YouTube AI error costs creator his channel over alleged link to Japanese account

AI why? Where’s the proof that AI was involved in this?

"A popular tech YouTuber with over 350,000 subscribers has lost his channel after YouTube’s automated systems flagged him for an alleged connection to a completely unrelated Japanese channel that received copyright strikes."

Automated systems existed much much before LLMs were widely deployed.

That's not necessarily AI.

That could be as simple as a database lookup against flagged accounts or a simple heuristic score.

We're over-AI-ing everything.

Everything is AI now. Magic fairy dust.

This reminds me how all of my friends in the last years have replaced special effects/CGI term in movies/ads/shows with AI.

I'm always confused because they've seen special effects for decades, and now the very same explosions, filters, etc, cgi is suddenly AI.

Or under AIing.

It's automated. It's based on information. Why is it not intelligent. Why is it demonstrably less intelligent than an LLM which may make no attempt to retrieve information from other sources but merely has what it was created with.

An algorithm that does a database lookup is more intelligent than an LLM.

Wouldn’t a heuristic score be AI? It could very probably *not* be an LLM or Stable Diffusion or similar which has coopted the overall term “AI”; but that doesn’t make it not an expert system, or an SLM used for categorization, or even A* search, all of which fell under the umbrella of “AI” for a long time.

AI doesn’t imply LLM.

Having been told to implement this sort of thing at least once, it's not even heuristic. It was exactly things like "if firstName == 'Saddam' and lastName == 'Hussein' then lockAccount()". Keep in mind this was years after he was dead, for some reason the big list of prohibited persons the US circulates includes lots of dead people

Because the current hype cycle of "AI" has subsumed the terms "Algorithm" or "Machine Learning" to classify automated decision making processes that rely on some level of modeling / applied statistics instead of deterministic code.

I don't love the way language around this is evolving as it is mostly a marketing tool to make these tools seem much more than they are. Primarily this is driven by the current generative "AI" bubble

The term "the algorithm" has been replaced by "the AI" in modern parlance, and doesn't refer to any specific AI architecture - just something that makes a decision without a human in the loop.

A heuristic can be a single if statement.

Heuristics are just rules of thumb without necessarily having a rigid law or clean classification.

You can derive heuristics from mathematically modeling something or even applying machine learning, but they need not necessarily involve either set of techniques.

I assure you, they still do.

Artificial Intelligence is a thriving and active discipline of the Computer Science field.

It includes things like A* search and expert systems even now, despite current popular parlance shoving LLMs into the spotlight as "AI" and implying that that's all the term means.