philip lelyveld The world of entertainment technology

‘You Can’t Lick a Badger Twice’: Google Failures Highlight a Fundamental AI Flaw

Here’s a nice little distraction from your workday: Head to Google, type in any made-up phrase, add the word “meaning,” and search. Behold! Google's AI Overviews will not only confirm that your gibberish is a real saying, it will also tell you what it means and how it was derived. ...

But two of its defining characteristics come into play when it explains these invented phrases. First is that it’s ultimately a probability machine; while it may seem like a large-language-model-based system has thoughts or even feelings, at a base level it’s simply placing one most-likely word after another, laying the track as the train chugs forward. That makes it very good at coming up with an explanation of what these phrases would mean if they meant anything, which again, they don’t. ...

The other factor is that AI aims to please; research has shown that chatbots often tell people what they want to hear. In this case that means taking you at your word that "you can't lick a badger twice"  is an accepted turn of phrase. In other contexts, it might mean reflecting your own biases back to you, as a team of researchers led by Xiao demonstrated in a study last year. ...

Compounding these issues is that AI is loath to admit that it doesn’t know an answer. When in doubt, it makes stuff up. ...

See the full story here: https://www.wired.com/story/google-ai-overviews-meaning/

The Man Who Wants AI to Help You ‘Cheat on Everything’

Last month, Roy Lee was suspended from Columbia after he was accused of using AI to “cheat” on technical job interviews for Amazon, Meta, and TikTok. On Sunday, he announced that he raised $5.3 million to start Cluely, a new startup that aims to allow users to similarly “cheat on everything.”  ...

Lee wants people to use Cluely on “Sales calls. Meetings. Negotiations,” the “manifesto” on the software’s website says. “If there’s a faster way to win—we’ll take it. We built Cluely so you never have to think alone again. It sees your screen. Hears your audio. Feeds you answers in real time. While others guess—you’re already right.” ...

“The second realm of negative feedback is that ‘cheating is inherently unethical and immoral.’ This is where we’re trying to do a lot of reframing around what exactly defines cheating, what cheating really is, and what cheating will look like in a future that is AI native,” he said. ...

I tried out a test version of Cluely in a mock interview with Emanuel Maiberg. We pretended Emanuel was interviewing me for a position at 404 Media and I used Cluely to shape my answers. ...

Then Cluely would answer the question it had just heard Emanuel ask me and put its answers in an overlay at the top of my screen. It would come in two pieces: a bright white script I could read verbatim and a darker piece of text below that explained the AI’s reasoning. Because of the way Cluely is programmed, it doesn’t show up in recordings or even screenshots. ...

It also wasn’t that impressive. A few questions into the conversation and Emanuel and I realized it was just feeding me ChatGPT answers to the questions as if I’d typed them in a browser. It also took 20 seconds each time to generate, with Emanuel and I staring at each other while we waited. ...

“The only thing the product really showcased [in the commercial] was ChatGPT in your glasses,” Lee said. ...

Both Lee and the marketing copy for Cluely hammer home that technical innovations are often thought of as cheating. Math teachers worried about the advent of calculators. ...

Every single thing that is rote memorization, that relies on facts that you don’t need in the moment, that are not intrinsically necessary for a human to learn, you won’t need that anymore.”

“Entire school systems will be gone. The entire K-12 education. Everybody sits in a room eight hours a day, takes multiple choice exams. In my future, everything we understand education to be, will be completely gone.” He said that people will explore whatever topics they want from the earliest ages in the “most efficient way possible.” ...

See the full story here: https://www.404media.co/the-man-who-wants-ai-to-help-you-cheat-on-everything/?utm_source=substack&utm_medium=email

Character.AI Introduces New Video Generator in Closed Beta

Character.AI, a platform offering AI chatbots for socializing and role play, has released a video generation model called AvatarFX in closed beta. Promising the ability to make photorealistic images “come to life — speak, sing and emote — all with the click of a button,” the technology combines audio and video to create a variety of visual style and voice, from realistic 3D — including “non-human faces (like a favorite pet)” — to 2D animations, according to the company. AvatarFX also has the ability “to maintain strong temporal consistency with face, hand and body movement” and can “power videos with multiple speakers.” ...

Character.AI says its AvatarFX video will be watermarked so people know it’s not “real.” Additionally, it plans to block the generation of video featuring minors and will run images of recognizable people through a filter that changes details to obscure their identity, according to TechCrunch, which concludes that “since AvatarFX is not widely available yet, there is no way to verify how well these safeguards work.” ...

“The voices in these videos are generated by Character.AI’s own text-to-speech (TTS) technology, making conversations and singing sound natural and smooth,” writes Mint.  ...

See the full story here: https://www.etcentric.org/character-ai-introduces-new-video-generator-in-closed-beta/

Shelly Palmer email

...

We’re just months away from living in a world where any given image (especially in media, marketing, ecommerce, and presentations) has a high likelihood of being AI-generated. The political, cultural, economic, and creative implications will be massive.

Still images are just the beginning. Once video generation APIs arrive, the shift won’t just be disruptive — it will be transformative. ...

UN AI Report

... The Body’s final report, outlines a blueprint for addressing AI-related risks and sharing its transformative potential globally, including by:

• Urging the UN to lay the foundations of the first globally inclusive and distributed architecture for AI governance based on international cooperation.
• Proposing seven recommendations to address gaps in current AI governance arrangements.
• Calling on all governments and stakeholders to work together in governing AI to foster development and protection of all human rights.

...

This is clearly reflected in the four main areas of generative AI of concern to the United Nations:

• Peace and Security: AI-powered disinformation is already endangering UN peace and humanitarian operations, putting staff and civilians at risk. Over 70% of UN peacekeepers responding in a recent survey said mis- and disinformation severely hampered their ability to carry out their work.
• Human rights violations: AI is being used to create and spread harmful content, including child sexual abuse material and nonconsensual pornographic images, especially targeting women and girls. The UN is also deeply concerned that anti-Semitic, Islamophobic, racist and xenophobic content could be supercharged by generative AI.
• Democracy at risk: The potential of AI to manipulate voters and sway public opinion during elections poses a significant threat to democratic processes around the world.
• Undermining science and public institutions: For example, AI tools could escalate decades-long disinformation campaigns to derail climate action by amplifying false information about climate change and renewable energy.

...

See the full story here: https://www.un.org/en/global-issues/artificial-intelligence

“Periodic table of machine learning” could fuel AI discovery

... “Perhaps the most challenging aspect of being a machine-learning researcher these days is the seemingly unlimited number of papers that appear each year. In this context, papers that unify and connect existing algorithms are of great importance, yet they are extremely rare. I-Con provides an excellent example of such a unifying approach and will hopefully inspire others to apply a similar approach to other domains of machine learning,” says Yair Weiss, a professor in the School of Computer Science and Engineering at the Hebrew University of Jerusalem, who was not involved in this research. ...

See the full story here: https://news.mit.edu/2025/machine-learning-periodic-table-could-fuel-ai-discovery-0423

New study introduces a test for artificial superintelligence

...

SuperARC defines intelligence in terms of recursive compression repeatedly condensing information to reveal deeper patterns not apparent to tools such as Large Language Model chatbots (LLMs). The test employs a type of specialised probability, drawing upon the equivalence between compressibility and predictability established in the theory of randomness. The paper proves mathematically the equivalence between compression and prediction and exploits it to show how model abstraction and planning in the context of AI are formally two sides of the same coin.

The authors argue that intelligence is best measured by the ability to produce approximations to short computable hypotheses—one that can not only reconstruct but also predict data by running code in parallel to simulate many future states and pick the one that is closer to the observation at any given time. This perspective moves away from conventional, human-centric IQ-style tests, aiming for a more fundamental and agnostic measure of natural and artificial higher cognitive ability not based on human-centric single answers. ...

The authors of this study propose that future AI progress hinges on integrating symbolic inference with machine learning, arguing that “pure memorisation” approaches fall short of genuine comprehension. A shift to neurosymbolic models may be required to bridge the gap between advanced pattern recognition and true algorithmic inference. ...

See the full story here: https://www.kcl.ac.uk/news/new-study-introduces-a-test-for-artificial-superintelligence

Virtual Companions Are Now Inspiring Art, Fashion, Music, And Even Romance

... [Media artist Refik Anadol] His large-scale installations often use AI to process massive datasets – from architectural archives to images of nature – transforming them into dynamic, mesmerizing visual experiences. ...

This approach, treating data as pigment and algorithms as collaborators, showcases how AI can not only assist but actively inspire new forms of artistic expression, challenging traditional notions of authorship and creativity. ...

The influence of the virtual extends prominently into the worlds of fashion and music. Digital fashion houses are pioneering clothing that exists only in virtual spaces, designed for avatars and online self-expression. ...

Platforms offering AI girlfriend or boyfriend experiences, such as HeraHaven AI, allow users to design customized companions, choosing physical attributes and personality traits to foster specific types of connections, from casual chat to more intimate interactions. ...

See the full story here: https://hauteliving.com/2025/04/virtual-companions-are-now-inspiring-art-fashion-music-and-even-romance/768621/

Undergrads Outsmart Big Tech AI

Two undergraduates from Korea have built "Dia," an open-source speech AI model that competes with tools like ElevenLabs and Google’s NotebookLM. With no prior funding and minimal experience, they used Google’s TPU Research Cloud to train the model. Dia is now publicly available on Hugging Face and GitHub.

The Decode:

• Dia Offers  Rich Voice Control - The 1.6B parameter model can generate podcast-like dialogues with control over tone, speaker tags, disfluencies, and even nonverbal sounds like coughs or laughter. Users can prompt Dia to generate random voices or clone real ones. In early demos, it rivaled commercial tools in quality and flexibility. ...

• Voice Cloning and Lack of Safeguards Raise Flags - Dia enables simple voice cloning, and its open access lacks strong safeguards. While the creators discourage misuse, they also disclaim responsibility. The data used for training hasn’t been disclosed, raising potential copyright concerns. ...

See the full story here: https://decodeai.ghost.io/undergrads-outsmart-big-tech-ai/

How Musicians Are Sabotaging AI With Inaudible Tricks

... Academics have also jumped on board. A team from the University of Tennessee and Lehigh University developed HarmonyCloak, a similar tech that hides distortion beneath the surface of songs. The system introduces subtle sound signals that corrupt data used by AI training models, while preserving the original listening experience for people. ...

Despite the innovation, not everyone’s cheering. Legal experts caution that intentionally corrupting datasets might stir up drama in the courts. Could these poison pills be classed as sabotage? Or are they just digital self-defence? Until clear legal guidance emerges, the tactics remain in a bit of a grey area. ...

See the full story here: https://www.bay939.com.au/trending/entertainment/how-musicians-are-sabotaging-ai-with-inaudible-tricks/