Recently, I came across some fascinating developments about how YouTube plans to keep teens safe on its platform—particularly here in the US. Following age-check crackdowns in the UK and Australia, YouTube announced it’s deploying artificial intelligence to estimate users’ ages in an effort to show age-appropriate content. Given YouTube’s massive reach and the ongoing debate about kids’ safety online, this struck me as a major shift in how tech giants are handling age verification.

Why age checks are trending globally (and why YouTube is finally onboard)

Just days before the US rollout announcement, Australia banned kids under 16 from using YouTube and other social networks — a huge move considering how integral these platforms are to young people’s daily lives. Meanwhile, the UK implemented sweeping age verification rules around the same time via the Online Safety Act, targeting everything from porn access to harmful content.

It turns out YouTube’s new AI feature is partly a response to these tightening regulations internationally. While the company historically opposed mandatory age checks, it now seems to be reluctantly complying by using AI to infer age rather than relying solely on user-submitted info.

According to James Beser, YouTube’s director of product management for youth, the AI will estimate age by analyzing behavior patterns like video searches, watch categories, and account longevity. This machine learning approach aims to better distinguish teens from adults, allowing YouTube to activate protective measures for younger users—like disabling personalized ads and enabling stricter content filters.

How does YouTube’s AI age estimation actually work?

The technology is pretty intriguing. Instead of just trusting the birthdate users enter (which many kids might fudge), the AI looks at subtle digital footprints and signals. For example, it pays attention to what types of videos users watch or search for, how long their accounts have been active, and other behavioral cues. This layered approach is what makes it more robust – but it’s not flawless.

Here’s the catch: if YouTube’s AI guesses your age incorrectly, the platform will ask users to verify their age through more traditional means—like submitting a credit card, a government ID, or even a selfie. This fallback gives a manual verification route, but also raises questions about privacy and data security for users, especially minors.

What this means for parents and teens in the US

In the US, where regulations vary widely state by state, and platforms aren’t uniformly forced to impose age checks, YouTube’s adoption of AI age verification is a landmark step. While some states have laws targeting social media age verification, YouTube’s move signals a more standardized approach that could influence other platforms.

For teens, this could mean safer feed curation and less exposure to inappropriate content or targeted ads. For parents, it’s a double-edged sword—while the AI might improve protections, it relies on collecting and analyzing behavioral data, which may feel invasive or raise privacy concerns.

It’s also worth noting that when this AI was tested in Australia recently, it wasn’t guaranteed to be effective, leaving open questions about accuracy and enforcement. Still, YouTube’s decision to implement these measures voluntarily here might be a sign of how seriously tech companies are taking youth online safety amid political and regulatory pressures worldwide.

Key takeaways

• YouTube is using AI to estimate user ages by analyzing behavior, not just declared birthdates, aiming for better protection of teens.
• If the AI estimate is off, users can verify age with credit card, government ID, or selfie, which opens new questions about privacy and security.
• This rollout in the US follows similar legal moves in the UK and Australia, reflecting a global push for stricter youth online safety.

Wrapping it up

Seeing YouTube embrace AI for age verification feels like a meaningful step in tackling the tricky balance between online safety and user privacy. It’s clear the platform is responding not just to regulators but to a cultural push for safer digital spaces for younger users. However, the technology is not foolproof, and the introduction of personal data for verification will reignite debates about privacy.

For parents, educators, and even teens, this development signals that digital platforms are evolving rapidly — and so must our conversations about responsible, transparent tech use. I’ll certainly be watching how this AI age estimation performs live and what feedback emerges from real users in the US.

The post YouTube’s AI age checks in the US: What it means for teens and parents appeared first on Aiholics: Your Source for AI News and Trends.

Thinking about AGI by 2030 always sparks some fascinating questions. How will we know when we’ve truly reached it? What will that breakthrough moment actually look like? I recently came across some insights that paint a vivid picture of what these milestones might be – far beyond just more powerful computation or incremental upgrades.

Defining AGI: It’s about consistency across all cognitive domains

The first thing to tackle is defining what AGI actually means, and it turns out that’s more complicated than it seems. The bar isn’t just about excelling at one task or dominating a niche like today’s systems. Instead, true AGI is about matching the brain‘s broad cognitive capabilities consistently. Think about it: our brains didn’t just invent civilization by being great at chess or language alone—they operate as highly general “thinking machines.”

Current AI, however brilliant in specific areas, often shows glaring inconsistencies. It’s like a patchwork of sharp spots and blind spots, excelling spectacularly at some tasks but failing at others—something the experts call “jagged intelligence.” For example, a system might generate near-perfect chess moves but struggle with creative scientific insight or long-term reasoning.

Testing for AGI might involve a massive battery of tens of thousands of cognitive tasks that humans can tackle. Beyond that, imagine a panel of hundreds of the world’s top specialists — terrors in their respective fields — trying for months to find any glaring holes or weaknesses. If none are found, then maybe we’re there. But the real magic might be those rare, lighthouse moments – the “move 37” of AGI.

The elusive “move 37” and other landmark breakthroughs

The “move 37” reference comes from a stunning moment in the game of Go where AI surprised everyone with a deeply creative, non-intuitive play. What would a move 37 look like in AGI? One idea is inventing a new scientific conjecture or hypothesis, something revolutionary like Einstein did with relativity.

Imagine training an AGI only on scientific knowledge up to 1900, then seeing whether it could independently come up with special and general relativity. That kind of breakthrough would be an unmistakable sign of true general intelligence — creative, theoretical, and deep. Another marker could be inventing a brand-new game with richness and elegance comparable to Go, showing not just mastery but true innovation.

These leaps count for more than just checking boxes on cognitive tests. They demonstrate an AI that can invent brand new knowledge or culture, not just remix existing patterns. It’s about the ability to surprise even the best human experts, producing insights or moves they might initially dismiss but later come to fully appreciate.

Incremental upgrades vs breakthrough leaps: the path to AGI

We often talk about AI progress as a race of scaling up compute or training on more data. But the path to AGI seems to require a hybrid approach: both many incremental improvements and a few game-changing breakthroughs.

Systems like AlphaEvolve already showcase the power of recursive self-improvement — fine-tuning code or enhancing performance through many small steps. But whether this kind of steady hill-climbing alone can get us to AGI is dubious. We might need at least one or two major paradigm shifts, the AI equivalents of transformers or the transformer architecture revolution of 2017.

Scaling compute and data remain crucial. Interestingly, there’s still a lot of room to grow in pre-training, post-training, and inference compute, especially as billions of users worldwide demand responsive, intelligent AI. Yet, the biggest leap might come from the research bench – the creative minds who can crack new scientific or conceptual codes.

And on the data front, running out of high-quality human-like data might not be the bottleneck. Synthetic data generation and simulation offer promising ways to keep feeding AI systems the right information, sustaining progress without hitting a wall.

Practical insights for AIholics and the future

What can we take away from this perspective?

• AGI is about consistent, general cognition, not narrow prowess. True intelligence won’t just ace chess or coding, but operate robustly across domains without glaring blind spots.
• Breakthrough moments matter as much as scaling. Expect landmark achievements — like a novel scientific theory or a brand-new complex game — that showcase real creativity and insight.
• Scaling compute and data remain important, but innovation drives the hardest challenges. AI progress depends equally on deep research and system engineering, so organizations with strong research teams remain key players.

So where does this leave us in 2024? There’s a roughly 50% chance AGI could arrive by 2030, according to recent expert insights. But even when it happens, it may look more like a tapestry of steady improvements punctuated by brilliant, eye-opening leaps. We should watch closely not only for raw performance but for those breathtaking moments of original creativity that redefine what’s possible.

And as these systems evolve, so will the dynamic between human experts and AI — sometimes challenging our assumptions, sometimes elevating us to new heights of understanding. It’s an exciting journey that’s just beginning, with countless surprises ahead.

The post What AGI by 2030 could really look like: consistency, creativity, and the move 37 moment appeared first on Aiholics: Your Source for AI News and Trends.

Apple is likely to delay the release of its new AI features in Europe this year due to regulatory concerns. The company announced that it does “not believe” it will roll out Apple Intelligence, iPhone Mirroring, and SharePlay Screen Sharing to EU users in 2023, citing the uncertainties brought about by the Digital Markets Act (DMA).

The DMA, which regulates large digital platforms to ensure fairness and competition in the EU market, imposes several “do’s” and “don’ts” on “gatekeepers,” or large platforms offering digital services. One key provision is that these platforms cannot use collected data from third parties to compete with them.

Apple expressed concerns that complying with the DMA could impact user privacy and security. “Specifically, we are concerned that the interoperability requirements of the DMA could force us to compromise the integrity of our products in ways that risk user privacy and data security,” an Apple spokesperson said in a statement shared with Quartz. Apple emphasized its commitment to collaborating with the European Commission to find a solution that would allow the company to deliver these features without compromising user safety.

In China, Apple is seeking a partner to help roll out its AI features in its second-largest iPhone market. Apple is integrating OpenAI’s ChatGPT into its newest operating systems, but ChatGPT and other foreign AI models are not available in China. Apple has reportedly discussed deals with Chinese AI developers, including Baidu, Alibaba, and Beijing-based startup Baichuan AI.

The Cyberspace Administration of China requires AI models to undergo a “security assessment” and ensure that content generated by chatbots reflects core socialist values and avoids subversion of state power. As of March, the CAC had approved 117 generative AI models, all developed within China, according to the Wall Street Journal.

The post Apple postpones AI features in Europe amid regulatory hurdles appeared first on Aiholics: Your Source for AI News and Trends.