How and why, roughly 2 million years ago, early human ancestors evolved large brains and began fashioning relatively advanced stone tools, is one of the great mysteries of evolution. Some researchers argue these changes were brought about by the invention of cooking. They point out that our bite weakened around the same time as our larger brains evolved, and that it takes less energy to absorb nutrients from cooked food. As a result, once they had mastered the art, early chefs could invest less in their digestive systems and thus invest the resulting energy savings in building larger brains capable of complex thought. There is, however, a problem with the cooking theory. Most archaeologists (考古学家）believe the evidence of controlled fire stretches back no more than 790,000 years.

Roger Summons of the Massachusetts Institute of Technology has a solution. Together with his team, he analyzed 1.7 million-year-old sand-stones that formed in an ancient river at Olduvai Gorge in Tanzania. The region is famous for the large number of human fossils (化石) that have been discovered there, alongside an impressive assembly of stone tools. The sand-stones themselves have previously yielded some of the world’s earliest complex hand axes — large tear-drop-shaped stone tools that are associated with Homo erectus (直立人) . Creating an axe by repeatedly knocking thin pieces off a raw stone in order to create two sharp cutting edges requires a significant amount of planning. Their appearance is therefore thought to mark an important moment in intellectual evolution. Trapped inside the Olduvai sand-stones, the researchers found distinctive but unusual biological molecules（分子）that are often interpreted as biomarkers for heat-tolerant bacteria. Some of these live in water between 85°C and 95°C. The molecules’ presence suggests that an ancient river within the Gorge was once fed by one or more hot springs.

Dr. Summons and his colleagues say the hot springs would have provided a convenient “pre-fire” means of cooking food. In New Zealand，the Maori have traditionally cooked food in hot springs, either by lowering it into the boiling water or by digging a hole in the hot earth. Similar methods exist in Japan and Iceland, so it is plausible, if difficult to prove, that early humans might have used hot springs to cook meat and roots. Richard Wrangham, who devised the cooking theory, is fascinated by the idea. Nonetheless, fire would have offered a distinct advantage to humans, once they had mastered the art of controlling it since, unlike a hot spring, it is a transportable resource.

● Stocky, slow-moving whale, rarely grows beyond 15 metres in length

● Flippers are a third of body length; variable dorsal fin size and shape; saw-toothed trailing edge on flukes, often raised when diving

● Bumpy tubercles on top of head

● Body colour is dark brown to black; often extensive white on flippers and underside of body and flukes; such patterns enable individual recognition

● Bushy blow, occasionally V-shaped

● 270-400 olive baleen plates

Humpback whales belong to the rorqual (groove-throated) family, which includes fin, sei, Bryde’s, minke and blue whales. The big family migrate between winter tropical breeding areas (North West Shelf, Great Barrier Reef, New Caledonia, Vanuatu, Fiii, Tonga) and summer Antarctic feeding areas. Once common in New Zealand waters, humpbacks are now rarely seen and may migrate further offshore. Males compete for mates either by physical fight or by song. Females give birth to their young every two to three years; some non-breeding females probably remain in the southern waters during winter. Young humpback whales return to their area of birth but in later life some wander between breeding areas. Humpbacks eat small shrimps and other schooling prey, such as fish, forming small, cooperative groups of two to three individuals to feed.

Similar species: Easily identifiable due to a ‘hump’ back when submerging, but at a distance may be confused with other species that raise their flukes when diving, such as sperm, right and blue whales.

Protection status: Recovering well from past whaling and now numerous in some former migration and aggregation areas, rarely seen in others.

Let us all raise a glass to AlphaGo and the advance of artificial   intelligence. AlphaGo，

DeepMind’s Go-playing AI，just defeated the best Go-playing human，Lee Sedol. But as we drink to its success. we should also begin trying to understand what it means for the future.

The number of possible moves in a game of Go is so huge that. in order to win against a player like Lee. AlphaGo was designed to adopt a human—like style of gameplay by using a relatively recent development--deep learning. Deep learning uses large data sets，“machine learning”algorithms （计算程序） and deep neural networks to teach the AI how to perform a particular set of tasks. Rather than programming complex Go rules and strategies into AlphaGo，DeepMind designers taught AlphaGo to play the game by feeding it data based on typical Go moves. Then，AlphaGo played against itself, tirelessly learning from its own mistakes and improving its gameplay over time. The results speak for themselves.

Deep learning represents a shift in the relationship humans have with their technological creations. It results in AI that displays surprising and unpredictable behaviour. Commenting after his first loss，Lee described being shocked by an unconventional move he claimed no human would ever have made. Demis Hassabis. one of DeepMind's founders，echoed this comment：“We're very pleased that AlphaGo played some quite surprising and beautiful moves. ”

Unpredictability and surprises are—or can be—a good thing. They can indicate that a system is working well，perhaps better than the humans that came before it. Such is the case with AlphaGo. However，unpredictability also indicates a loss of human control. That Hassabis is surprised at his creation's behaviour suggests a lack of control in the design. And though some loss of control might be fine in the context of a game such as Go，it raises urgent questions elsewhere.

How much and what kind of control should we give up to AI machines? How should we design appropriate human control into AI that requires us to give up some of that very control? Is there some AI that we should just not develop if it means any loss of human control? How much of a say should corporations，governments，experts or citizens have in these matters? These important questions, and many others like them，have emerged in response，but remain unanswered. They require human，not human - like，solutions.

So as we drink to the milestone in AI, let's also drink to the understanding that the time to answer deeply human questions about deep learning and AI is now.