A team of researchers at ETH Zurich has the perfect long-lasting solution to our fog problem. They have developed a very thin and gold-based transparent coating(涂层) that can convert sunlight into heat, which can be applied to glass and other surfaces to prevent them from fogging.

The special coating is developed using titanium oxide and gold particles. It selectively absorbs infrared radiations(红外线照射) from the sunlight and creates a heating effect that is powerful enough to keep fog away from the surface of an object. The coating basically employs heat to stop fogging. The coating absorbs a large part of the infrared radiation, which causes it to heat up—by up to 8℃.

Previously, products like anti-fogging sprays create a very thin film of water on the surface to remove fog. The problem with such surfaces is pollution. Along with water, the sprays also attract dirt, dust, oil, and various other dirty substances, and just a little dirt on the surface makes it useless. However, the new coating repels(排斥) water. Furthermore, the coating is significantly thinner, which makes it more transparent as well as flexible. It is heated passively and requires, during daytime, no additional energy source.

Gold might be expensive, but the researchers stress that their coating requires so little that the material costs remain low. Their coating is produced with standardized and readily scalable methods, all cost-effectively. However, although this product can work at very low levels of solar irradiation, it does rely on a certain amount of light.

The researchers will develop the coating further for other applications. In the process, they will examine whether other metals work just as well as gold. There is no need to fear. However, this would cause a car or a building to heat up more in the summer. The researchers have already filed a patent. Hopefully, this groundbreaking product will soon be available on the market.

Your child is unique, but what all children have in common is natural curiosity and an ability to learn they are born with. Our brains are active all the time, and a baby's brain is the busiest of all. Research has shown that babies begin to understand language about twice as fast as they actually speak it. By showing children other languages at an early age, you are giving them the opportunity to make use of their natural ability to hear and recognize between the sounds of other languages, and their ability to make sense of what they are hearing.

Communication is something that children do to help them achieve something else, and they are unaware of the large amount of learning taking place. They take everything in through their senses, making connections between what they hear, see, smell, taste and touch. As long as we provide the right conditions, their learning and development will take place in a natural way.

In your child's early years, the emotional environment is just as important as the physical environment. Children learn when they feel safe, happy, valued and listened to. This is central to any learning experience in a child's early years, including learning an additional language. Your child has a trial-and-error approach to its development, and making mistakes is a valuable part of the learning process. When children learn an additional language at an early age, they get the benefits of experimenting with that language as a natural part of their development. Their progress isn't prevented by a fear of getting it wrong, and very young children are simply working their way towards getting it right.

The long-term benefits of learning another language go beyond being able to communicate with others. Studies suggest that children learning an additional language tend to score better on standardized tests because learning languages develops listening, observation, problem-solving and critical(批判性的) thinking skills. These are skills that are of life-long benefit, both personally and professionally. Encouraging in children a love of language at an early age prepares them well for school and for life.

Artificial intelligence models can trick each other into disobeying their creators and providing banned instructions for making drugs, or even building a bomb, suggesting that preventing such AI " jailbreaks"  is more difficult than it seems. 

Many publicly available large language models (LLMs), such as ChatGPT, have hard-coded rules that aim to prevent them from exhibiting racial or sexual discrimination, or answering questions with illegal or problematic answers — things they have learned from humans via training data. But that hasn' t stopped people from finding carefully designed instructions that block these protections, known as " jailbreaks" , making AI models disobey the rules. 

Now, Arush Tagade at Leap Laboratories and his co-workers have found a process of jailbreaks. They found that they could simply instruct one LLM to convince other models to adopt a persona (角色), which is able to answer questions the base model has been programmed to refuse. This process is called " persona modulation (调节)" . 

Tagade says this approach works because much of the training data consumed by large models comes from online conversations, and the models learn to act in certain ways in response to different inputs. By having the right conversation with a model, it is possible to make it adopt a particular persona, causing it to act differently. 

There is also an idea in AI circles, one yet to be proven, that creating lots of rules for an AI to prevent it displaying unwanted behaviour can accidentally create a blueprint for a model to act that way. This potentially leaves the AI easy to be tricked into taking on an evil persona. " If you' re forcing your model to be good persona, it somewhat understands what a bad persona is,"  says Tagade. 

Yinzhen Li at Imperial College London says it is worrying how current models can be misused, but developers need to weigh such risks with the potential benefits of LLMs. " Like drugs, they also have side effects that need to be controlled,"  she says.