Hollywood strikes and AI - H+ Weekly - Issue #424
This week - Meta releases Llama 2; proof that ChatGPT gets dumber; a startup that aims to make lab-grown human eggs; CRISPR breeds greener poplar trees; and more!
On May 2nd, 2023, 11,500 TV and movie screenwriters went on strike in the US. On July 14th, actors from SAG-AFTRA union joined the strike. It is the first time since 1960 that both actors and screenwriters went on strike together and brought Hollywood to a stop.
The main reason for this strike is the changing landscape and the business model of making movies and TV without adequate changes in actors’ and screenwriters’ contracts. The streaming revolution resulted in shorter periods of time when actors and screenwriters are working. Previously, screenwriters could count on 20-40 weeks of guaranteed work, but now they face a mere 6-10 weeks. There is also the issue of residuals. During the time of network TV, screenwriters and actors could count on payments from reruns on TV. In streaming, those residuals are far smaller. This caused their jobs to be less stable with lower income.
And then there is AI.
Screenwriters and actors are concerned their work could be replaced by AI. Creating scripts for new TV shows and movies could be delegated to tools like GPT-4, reducing the need for hiring human screenwriters.
Actors are concerned about studios using AI to recreate their likenesses without consent or without compensation. Studios approached striking actors with a “groundbreaking deal” suggesting that they could scan background actors and use their likenesses indefinitely, offering only a one-time payment equivalent to a single day's pay.
AI tools for making movies are already here. There is an entire suite of such tools as Runway which makes editing easier and can generate videos from prompts or change the style of footage. There is ChatGPT and other text generators that can generate an entire script or at least a good enough outline for a human screenwriter to build upon. With the Unreal Engine, it is possible to relatively easy build entire virtual sets with high enough quality to look like real sets. Unreal also offers Metahuman - a tool that enables the creation of realistic virtual human characters. And there are voice cloning tools and voice generators specifically designed to replace human voice actors.
Just this week, researchers unveiled SHOW-1 - an AI that “will write, animate, direct, voice, edit for you”. As an example, they showed an entirely new episode of South Park generated from a prompt.
Filmmakers are already experimenting with these tools to see what is possible. For example, the short movie below has been created using AI-powered tools to help with writing the script and creating the concept art. The AI also generated all the voices and participated in some creative decisions. But everything else - editing, the idea, sound design, etc. was made by a human.
Hollywood and the movie industry are in crisis. The studios are playing safe. The movies we see in theatres are either superhero movies, remakes, sequels or movies based on established brands like Barbie. Many of those movies, with budgets in hundreds of millions of dollars, struggle to break even.
The last time both actors and screenwriters went on strike was in 1960. Back then, a new emerging technology threatened the movie industry - the TV. Studios played safe and were making only westerns, war movies or musicals. The big blockbuster movies such as Cleopatra were financial disasters. The old studios went out of business or had to reinvent themselves. From this dark time in Hollywood history emerged a new wave of creative filmmakers such as Francis Ford Coppola, Stanley Kubrick, George Lucas, Steven Spielberg and others who had good stories to tell. This wave of New Hollywood created many of the best movies ever created - Star Wars, Jaws, The Shining, Blade Runner, A Clockwork Orange, and every iconic movie of the late 1960s, 70s and 80s.
Maybe something similar will happen now. Maybe a new generation of creative filmmakers will see AI as a tool to enhance new and interesting stories. Maybe we will see an explosion of creativity, first on YouTube and TikTok, and then on small and big screens. Maybe from the current crisis, a new, healthier movie industry will emerge, just like the New Hollywood emerged in the 1960s, full of new creators enhancing their stories with AI.
Becoming a paid subscriber now would be the best way to support the newsletter.
If you enjoy and find value in what I write about, feel free to hit the like button and share your thoughts in the comments. Share the newsletter with someone who will enjoy it, too. That will help the newsletter grow and reach more people.
🦾 More than a human
Startup aims to make lab-grown human eggs, transforming options for creating families
Not so long ago, scientists found a way to turn any human cell into an egg cell and now we have a company that tries to commercialise this technique. The company, named Conception, promises that this method can revolutionise how humans reproduce and allow women who have lost their eggs to cancer treatment, women who have never been able to produce healthy eggs and women whose eggs are no longer viable because of their age to have children. This method also opens the possibility for same-sex couples to have biological children. However, the company has released few details about its experiments and hasn't published its results in a scientific journal. Independent scientists haven't been able to validate their claims.
This company plans to transplant gene-edited pig hearts into babies next year
The biotech company eGenesis has successfully transplanted pig hearts into infant baboons and plans to perform the same operation on human babies next year. eGenesis uses CRISPR gene-editing to modify around 70 genes in the pig's genome to make the organs suitable for human transplantation. The company's goal is to provide temporary pig heart transplants for human babies while they wait for human heart donors.
🧠 Artificial Intelligence
Introducing Llama 2
Meta has released Llama 2 - an open-source successor to their accidentally open-source large language model (Llama 1 leaked early this year and became a base for many open-source projects). The new model has been trained on 40% more data than Llama 1 and has doubled the context to 4096 tokens. Llama 2 is also available for free for research and commercial use (Llama 1 licence only covered research use). Perfomance-wise, Llama 2 is not the best model out there available but thanks to it being open source it might get improved very quickly. For more details about Llama 2, you can read the paper describing the model in a surprisingly open and detailed way, or check this post from Charlie Guo which does a good job of explaining Meta’s newest model.
China unveils provisional rules for generative AI, including a licensing regime
China’s top cyberspace regulator unveiled a set of provisional rules to govern generative AI services, including API providers, that serve China-based users. The rules focus on balancing development and security and will require providers to adhere to core socialist values, register algorithms that influence public opinions, and obtain administrative licenses. They must also ensure non-discriminatory and anti-addiction practices for users. Providers are responsible for identifying and stopping the generation of illegal content and sharing details of AI models with regulators. The new rules also call for “self-reliant innovation” in AI algorithms, frameworks, chips, software platforms and other infrastructure, while still encouraging “equal and mutually beneficial” international cooperation.
How Is ChatGPT’s Behavior Changing over Time?
Recently, more and more people started to share the impression that ChatGPT has become dumber. Thanks to this paper, we now have concrete data to support this claim. The study compares the performance of ChatGPT in tasks such as math problem-solving, code generation, and visual reasoning between the March and June versions of GPT-3.5 and GPT-4. The researchers have indeed found that ChatGPT's performance has changed over time. In some cases, they observed a significant downgrade in its capabilities.
Apple is testing a ChatGPT-like AI chatbot
Apple has joined the generative AI train. Reports emerged that the company is working on its own chatbot which is currently used internally by Apple employees. Apple’s chatbot is similar to Bard, ChatGPT and Bing AI, as it doesn’t feature any additional features that separate it from what’s currently commercially available.
Bill Gates: The risks of AI are real but manageable
”Whether it was the introduction of cars or the rise of personal computers and the Internet, people have managed through other transformative moments and, despite a lot of turbulence, come out better off in the end”, writes Bill Gates. Gates acknowledges there are and will be issues caused by the misuse of AI but there are historical precedents that show these challenges can be managed successfully. Gates believes that with proper management, regulations, and collaboration, AI's benefits can outweigh its risks, and society can navigate through the transformative impact of AI successfully.
🤖 Robotics
Robotic Bees Could Support Vertical Farms Today and Astronauts Tomorrow
Vertical farms have many problems to overcome before they can replace traditional farms. One of them is that some plants need to be pollinated by bees and bees have trouble navigating under artificial light. Pollinating by hand is too complicated. So people are turning to robotics for solutions. But don’t expect tiny robots resembling bees. Current solutions include drones and multi-arm robots.
Humanoid robot GR-1 to aid aging population in China
Chinese company Fourier Intelligence has revealed plans to manufacture 100 of its GR-1 general-purpose humanoid robots by the end of 2023 to help patients with rehabilitation. Zen Koh, CEO and Co-founder of Fourier Intelligence said, “As we move forward, the entire GR-1 could be a caregiver, could be a therapy assistant, can be a companion at home for the elderly who stay alone.” The new robot design is just over 5 feet (1.64m) tall and weighs 120 pounds (55kg). It can walk, avoid obstacles and perform simple tasks like holding bottles. The robot design is currently in the research and development phase, and the prototype will take up to two-three years to complete.
Robots are transforming every aspect of human life. This also includes warfare. In this video, Isaac Arthur takes the battle robots we see in science fiction and checks how feasible their designs are and what is the most likely path robotics will take on the near and far future battlefields.
🧬 Biotechnology
CRISPR breeds greener poplar trees
Researchers used a CRISPR gene-editing system to breed poplar trees with reduced levels of lignin, the major barrier to the sustainable production of wood fibres, while improving their wood properties. The findings hold promise to make fibre production for everything from paper to diapers greener, cheaper, and more efficient.
MIT’s vaccine-enhanced CAR-T cell therapy destroys solid tumors
MIT researchers have demonstrated how pairing CAR-T cell therapy with a cancer vaccine boosts its efficacy in mice with two deadly types of solid tumours: glioblastoma, a fast-growing brain cancer, and melanoma, the most serious kind of skin cancer. “In principle, this should apply to any solid tumour where you have generated a CAR-T cell that could target it,” said senior author Darrell Irvine.
Generative AI imagines new protein structures
Researchers from MIT CSAIL lab have created a new tool named FrameDiff which uses stable diffusion, a technique used in image generators such as Midjourney, to generate new protein structures. Some of its creations are completely new, never seen in nature structures.
H+ Weekly sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.
A big thank you to my paid subscribers and to my Patreons: whmr, Florian, dux, Eric and Andrew. Thank you for the support!
You can follow H+ Weekly on Twitter and on LinkedIn.
Thank you for reading and see you next Friday!
Very nice summary.