from growers. Our wine-making team is responsible for the end-to-end process of getting that into a bottle.

Selling alcohol online is a difficult business. Shipping it is equally difficult owing to the weight and fragility.

U.S. laws surrounding the sale of beverage alcohol date to the 1920s prohibition era. It’s a three-tier distribution system of complex rules and regulations.

For example, some states have lifetime caps on the amount of alcohol to ship into that region. We can no longer ship there once we’ve hit a specific lifetime value — ever. For other states, it depends on where the wine was produced or bottled.

Plus, states have various marketing regulations. We can say “shipping included” and “zero-dollar shipping” but not “free shipping.”

Bandholz: You have an innovative subscription model.

Dolwani: Two years ago, we transitioned to a credits-based system. We acquire subscriptions through a discounted first-time purchase. After that, customers receive 60 credits on their accounts every month. Those credits roll over and never expire. Customers do not have to order every month.

We previously had the traditional model of receiving four wines every month or every quarter. But with automatic shipments, we had a lot of delivery headaches as, by law, customers had to be home to sign for the shipment.

We switched to the credit model for that reason and from customer feedback.

An added benefit of the new system is better engagement. Digital customers coming to the site, viewing our products, and selecting what they want provides key data on what has the best chance of success in physical wholesale channels.

Bandholz: What happens if customers don’t use their credits?

Dolwani: We want buyers to use 100% of their credits. If they’re not using the product, they will not be a long-term customer. We’re consistently emailing them if they have unused credits, saying, “You have a lot of credits. You should probably use them.” If they’re unresponsive to emails, we’ll offer incentives and, also, use direct mail.

But it’s a tricky balance. Reminding customers of unused credits can prompt them to cancel, as they aren’t using the service. So it’s important to communicate in a way that’s merchandised and product-forward, not necessarily highlighting large discounts or the lack of use.

Bandholz: Tell us more about customer acquisition.

Dolwani: We have a traditional, three-fold mix — Facebook, Google, and affiliates. Our ability to scale on Facebook through iOS 14.5 and increased shipping costs was possible only because of continuous improvements on ad creatives and looking at the sales funnel holistically.

In June 2021, we overhauled all of our advertising to use creators and landing pages with better ad-to-page relevancy. We retooled our entire acquisition funnel for the next generation.

Looking at the entire funnel helps keep Facebook a big part of our mix. Google is steady. It doesn’t scale too far up or down.

Our affiliate network has been huge for us. It accounts for a good, reliable portion of our customer acquisition. Using pay-per-post influencers was incredibly successful for us. But much of the engagement shifted from Instagram Stories to TikTok.

Bandholz: Where can listeners support you?

Dolwani: They can buy our products at Winc.com. My Twitter is @jaidolwani, and I’m on LinkedIn.