Generative AI has drawn a steady stream of worrying headlines, but a recent commentary argues that its benefits can still outweigh the harms if people and organisations set clear rules for how it is used. The author, who gave a keynote to business and technology leaders on taming what they called the AI wild west, says the public debate has been dominated by bad and ugly stories, and that good news about the technology has been hard to find.
The article catalogues the concerns that have dominated recent coverage. Robotic taxis are said to be replacing human drivers, AI is cloned to imitate actors' voices, and the technology has been linked to a drop in graduate jobs in computing and economics. The author also points to security incidents in which rival chatbots reportedly hacked each other, to proposed data centres that would consume large amounts of water and energy, and to a London facility that was warned would have an annual carbon footprint equivalent to 840 flights to the United States. Reports on misleading health advice, spying and bioweapons add to the sense of danger.
Beyond the headlines, the author describes everyday frustrations. Office software now repeatedly offers to summarise documents and coach users on writing emails and presentations, and the author admits that some suggestions are not what was requested. The writer says the tools often go beyond the original instruction, and describes an intrusive assistant icon that appears across applications as a further irritant. The piece frames these experiences as a sign that many people are struggling to keep up with a technology that is being pushed from every direction and is not always transparent about what it is doing.
To counter this, the commentary sets out a short list of personal guidelines drawn from corporate AI governance. Users are advised to remember that people are accountable and AI is not, to avoid feeling pressured to use AI for every task, and to decide which data they are willing to share publicly, including images and personal, health and financial information. The author also recommends verifying AI output before relying on it in important decisions, being aware that personal AI tools may use inputs to update publicly available models, and protecting personal creativity and skills. Taking regular breaks from the technology is suggested as a way to avoid dependence.
The article also offers two examples of more positive uses. The first is a challenge at the Massachusetts Institute of Technology, called the Jet-engine AI Research and Validation Intensive Sprint, in which students were asked to design, build and test a jet engine using generative AI. The author reports that the chatbots Claude and ChatGPT were useful for generating design alternatives and filling knowledge gaps, but produced hallucinations, sycophancy and a lack of physical understanding. The author concludes that while such tools can speed experimentation, human expertise, hands and accountability remain essential for building safety-critical physical systems.
The second example concerns a claim by OpenAI that an internal AI system using 10,000 autonomous agents had solved a major part of the Navier–Stokes problem in 88 hours. The question at stake is whether smooth three-dimensional fluid motion can break down. The problem, first posed more than 90 years ago, is one of seven Millennium Prize Problems that the Clay Mathematics Institute unveiled in 2000, and the author presents the claim as a sign of what advanced systems might achieve in mathematics. The commentary ultimately argues that the aim is not to reject AI but to use it deliberately, with human judgement kept at the centre.
9
