‘I feel useless': ChatGPT-5 is so smart, it has spooked Sam Altman, the man who started the AI boom

Economic Times

5 days ago

OpenAI is on the verge of releasing GPT-5, the most powerful model it has ever built. But its CEO, Sam Altman, isn't celebrating just yet. Instead, he's sounding the alarm.In a revealing podcast appearance on This Past Weekend with Theo Von, Altman admitted that testing the model left him shaken. 'It feels very fast,' he said. 'There are moments in the history of science, where you have a group of scientists look at their creation and just say, you know: 'What have we done?''His words weren't about performance metrics. They were about consequences.Altman compared the development of GPT-5 to the Manhattan Project — the World War II effort that led to the first atomic bomb. The message was clear: speed and capability are growing faster than our ability to think through what they actually mean.He continued, 'Maybe it's great, maybe it's bad—but what have we done?'
This wasn't just about AI as a tool. Altman was questioning whether humanity is moving so fast that it can no longer understand — or control — what it builds. 'It feels like there are no adults in the room,' he added, suggesting that regulation is far behind the pace of innovation.Exact specs for GPT-5 are still under wraps, but reports suggest significant leaps over GPT-4: better multi-step reasoning, longer memory, and sharper multimodal capabilities. Altman himself didn't hold back about the previous version, saying, 'GPT-4 is the dumbest model any of you will ever have to use again, by a lot.'For many users, GPT-4 was already advanced. If GPT-5 lives up to the internal hype, it could change how people work, create, and think.In another recent conversation, Altman described a moment where GPT-5 answered a complex question he couldn't solve himself. 'I felt like useless relative to the AI,' he admitted. 'It was really hard, but the AI just did it like that.'OpenAI's long-term goal has always been Artificial General Intelligence (AGI). That's AI capable of understanding and reasoning across almost any task — human-like intelligence.Altman once downplayed its arrival, suggesting it would 'whoosh by with surprisingly little societal impact.' Now, he's sounding far less sure. If GPT-5 is a real step toward AGI, the absence of a global framework to govern it could be dangerous.
AGI remains loosely defined. Some firms treat it as a technical milestone. Others see it as a $100 billion opportunity, as Microsoft's partnership contract with OpenAI implies. Either way, the next model may blur the line between AI that helps and AI that acts. OpenAI isn't just facing ethical dilemmas. It's also under financial pressure.Investors are pushing for the firm to transition into a for-profit entity by the end of the year. Microsoft, which has invested $13.5 billion in OpenAI, reportedly wants more control. There are whispers that OpenAI could declare AGI early in order to exit its agreement with Microsoft — a move that would shift the power balance in the AI sector dramatically.Microsoft insiders have reportedly described their wait-and-watch approach as the 'nuclear option.' In response, OpenAI is said to be prepared to go to court, accusing Microsoft of anti-competitive behaviour. One rumoured trigger could be the release of an AI coding agent so capable it surpasses a human programmer — something GPT-5 might be edging towards.Altman, meanwhile, has tried to lower expectations about rollout glitches. Posting on X, he said, 'We have a ton of stuff to launch over the next couple of months — new models, products, features, and more. Please bear with us through some probable hiccups and capacity crunches.'While researchers and CEOs debate long-term AI impacts, one threat is already here: fraud.
Haywood Talcove, CEO of the Government Group at LexisNexis Risk Solutions, works with over 9,000 public agencies. He says the AI fraud crisis is not approaching — it's already happening.
'Every week, AI-generated fraud is siphoning millions from public benefit systems, disaster relief funds, and unemployment programmes,' he warned. 'Criminal networks are using deepfakes, synthetic identities, and large language models to outpace outdated fraud defences — and they're winning.' During the pandemic, fraudsters exploited weaknesses to steal hundreds of billions in unemployment benefits. That trend has only accelerated. Today's tools are more advanced and automated, capable of filing tens of thousands of fake claims in a day.
Talcove believes the AI arms race between criminals and institutions is widening. 'We may soon recognise a similar principle for AI that I call 'Altman's Law': every 180 days, AI capabilities double.' His call to action is blunt. 'Right now, criminals are using it better than we are. Until that changes, our most vulnerable systems and the people who depend on them will remain exposed.'Not everyone is convinced by Altman's remarks. Some see them as clever marketing. But his past record and unfiltered tone suggest genuine concern.GPT-5 might be OpenAI's most ambitious release yet. It could also be a signpost for the world to stop, look around, and ask itself what kind of intelligence it really wants to build — and how much control it's willing to give up.