We live in an era of contrasts when it comes to artificial intelligence. On the one hand, almost apocalyptic predictions about AI rebelling against humans; on the other, billion-dollar investments from big technology companies and educational initiatives trying to democratize knowledge. In the last 24 hours, what we have seen is precisely this contrast of extremes that characterizes the current moment in the race for AI.<\/p>\n
The news that attracted the most attention was, without a doubt, the prediction that AI could surpass humans and rebel as early as 2027. According to a report from the AI Futures Project<\/a>, a Berkeley nonprofit organization<\/strong>, we are rapidly moving towards the development of so-called General AI \u2013 capable of performing any intellectual task at human level or higher.<\/p>\n The report is led by Daniel Kokotajlo, a former OpenAI researcher who left the company after he believed it was acting recklessly. Together with Eli Lifland, they developed \u201cAI 2027,\u201d a fictional scenario of what a world in which AI systems surpass our intelligence would look like.<\/p>\n I am often asked about these almost apocalyptic predictions. In my view, we need a balance: neither naive techno-optimism that ignores real risks, nor alarmism that paralyzes innovation. The development of General AI is indeed a likely horizon, but between its existence and a supposed \u201crebellion\u201d there is a conceptual and practical chasm.<\/p>\n Having worked with startups and innovation for over two decades, I have learned that emerging technologies are often overestimated in the short term and underestimated in the long term. If history teaches us anything, it is that regulation and development of security protocols tend to follow (albeit with delay) the most disruptive advances.<\/p>\n While some predict dystopian scenarios, Mark Zuckerberg announced an investment of up to US$1.4 billion<\/a> for the expansion of Meta's AI infrastructure<\/strong>. The plan includes the construction of a gigantic data center, with more than 2 gigawatts of capacity \u2013 to give you an idea, approximately the size of Manhattan.<\/p>\n This isn\u2019t just an impressive investment, but a clear sign that the race for AI supremacy is just beginning. Meta plans to activate around 1.3 million specialized GPUs by the end of the year, requiring more than 1 gigawatt of power to operate.<\/p>\n These numbers reveal an inconvenient truth: the computing power needed to train and operate advanced AI models is becoming an increasingly scarce resource, concentrated in the hands of a few companies. The question we should ask is not whether AI will rebel against humans, but what the socioeconomic impact of this concentration of technological power will be.<\/p>\n In Brazil, an initiative that deserves to be highlighted comes from XP, which is offering 10 thousand scholarships<\/a> for training in generative artificial intelligence and cloud automation<\/strong>. In partnership with the DIO platform, the free online course will cover everything from basic concepts to advanced techniques such as prompt engineering and using GitHub Copilot.<\/p>\n This initiative is exactly the kind of action we advocate when we talk about the need to democratize access to technology. In my years working with startups, it has become clear that one of the biggest bottlenecks in Brazil is not the lack of talent, but rather the lack of adequate and up-to-date training.<\/p>\n The bootcamp has 49 hours of content divided into five modules, and registration is open until May 11. What I particularly like about this initiative is its practical focus: live mentoring with experts and applied challenges. It's the kind of training that develops the skills that the market really needs.<\/p>\nMeta Invests up to US$65 Billion in AI Infrastructure<\/h2>\n
XP Invests in Training: 10,000 Scholarships for AI Training<\/h2>\n
The Invisible Revolution of AI at Work<\/h2>\n