Despite not receiving quite the same buzz as OpenAI’s ChatGPT or Google’s Gemini, Meta AI is improving at an astounding rate, says Meta CEO Mark Zuckerberg. Not only that, but in a new post shared to Meta’s website titled Personal Superintelligence, Zuckerberg claims that Meta AI is improving on its own, without the help of any outside influence.
The Meta CEO says that the improvements are slow, for now, but that they are undeniable. While this might be exciting for some people — like Zuckerberg — it’s also a bit of a nightmare for those who have been pushing for better AI safety guidelines, especially when you consider that some believe big tech is downplaying the risks of AI entirely.
Zuckerberg goes on to talk about the future he sees for AI as a “personal superintelligence” that empowers more people, as well as Meta’s goals to put that power in the hands of as many people as possible. It sounds like a relatively noble goal, but all it really does is distract from the bigger concern at hand: AI is somehow improving itself.
Why self-improving AI is scary
There are a number of reasons why self-improving AI should concern us. For starters, the AI bubble has been expanding quickly — more quickly than anyone can truly keep up with it. And with the Trump administration and others pushing for an AI Action Plan that doesn’t regulate AI, many are concerned about just how far AI will advance before there are any safeguards put in place to manage its growth. In fact, we’re already seeing AI implemented across all of our services and devices, from being built into our phones to verifying our ages across Google services.
Then there’s the fact that AI is already learning things it wasn’t taught. Of course, this isn’t the first time we’ve seen self-improving AI, as previously researchers made AI that could learn from other AI. But seeing it on such a large scale does raise some questions. And all of this is especially concerning in a world where big tech already harvests and sells our data to the highest bidder.
How much of our data will AI be given access to as these services expand? How much control will we really have over it? These are questions we will likely never have answers to. And if Meta AI is already showing signs of self-improvement, how long until other models like GPT and Gemini do the same? Based on how slowly we’re responding to worrying AI advancements right now, I highly doubt we’re ready for that.
