The jump from GPT-4 to GPT-5 feels like an acceleration of time. OpenAI chief Sam Altman explains what the new model can do, what it can't yet do, and how to stay sober in a world where software does in minutes what experts took hours to months to do. For the first time, he experiences a model that provides a usable answer to almost any difficult scientific or technical question. You especially notice the difference in coding, from a simple game to complex applications that can be modified live. Whereas GPT-4 mostly impressed with exams and tests, GPT-5 builds faster, more consistently and writes more naturally. At the same time, Altman cautions that the model remains limited in other respects; expectations will continue to rise as people work with it.
When Altman talks about "superintelligence," he means a system that does AI research better than a top team AND can run an organization better than the CEO. That sounded like science fiction years ago but is now, he says, "visible through the fog." The next milestone on that path is AI making a meaningful scientific discovery. The moment, he says, lies in the second half of this decade, although the definition of "meaningful" remains subject to debate.
Suriname needs to implement these new speeds. With government, industry and universities, establish a small but targeted AI action program, a national data agreement (privacy, archiving, access rights) and a pilot regulatory environment so pilots can run safely or a fund for access to computing power and copilots for start-ups, media and SMEs. What can also be done are educational modules that teachers and students have built instead of just typing prompts. Start with pilot projects where the impact is greatest-for example, triage and translation assistance in primary care, origin verification and deepfakes in police and journalists, and predictive maintenance for roads, bridges and energy.
The line between real and made will continue to blur. Altman expects a mix of cryptographic proofs of origin and social habituation; media have long been "a little bit real and a little bit not." More importantly, a basic model will become culturally and personally tailored to the user, rather than a proliferation of totally different models. That also explains why chatbots with memory feel more personal because context increasingly determines output.
In the labor market, entry-level positions will disappear, jobs will change and new occupations will emerge. Young people can benefit from this, a person can use AI to achieve start-up speed and clout that used to require hundreds of employees. More vulnerable is the group around age sixty, who are less willing or able to retrain, there lies a policy challenge to make transitions fairer. Those who learn to steer now will soon determine the route.
In healthcare, the benefits are immediate. Short-term improvements revolve around more accurate health advice with fewer hallucinations. Heading toward 2035, Altman draws a workflow in which AI builds hypotheses, plans experiments, processes results and pushes new treatments toward the clinic step by step. The promise is tangible and so is the downside; the same capacity requires strict safeguards in biosecurity.
The real brake is not in vision, but in infrastructure. The AI wave rests on computing power, data and algorithms plus an often forgotten fourth pillar, the products people actually want to use. Computing power is the biggest bottleneck. Mega data centers require gigawatt levels of power, specialized chips and a supply chain that is not yet up to speed. As long as demand grows faster than capacity, crunches and queues will remain. On the data side, the value of synthetic data is increasing and user-supplied, increasingly difficult tasks are becoming more important; models must also learn to discover what is not yet in data sets. Algorithmically, Altman still sees "orders of magnitude" of efficiency gains ahead.
Every industrial leap creates messes that politics and society must clean up. Altman argues not for panic, but for new frames of mind in the social contract especially around fair access to computing power and preventing geopolitical friction over it. At the same time, he insists on a simple piece of advice: use the tools. Not to escape thinking, but to deepen it. With that, the core of his message becomes clear, man co-evolves with his tools. Technology is accelerating, but the direction remains ours provided we learn, regulate and build with our eyes open.
