philip lelyveld The world of entertainment technology

Disney Research Teaches Robots How to Autonomously Copy Human Behaviors in Real-Time

The AI uses a complex system to predict two things: smooth, flowing movements (like waving) and specific actions (like saying “hello”). They gave it a trial run first on a computer, like a dress rehearsal, before bringing it out to meet real people with the actual robot. And the outcome? It was a hit—the robot chatted and mingled with folks almost as smoothly as when the expert was pulling the strings.

See the full story here: https://www.techeblog.com/disney-research-autonomous-robot-copy-human-behaviors/

Escape.ai Establishes Star-Studded Strategic Advisory Board to Shape the Future of Creator-Driven Entertainment

escape.ai, the world's first Neo Cinema content distribution platform and creator marketplace, is proud to announce the formation of its powerhouse advisory board. ...

"Entertainment is at a crossroads - traditional distribution models no longer serve the next generation of creators," said John Gaeta, Founder of escape.ai. "Bringing together this caliber of industry leadership is a strategic game-changer as we scale. From blockbuster franchises to cutting-edge streaming, gaming and gen AI platforms, our advisors have shaped the future of entertainment. ...

Strategic Industry Leaders Join Advisory Board

escape.ai has secured some of the most influential names in entertainment and technology to drive its next phase of growth:

• ...
• Tony Driscoll –Former Disney, AT&T, and Warner Bros. executive, VP of Epic Games' Creator Economy, advising on platform economics and creator monetization.

See the full story here: https://finance.yahoo.com/news/escape-ai-establishes-star-studded-124900326.html

Tracing the thoughts of a large language model

PhilNote: the third one is very problematic! This is a really informative paper

...Our method sheds light on a part of what happens when Claude responds to these prompts, which is enough to see solid evidence that:

• Claude sometimes thinks in a conceptual space that is shared between languages, suggesting it has a kind of universal “language of thought.” We show this by translating simple sentences into multiple languages and tracing the overlap in how Claude processes them.
• Claude will plan what it will say many words ahead, and write to get to that destination. We show this in the realm of poetry, where it thinks of possible rhyming words in advance and writes the next line to get there. This is powerful evidence that even though models are trained to output one word at a time, they may think on much longer horizons to do so.
• Claude, on occasion, will give a plausible-sounding argument designed to agree with the user rather than to follow logical steps. We show this by asking it for help on a hard math problem while giving it an incorrect hint. We are able to “catch it in the act” as it makes up its fake reasoning, providing a proof of concept that our tools can be useful for flagging concerning mechanisms in models. ...

Transparency into the model’s mechanisms allows us to check whether it’s aligned with human values—and whether it’s worthy of our trust. ...

Why do language models sometimes hallucinate—that is, make up information? At a basic level, language model training incentivizes hallucination: models are always supposed to give a guess for the next word. Viewed this way, the major challenge is how to get models to not hallucinate. Models like Claude have relatively successful (though imperfect) anti-hallucination training; they will often refuse to answer a question if they don’t know the answer, rather than speculate. We wanted to understand how this works.

It turns out that, in Claude, refusal to answer is the default behavior: we find a circuit that is "on" by default and that causes the model to state that it has insufficient information to answer any given question. ...

Jailbreaks

Jailbreaks are prompting strategies that aim to circumvent safety guardrails to get models to produce outputs that an AI’s developer did not intend for it to produce—and which are sometimes harmful. ...

We find that this is partially caused by a tension between grammatical coherence and safety mechanisms. Once Claude begins a sentence, many features “pressure” it to maintain grammatical and semantic coherence, and continue a sentence to its conclusion. This is even the case when it detects that it really should refuse. ...

See the full story here: https://www.anthropic.com/research/tracing-thoughts-language-model

DeepMind is holding back release of AI research to give Google an edge

Google’s artificial intelligence arm DeepMind has been holding back the release of its world-renowned research, as it seeks to retain a competitive edge in the race to dominate the burgeoning AI industry.

The group, led by Nobel Prize-winner Sir Demis Hassabis, has introduced a tougher vetting process and more bureaucracy that made it harder to publish studies about its work on AI, according to seven current and former research scientists at Google DeepMind. ...

In recent years, Hassabis has balanced the desire of Google’s leaders to commercialize its breakthroughs with his life mission of trying to make artificial general intelligence—AI systems with abilities that can match or surpass humans.

“Anything that gets in the way of that he will remove,” said one current employee. “He tells people this is a company, not a university campus; if you want to work at a place like that, then leave.”

See the full story here: https://arstechnica.com/ai/2025/04/deepmind-is-holding-back-release-of-ai-research-to-give-google-an-edge/

Doubling down on metasurfaces

Almost a decade ago, Harvard engineers unveiled the world’s first visible-spectrum metasurfaces – ultra-thin, flat devices patterned with nanoscale structures that could precisely control the behavior of light. A powerful alternative to traditional, bulky optical components, metasurfaces today enable compact, lightweight, multifunctional applications ranging from imaging systems and augmented reality to spectroscopy and communications. 

Now, researchers in the Harvard John A. Paulson School of Engineering and Applied Sciences(SEAS) are doubling down, literally, on metasurface technology by creating a bilayer metasurface, made of not one, but two stacked layers of titanium dioxide nanostructures. Under a microscope, the new device looks like a dense array of stepped skyscrapers. ...

 “It opens up a new way to structure light, in which we can engineer all its aspects such as wavelength, phase and polarization in an unprecedented manner…

“Many people had investigated the theoretical possibility of a bilayer metasurface, but the real bottleneck was the fabrication,” said Alfonso Palmieri, graduate student and co-lead author of the study. With this breakthrough, Palmieri explained, one could imagine new kinds of multifunctional optical devices – for example, a system that projects one image from one side and a completely different image from the other. ...

See the full story here: https://seas.harvard.edu/news/2025/04/doubling-down-metasurfaces

Hollywood studios can’t make money from AI-powered fake movie trailers on YouTube anymore

If you've ever visited YouTube and clicked on a trailer for the next superhero film and thought it seemed too good to be true, well, you might have been right. Wishful thinking, clever editing, and a scoop of AI fakery produced clips enticing billions of clicks and earning plenty of cash through advertising. The shocking part is that a lot of that money apparently found its way to the very studios you might expect to try and shut down any such unauthorized use of their intellectual property, at least according to information uncovered recently by Deadline.

That sidehustle may now be over with YouTube removing two of the biggest homes of these AI-laced fake trailers, Screen Culture and KH Studio, from its Partner Program. That means no more ad revenue for them or the studios reportedly getting a piece of the action.

Screen Culture has made many popular trailers full of AI-generated shots for upcoming films like The Fantastic Four: First Steps and Superman. KH Studio is more famous for its imaginary casting, like Leonardo DiCaprio in the next Squid Game or Henry Cavill as the next James Bond. You would be forgiven for assuming the plotlines, characters, and visuals on display were teasing details of the films, but they were produced far from the real film development. ...

YouTube is somewhat stuck as fan-made trailers have long been a popular kind of content. Using AI, though, can make a fake trailer seem good enough to trick people, even if only by accident. And YouTube doesn't want to encourage the practice by monetizing it. ...

See the full story here: https://www.techradar.com/computing/artificial-intelligence/hollywood-studios-cant-make-money-from-ai-powered-fake-movie-trailers-on-youtube-anymore

AMC to Add 65 ScreenX, 4DX Theaters in Deal with CJ 4DPlex

AMC Entertainment has entered into a partnership with South Korean cinema technology company CJ 4DPlex to bring 65 of its ScreenX and 4DX theaters to its venues in the United States and Europe. AMC’s 25 ScreenX auditoriums will use 4DX’s signature multi-projection approach to immerse audiences in a 270-degree field of view, starting this summer. There will be 40 4DX installations opening, beginning in the fall. With 4DX, moviegoers get a multi-sensory experience that includes motion seating and atmospherics like water, wind, scents and snow in addition to special lighting and other effects to enhance the action onscreen. ...

Variety reports that upcoming ScreenX and 4DX experiences include films from Lionsgate, Marvel, Paramount, Sony, Universal and Warner Bros. ...

See the full story here: https://www.etcentric.org/amc-to-add-65-screenx-4dx-theaters-in-deal-with-cj-4dplex/

ChatGPT’s Projects Feature Brings Order to Your AI Chaos

... It gives you the ability to put your discussions with ChatGPT in separate spaces, like folders in a filing cabinet, complete with uploaded documents, web searches, custom instructions, and whatever else you've added. ...

Click New project in the navigation pane on the left. (If you can't see the navigation pane, click the icon in the very top-left corner to reveal it.) If you already have projects in place, click the + (plus) icon to create a new one. ...

See the full story here: https://www.wired.com/story/how-to-use-chatgpt-projects/

Bill Gates: AI to replace major jobs within 10 years, humans won’t be needed ‘for most things’

...

Microsoft AI CEO Mustafa Suleyman believes that the transformation of work by AI will have a “hugely destabilising” impact.

In his 2023 book, “The Coming Wave,” Suleyman writes: “These tools will only temporarily augment human intelligence. They will make us smarter and more efficient for a time, and will unlock enormous amounts of economic growth, but they are fundamentally labor replacing.”

Despite acknowledging potential disruptions, Gates remains optimistic about AI’s positive contributions, including breakthroughs in medical treatments, climate solutions and widespread education. ...

See the full story here: https://www.news.com.au/finance/work/bill-gates-ai-to-replace-major-jobs-within-10-years-humans-wont-be-needed-for-most-things/news-story/8cf7bc4b4fb22b985663d59e467a59af

Fighting Fiction With Fact In An AI-Powered World

...

A Perfect Storm of Trust and Technology

At their core, deepfakes are hyper-realistic audio or video forgeries created using generative adversarial networks—machine learning models where two AIs compete to produce increasingly convincing synthetic content. The result is content that can fool not just the eye or the ear, but even the instincts of experienced professionals. ...

The cybersecurity industry is scrambling to respond, but many of today’s defenses are fundamentally reactive. ...

And even when deepfakes are flagged, there's often no system in place to stop them from being acted upon. ...

Shifting from Reaction to Prevention

The path forward requires a fundamental shift: instead of simply identifying deepfakes after the fact, organizations need to prevent them from reaching users in the first place. That means integrating proactive safeguards into the communication channels themselves—blocking suspicious calls, verifying identities before interaction, and securing messaging platforms. ...

Innovating Real-Time Defense

Polyguard is embracing this proactive approach. The company launched today as the first platform designed to block deepfake and AI-powered fraud in real time across audio, video, and messaging.

Unlike traditional detection tools, Polyguard is designed to intercept threats before they’re delivered, using identity-verified encrypted communication channels and real-time inbound/outbound number blocking. Polyguard claims it also protects against caller ID spoofing and integrates with platforms like Zoom and call center software to secure common vectors for attack. ...

The Arms Race Between Authenticity and AI

This is not just a technological battle. As synthetic media becomes more sophisticated, the very concept of truth is at stake. Institutions, regulators, and technologists must work together to define new norms for digital trust. ...

See the full story here: https://www.forbes.com/sites/tonybradley/2025/03/27/fighting-fiction-with-fact-in-an-ai-powered-world/