【AI产业】大量公司利用AI来寻找矿产

THE FUTURE is electric. That means it will need a lot of batteries, motors and wires. That, in turn, means a lot of cobalt, copper, lithium and nickel with which to build them. Great times, then, for prospec­tors, and particularly for any who think they can increase the efficiency of their profession. Several firms are applying arti­ficial intelligence (AI) to the process, both to improve the odds of surface strikes and to detect underground ore bodies that are invisible to current techniques.

未来是电动的。这意味着将需要大量的电池、电动机和电线。而这反过来又意味着需要大量的钴、铜、锂和镍来制造它们。对于勘探者来说，这是一个好时机，特别是对于那些认为他们可以提高自己的专业效率的人。一些公司正在将人工智能（AI）应用于这一过程，既可以提高地表找矿的成功率，也可以探测对当前技术来说是看不见的地下矿体。

KoBold Metals in Berkeley, California, Earth AI in San Francisco and VerAI in Bos­ton are tiddlers at the moment, as are Sens­Ore, in Melbourne and OreFox, in Brisbane. But at least one bigger fish–Rio Tinto, an Australian-British firm–is also keen. They are garnering reams of geological, geo­chemical and geophysical data to feed to software models. These, they hope, will spot patterns and draw inferences about where to sink new mines.

加利福尼亚州伯克利市的KoBold Metals、旧金山的Earth AI、波士顿的VerAI目前都是小规模企业，墨尔本的SensOre和布里斯班的OreFox也是如此。但至少有一家较大的公司——澳大利亚英国公司力拓（Rio Tinto）也很感兴趣。它们正在收集大量地质、地球化学和地球物理数据，用于供给软件模型。它们希望这些模型能够发现模式并推断出在哪里挖掘新的矿井。

Some of the data are new. But a lot once mouldered in the archives of national geo­logical surveys, journals of geology and other historical repositories–or, in the case of Rio Tinto, which has been operat­ing for 150 years, sat in the form of rock cores in various sheds around the world.

一些数据是新的。但很多曾经存在于国家地质调查档案、地质学杂志和其他历史资料库中，或者在力拓这样的公司，已经运营了150年，以岩心的形式储存在世界各地的各种仓库中。

Kobolds were mythical underground sprites that bedevilled miners in medieval Germany. (They also gave their name to co­balt.) Kurt House, KoBolds boss, hopes some of their magic will rub off. His firm has reformatted archive data from around the world, many of which are on paper and some of which go back to the 19th century, into machine-useable form. That has per­mitted it to build maps of areas of interest all over Earths surface.

Kobolds是中世纪德国的矿工们所害怕的神秘的地下小精灵。（它们也给钴起了名字。）KoBold的老板库尔特·豪斯希望它们的一些魔力会传下来。他的公司已经将来自世界各地的档案数据（其中许多是纸质的，有些可以追溯到19世纪）重新格式化为机器可用的形式。这使得他们能够制作出全球各地感兴趣区域的地图。

Some of those maps are used to train the companys AI models. Others are used to test that softwares effectiveness by checking how good it is at predicting known ore deposits on maps it has not pre­viously seen. If it passes, it can be let loose on under-explored places of interest, gen­erating leads for KoBolds geologists.

其中一些地图用于训练公司的AI模型。其他地图用于通过检查其在先前未见过的地图上对已知矿床的预测效果来测试该软件的有效性。如果通过测试，它就可以释放到未开发的感兴趣的地方，为KoBold的地质学家提供线索。

Earth AI , led by Roman Teslyuk, Sens­Ore, led by Richard Taylor, and OreFox, led by Warwick Anderson, have taken similar approaches, but have concentrated on Aus­tralia, which has particularly rich public geological records. VerAI , led by Yair Fras­tai, focuses on the western bits of North and South America, home to eight of the worlds ten biggest copper mines.

Earth AI由罗曼·特斯里克领导，SensOre由理查德·泰勒领导，OreFox由沃里克·安德森领导，它们采取了类似的方法，但专注于澳大利亚，该国拥有特别丰富的公共地质记录。VerAI由亚伊尔·弗拉斯塔伊领导，专注于北美和南美的西部地区，这里有世界十大铜矿中的八个。

Dr House is especially proud of his Al s ability to predict the shapes and distribu­tions of subterranean plutonic intrusions. These are bodies of igneous rock, often ore-bearing, that have risen as liquid mag­ma from Earths interior but solidified be­fore they reached the surface. They can be detected from the surface via magnetic anomalies which suggest that a particular group of rocks formed at a different time from its surroundings, a standard practice in the industry. But KoBolds AI is able to make more accurate predictions of the shapes of these intrusions, and so suggest the most effective places to drill.

豪斯博士对他的AI能够预测地下岩浆侵入的形状和分布感到特别自豪。这些是火成岩体，通常富含矿物质，是从地球内部升起的液体岩浆，但在达到地表之前凝固。它们可以通过表面的磁异常来检测，这表明一组特定的岩石形成的时间与周围不同，这是该行业的一种标准做法。但是KoBold的AI能够更准确地预测这些侵入体的形状，从而提出最有效的钻探位置。

And with success. Last year, KoBold an­nounced its discovery of a rich deposit of chalcocite, a sulphide of copper, in Zam­bia. Earth AI, meanwhile, has to its credit a big find of molybdenum (an important component of specialist steels) in New South Wales. VerAI has found ore contain­ing copper, gold and silver in Chile and Peru. SensOre has found a large source of lithium in Western Australia. And OreFoxs technology has turned up a potential gold mine in Victoria, plus several promising copper prospects.

而且取得了成功。去年，KoBold宣布在赞比亚发现了丰富的辉铜矿床，辉铜是一种含有铜的硫化物。与此同时，Earth AI在新南威尔士州成功发现了一处大型的钼矿（钼是专用钢的重要组成部分）。VerAI在智利和秘鲁发现了含有铜、金和银的矿石。SensOre在西澳大利亚发现了一个大型的锂资源。而OreFox的技术在维多利亚州发现了一个潜在的金矿，以及几个有希望的铜矿前景。

Rio Tinto is building what Russell Eley, its head of exploration data science, calls a "virtual core shed". This will bring together details of the many rock-core samples the firm has collected over the years. Software will then search these for patterns that will assist the interpretation of new cores, and tell geologists the best places to drill next.

力拓正在建设一所被 Russell Eley 称为“虚拟岩心仓库”的设施。这将汇集该公司多年来收集的许多岩心样本的详细信息。然后软件将搜索这些样本，寻找有助于解释新样本的模式，并告诉地质学家下一步最佳的钻探地点。

Dr House observes that 99% of explora­tion projects fail to turn into actual mines. AI therefore has plenty of room to improve things. It may also help with a more subtle problem. By greatly expanding the volume of rock which can be searched, it will en­able new strikes in familiar, well-governed countries, lessening the need to rely on what Mr Taylor diplomatically calls "exotic jurisdictions" for future supplies.

豪斯博士指出，99%的勘探项目最终无法成为实际矿井。因此，人工智能有很大的发展空间。它还可能帮助解决一个更微妙的问题。通过大幅扩大可搜索的岩石体积，它将使在熟悉且治理良好的国家中发现新矿脉成为可能，减少了依赖Taylor先生委婉称为“异国法域”未来供应的需求。