The week in AI: Showrunner - the team pioneering an 'AI Netflix' where you can create your own show

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each Thursday, we aggregate the major developments in artificial intelligence; we pass along the news, useful resources, tools and services, and highlight the top research in the field as well as exciting developments in open source. Even if you aren’t a machine learning engineer, we’ll keep you in touch with the most important developments in AI.

NEWS & OPINION

San Francisco based AI-entertainment startup The Simulation launches Showrunner

-------------------------

Nearly a year ago, a team named Fable Studio went viral for a showcase of their AI-generated South Park demo and related research paper. While the quality of the content was up for debate, what the demo represented was not. Even a year ago, it was clear that this team (now called The Simulation) was on to something - integrating cutting-edge technologies to generate AI-driven animated content. And now, they’ve announced Showrunner.

How does Showrunner work? Language models are trained on extensive datasets and can generate television show scripts, while state-of-the-art diffusion models create visual content that aligns with the stylistic and thematic elements of the shows. This process is brought to life by a multi-agent simulation system. The system utilizes data such as character histories, goals, and emotional states to inform the behavior and interactions of characters within the simulated narrative - et voilà, you have an AI-generated TV show.

Showrunner is in alpha and plans to launch with 10 original shows. Users will have the ability to generate new episodes and edit deeper into scripts, shots, and voices. If you’re interested, you can apply to leapfrog the 50,000+ waitlist, and select user-created episodes will eventually receive payment, revenue sharing, and IMDB credits.

Based on the previews available, some of these shows might have a shot to make something compelling, but the results are still very robotic and generally not very good. In our opinion, the Ikiru Shinu preview stands out perhaps in part because a) it’s a Japanese subtitle and therefore less jarringly AI-sounding for an English viewer and b) the detached monotone of the voices seems to fit the style. Even if there are some major hurdles to clear around this kind of AI-driven content creation, Showrunner could be brewing something big here.

But… there’s also a valid critical lense on this direction of the technology given the situation in Hollywood after the SAG-AFTRA strike. It’s not difficult to see why someone like Tyler Perry would cancel an $800m studio expansion over concerns about AI, why Scarlett Johansson is up in arms about OpenAI’s voice assistant sounding just like her, and why research estimates that up to 200,00 jobs in Hollywood could be displaced in the next three years by generative AI.

AI researchers are demanding the ‘Right to Warn’

-------------------------

Current and former employees from leading AI organizations, including OpenAI, Anthropic, and Google DeepMind, have authored a clarion call open letter highlighting the need for improved transparency and accountability in the development and deployment of AI technologies. The letter underscores a range of potential risks associated with advanced AI - from exacerbating social inequalities to the possibility of losing control over systems. The researchers want a framework where employees can voice concerns about these risks without facing retaliation. They emphasize the need to shift to a culture that prioritizes safety and ethical considerations over rapid advancement - something that would resonate with the average American’s opinion on AI.

The signatories propose several measures aimed at fostering an environment of open discourse and accountability within AI companies. These include:

• The elimination of agreements that prevent employees from openly criticizing the company's approach to AI risk

• The creation of anonymous processes for raising concerns directly to the company’s board or appropriate regulators

• Assurance that whistleblowers are protected against any form of retaliation

The letter has thus far been endorsed by renowned AI experts Yoshua Bengio, Geoffrey Hinton, and Stuart Russell. OpenAI recently had to backtrack on aggressive exit NDA’s, where they threatened to revoke vested equity in the company from anyone who criticized them. This letter sets the stage well for the last story we’re covering today:

Former OpenAI researcher releases 165-page essay on the trajectory of AGI

-------------------------

Former OpenAI researcher Leopold Aschenbrenner has unveiled a new essay that casts a profound vision on the trajectory of artificial general intelligence (AGI) and its implications. Aschenbrenner outlines the exponential scale-up in computing power, projecting that machines will surpass college graduates in general cognitive abilities by 2025/26 and will achieve superintelligence by the end of the decade. In the essay, he considers the transformative impact this will have on various sectors, particularly highlighting a potential escalation in national security measures and the increasing ferocity of the tech/AI race - mostly between the U.S. and China. He calls it a national security issue that needs its own ‘Manhattan Project’.

And it feels right to say this here, relative to the story above: Aschenbrenner revealed on the Dwarkesh podcast that he was fired from OpenAI after raising AI security concerns, while reiterating his viewpoints on the podcast.

The essay also elaborates on the critical milestones from GPT-4 to potential AGI, focusing on the rapid advancements due to increased computate power and algorithmic efficiencies. Aschenbrenner uses the term "counting the OOMs" (orders of magnitude) to predict future capabilities, suggesting another qualitative leap similar to a jump from preschool to high school level intelligence (GPT-2 to GPT-4) within the next few years. The projection is based on continuing trends in both hardware and machine learning breakthroughs, which have - so far - consistently delivered exponential improvements in model performance.

MORE IN AI THIS WEEK

• Breaking: AI giants Microsoft, Nvidia, and OpenAI will face a federal antitrust investigation

• A top secret meeting is kicking off in Madrid with the CEOs of Google DeepMind, Anthropic and Microsoft AI in attendance

• The CEO of Zoom wants AI clones in meetings

• Most downloaded US news app, ‘NewsBreak‘ has Chinese roots and 'writes fiction' using AI

• Elon Musk had Nvidia send AI chips intended for Tesla to X and xAI

• Google rolls back AI Overviews for some searches after flubs and flaws

• Stanford University team apologizes over claims they copied Chinese project for AI model

• Apple made a once-unlikely deal with Sam Altman to catch up in AI

• AI apocalypse? ChatGPT, Claude and Perplexity all went down at the same time

TECHNICAL, RESEARCH & OPEN SOURCE

Nvidia unveils next-gen AI chip platform at Computex in Taipei

-------------------------

Nvidia CEO Jensen Huang spoke about a number of new AI announcements during a keynote at the 2024 Computex conference, including a new AI-powered PC gaming assistant, and AI tools for creating lifelike avatars.

Huang also detailed a roadmap of Nvidia’s AI acceleration platforms, mentioning an upcoming Blackwell Ultra chip slated for 2025 and their next-gen AI platform called "Rubin" set for 2026.

Huang appeared hesitant to make the Rubin announcement, leading with, “I'm not sure yet whether I'm going to regret this.” It’s not hard to imagine why - Nvidia only three months ago revealed Blackwell, which is barely out of the gate and not yet widely shipping. There’s potential for an Osborne effect here, as Nvidia seems to be so far ahead of everyone else in AI that they are actively leapfrogging themselves. Announcing these major upgrades just over the horizon might impact sales of their current offerings.

The Rubin AI platform is expected in 2026 and will use HBM4 (high-bandwidth memory) and NVLink 6 Switch, operating at 3,600GBps. Huang did not provide extensive specifications for the upcoming products, but did make it clear that cost and energy savings were paramount factors related to continued development.

Nvidia currently possesses an estimated 70–95 percent of the data center GPU market share, so despite the Osborne effect the Rubin reveal seems like a confident and maybe even calculated ‘risk’. Huang has been consistently eager to showcase the company's plans to continue pushing silicon fabrication tech to its limits and widely broadcast that Nvidia plans to keep releasing new AI chips at a steady cadence.

MORE IN T/R/OS:

• Intel at Computex - “bringing AI everywhere”

• AMD at Computex - “the future of high performance computing in the AI era”

• Groqbook: Generate entire books in seconds using Groq and Llama3

• World's first bioprocessor uses 16 human brain organoids for ‘a million times less power’ consumption than a digital chip

• SaySelf: teaching LLMs to express confidence with self-reflective rationales

• (Microsoft research) Introducing Aurora: The first large-scale foundation model of the atmosphere

• Mistral launches fine-tuning API for Mistral 7B and Mistral Small models

That’s all for this week. We’ll see you next Thursday.