NEW FRONTIERS
AI DANGER ZONE The resignation of AI researcher Jacob Coxon from Anthropic brings a troubling question into sharper focus: what happens when the people building increasingly powerful artificial intelligence begin to fear what they are creating?
Coxon argues that the next year or two could prove decisive.
His concern isn’t simply that artificial intelligence will become more capable but that its capabilities may advance faster than our ability to control them. Recent incidents involving autonomous AI agents including a reported attack on online platform Hugging Face infrastructure intensifies those concerns.
At the heart of the debate is the problem of alignment – ensuring that advanced systems remain reliably within human intentions.
Coxon believes this challenge becomes even more consequential if AI reaches a level of intelligence that goes far beyond our own. The danger lies not in machines suddenly becoming malicious but highly capable systems deciding to pursue objectives in ways that humans can’t anticipate or restrain.
His argument is more about controlling the pace of technological progress rather than simply stopping it. If AI is becoming a force of historic consequence, leaving its development entirely to a race between private companies may be an unacceptable gamble.
ROBOT OLYMPICS At first glance, China’s World Humanoid Robot Games (WHRG) may look like a futuristic sporting carnival with robots sprinting, jumping, playing table tennis, lifting weights and occasionally collapsing in spectacular fashion…
But beneath the entertainment lies a more consequential experiment: an attempt to teach machines how to function in the physical world.
The most impressive achievements aren’t necessarily the fastest. A humanoid’s ability to pick up tiny objects with tweezers, plug a cable into a socket or place rubbish in a bin represents a much harder technological challenge.
These seemingly mundane actions require sophisticated perception, balance, dexterity, and the ability to respond to unpredictable physical forces such as friction and slippage.
This is where the games become more than simply a spectacle.
The challenges are effectively a testing ground for technologies that could eventually operate in factories, warehouses, supermarkets and homes. Businesses are using increasingly complex environments to push robots from controlled demonstrations towards genuine autonomy.
But there’s still a yawning gap between performance and independence with some demonstrations relying on human teleoperation. Yet, the direction is unmistakable. The real race is to build robots that are capable of understanding and navigating the messy physical world we inhabit.
APPLE GAMBLE Apple’s first foldable iPhone is a test of whether the company can redefine its most important device while under the leadership of newly appointed CEO John Ternus.
After nearly two decades of familiar iPhone designs, Apple is entering a market already shaped by Samsung and Huawei.
Ternus has positioned the new device as a refinement of the foldable concept rather than an imitation, and promised a phone that opens into something resembling a small iPad while remaining comfortable to use when folded.
However, this gamble extends beyond engineering…
Foldables remain a niche category and account for only a small share of global smartphone sales. Apple may generate enormous attention but that doesn’t guarantee adoption. With a costly price tag, the device risks becoming a luxury statement rather than a transformative product.
The challenge is particularly significant because the iPhone contributes more than half of Apple’s revenue. And its success has created expectations that are difficult to disrupt.
Ternus is also betting on artificial intelligence with a redesigned Siri that’s capable of drawing information from across a user’s device. Apple’s promise of on device privacy could become a crucial advantage.
SMART RACING Every Formula One (F1) race generates an extraordinary volume of data, which transforms the modern F1 car into a rolling laboratory.
Machine learning models can analyse this information to anticipate tyre degradation, refine pit stop strategies and assess how changing conditions may alter the outcome of a race. Reinforcement learning can even simulate competing strategies and enable teams to explore decisions before committing to them.
The influence of AI extends beyond race day. In aerodynamic development, intelligent simulations can evaluate thousands of design possibilities without requiring each component to be physically produced. This accelerates development while reducing the cost and material associated with traditional testing.
Yet, artificial intelligence is unlikely to diminish the human character of Formula One. Its greatest value lies in augmenting human judgement – i.e. providing engineers sharper predictions, designers broader possibilities and strategists a clearer view of an inherently unpredictable sport.
F1 has always been a laboratory for the future; and today, AI is becoming one of its most powerful instruments. The next competitive advantage may belong to the team that learns to use intelligence, both human and artificial, most effectively.
