Bringing back the dead with AI - H+ Weekly - Issue #410

This week - how OpenAI approaches AI safety; a new target to reverse ageing; Amazon joins generative AI race; cyborg cockroaches; turning stem cells into synthetic embryos; and more!

The advances in deepfake technology allow us to recreate any human being. This also includes those who are not alive anymore.

The public got the first taste of digital resurrection in 2012 when 2Pac, who died in 1996, was brought back as a hologram to perform at Coachella. Similar technology brought Micheal Jackson back to the stage for the 2014 Billboard Awards.

In 2016, Disney digitally recreated Peter Cushing (who died in 1994) to "play" Grand Moff Tarkin in Rogue One. And last year, James Earl Jones (who is 92 years old), the voice of Darth Vader, allowed Disney to recreate his iconic voice with AI in future Star Wars projects, meaning that at some time in the future, Jones’ voice will outlive him.

Although the first people digitally brought back to life with technology were celebrities, technology has become more accessible and available to anyone. We already leave tons of data about ourselves in form of voice and text messages, photos and videos. Powerful generative AIs like GPT-4 or open-source large language models like llama.cpp are easy to use. All you need to do to create a digital replica of someone is to feed that data into AIs and train them.

In 2017, James Vlahos created a digital replica of his dead father. Vlahos transcribed the recording of his father telling stories from his life and feed them into a chatbot. The chatbot then learned how to mimic his father’s way of speaking and, equipped with his memories, it was able to act like his father. In 2021, Joshua Barbeau used GPT-3 to recreate his dead fiancee Jessica.

These projects helped those who created them deal with the grief and get complete with the deaths of those close to them. I can see this technology being used in therapy to help people with grief.

Another possible application can be in education. For example, instead of learning about general relativity from a physics textbook, a digital replica of Albert Einstein could teach it. Or we can have interactive history lessons where AIs tell the stories as the people who experienced them first-hand.

And then there is always the entertainment industry, which as we have seen earlier, is already resurrecting dead artists. Not everyone, however, approves this. When a director of a documentary about Anthony Bourdain revealed some voice lines were generated with an AI simulation of the late chef's voice, some people questioned whether it was ethical. Some artists try to protect their likenesses from being used by AI. For example, Robin Williams restricted the use of his image for 25 years after his death and Keanu Reeves has a clause in his contracts forbidding any digital manipulations.

---

🦾 More than a human

Ageing studies in five animals suggests how to reverse decline
A new study investigated how ageing affects transcription in humans, fruit flies, rats, mice and worms and discovered the same ageing mechanism across all those different animals. As the animals grow older, the speed at which the enzyme that drives transcription, RNA polymerase II (Pol II), increases and that has an impact on overall ageing. Researchers tested this hypothesis on fruit flies. The ones with modified Pol II enzyme lived 10% to 20% longer than their non-mutant counterparts. These findings open the door to exploring how Pol II could be a target for drugs that slow down the ageing process.

Digital twins could save your life. Here’s how
The idea of a digital twin - a virtual replica of an existing object - is not something new. The growing computing power and a better understanding of how the human body works open up an interesting possibility - the creation of a human digital twin. A human digital twin will definitely find a place in medical research, potentially saving thousands of lives every year. Peter Coveney, one of the foremost global experts on the topic, believes virtual patients could be available for practical medical uses in about five years’ time.

🧠 Artificial Intelligence

OpenAI - Our approach to AI safety
OpenAI published this blog post outlining its approach to AI safety in response to the safety concerns around powerful AI systems. OpenAI stated it is “committed to keeping powerful AI safe and broadly beneficial” and explains how they worked for six months after GPT-4 was trained to make sure it is safer and more accurate than GPT-3.5.

Amazon announces 'Bedrock' AI platform to take on OpenAI
Amazon joins the generative AI race with a toolkit called Bedrock. It will contain a range of AI services, from multilingual large language models to image generators. All of them will be available on Amazon Web Services.

▶️ Yann LeCun and Andrew Ng: Why the 6-month AI Pause is a Bad Idea (32:03)
Well-known AI researchers, Yann LeCun and Andrew Ng, discuss the recent call to pause AI research and why it is a bad idea. They don’t see a point in slowing down the research when we need more research to make the technology safer. They also noted the potential positive impact of intelligence augmentation is tremendous for society and that stopping AI research is almost impossible - the best we can do is to maximize positive effects and minimize negative outcomes. The topic of AGI has also been raised and their conclusion is that we are far away from human-level AIs.

AI Index 2023
Stanford Institute for Human-Centered Artificial Intelligence released its annual AI Index report. There are some interesting findings in the report. Last year was the first year in the last decade in which private investment in AI decreased and the proportion of companies adopting AI has plateaued, but those who adopted AI are doing better. The demand for AI-related professional skills is increasing across virtually every American industrial sector. The number of AI-related accidents is also on the rise. And Chinese citizens are among those who feel the most positively about AI products and services. Americans… not so much. The full report is available here.

Generative Agents: Interactive Simulacra of Human Behavior
This is an interesting paper. Researchers from Stanford University and Google created a virtual town populated with 25 generative AI agents. What they have discovered is that these agents started to behave like humans. They wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day. The paper also notes an instance when one agent wanted to throw a Valentine’s Day party and other agents created invitations to spread the news. The authors of the paper hope their research will find a place in interactive applications ranging from design tools to social computing systems to immersive environments.

🤖 Robotics

Exploring movement optimization for a cyborg cockroach with machine learning
Remote-controlled cockroaches are nothing new (you can even buy a kit and make your own cyborg cockroaches). What scientists from Osaka University have done is built a system to make these cyborg cockroaches more efficient as robots with machine learning. "Equipped with such a system, the cyborg successfully increased its average search rate and travelled distance up to 68% and 70%, respectively, while the stop time was reduced by 78%," said the study authors.

▶️ Top Five Sci-Fi Robots (9:35)
Spacedock (an excellent YouTube channel talking mostly about spacecrafts from science fiction) made this list of the top five sci-fi robots. The list includes Geth from Mass Effect, Issac from The Orville and TARS from Interstellar. What would your top five be?

Once More, With Feeling: Exploring Relatable Robotics at Disney
Do you remember that adorable robot Disney engineers showed off at SXSW this year? In this article, Morgan Pope, one of the engineers working on that robot, shares how they built the robot and how they incorporated failure into this project to build an expressive robot that evokes human emotions.

🧬 Biotechnology

Scientists Turned Monkey Stem Cells Into ‘Synthetic Embryos’
Researchers in China have turned monkey stem cells into embryos. Although other researchers have created “synthetic” embryos before, it’s the first time anyone has done it with monkeys - animals closely related to humans - and tried to get them to implant in the uterus. This new method of creating embryos could allow studying how diseases, birth defects, and miscarriages arise, without the practical and ethical concerns raised by using actual embryos.

Cancer and heart disease vaccines ‘ready by end of the decade’
Dr Paul Burton, the chief medical officer of the pharmaceutical company Moderna, says that vaccines against jabs for cancer, cardiovascular and autoimmune diseases, and other conditions will be ready by 2030. “We will have that vaccine and it will be highly effective, and it will save many hundreds of thousands, if not millions of lives. I think we will be able to offer personalised cancer vaccines against multiple different tumour types to people around the world”, Burton said.

---

H+ Weekly is a free, weekly newsletter with the latest news and articles about AI, robotics, biotech and technologies that blur the line between humans and machines, delivered to your inbox every Friday.

Subscribe to H+ Weekly to support the newsletter under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

You can follow H+ Weekly on Twitter and on LinkedIn.

Thank you for reading and see you next Friday!