Climate change has emerged as the biggest threat to humanity, with devastating consequences such as extreme weather events, climate migration, and a sharp decline in biodiversity. While the brunt of climate action is shouldered by green parties and public activists like the young Greta Thunberg, in recent years many industries have stepped up innovation to try and do their bit.

The informatics industry in particular has been flexing its R&D muscle to propose bleeding-edge solutions. A recent paper published by a group of high-profile AI experts and IT professionals explores the potential that can be found “at the nexus of climate change and machine learning”.

Headed by David Rolnick, a Postdoctoral Research Fellow at the University of Pennsylvania, the paper puts a spotlight on “high-impact opportunities for real-world change” present in such ML fields as artificial intelligence, computer vision, unsupervised learning, and more.

Elsewhere in the industry, a number of digital giants and startups are already implementing their own twists on machine learning-driven solutions for climate change prevention and mitigation.

Let’s examine the three main benefits that ML and its derivatives can offer to the organizations fighting for the future of our planet.

Better predictions and estimates

AI and deep learning can be a good fit for data-rich climate science projects that typically provide datasets of necessary depth and variety, such as the research led by Google, the Mila institute, or the German Aerospace Center. The resulting models can, for example, evaluate energy consumption levels in buildings and electric vehicles more accurately, making it easier to discover trends and predict future consumption. Down the line, this leads to effective energy distribution and battery management — pivotal tasks for those trying to minimize their carbon footprint and make a switch to renewables.

Other examples of deep data analysis in climate change-related research include Facebook’s AI-based population density map for the majority of Africa and the application of DCNNs for detecting extreme weather events like tropical cyclones, atmospheric rivers and weather fronts.

Improved tracking and monitoring

Computer vision technologies such as object recognition are increasingly used to supercharge the analysis of visual climate data. Potential applications range from monitoring the scale of deforestation to generating the size and location data for solar panels to gathering information about climate adaptation policies across communities and nations.

With climate change affecting wildlife all around the planet, conservationists need reliable solutions for tracking populations of endangered species without disturbing them. Appsilon’s animal recognition project is addressing this challenge with an AI model capable of recognizing specific kinds of animals in images. The model is trained using the transfer learning technique, which ensures accuracy even with smaller datasets.

Optimized legacy systems and workflows

Artificial intelligence solutions have proven their effectiveness in streamlining complex business workflows and enhancing legacy systems. In the context of “clean tech”, AI is poised to have an impact on the design of environmentally-friendly buildings, sustainable supply chains, the efficiency of freight transportation, and more.

One of the most important sectors for this kind of innovation is agriculture. AI development enables so-called precision agriculture — solutions that enable waste and carbon emission reduction while improving crop yield.

So is AI going to save us all?

The challenge of the climate crisis can also be viewed as an opportunity for forward-thinking tech businesses to make their contribution. Nevertheless, none of the solutions described above is a silver bullet. It will take coordinated action and commitment from all of us, not just the best brains of the IT sector, to tackle the problem.

At the same time, machine learning and AI have demonstrated enormous potential in reshaping our world — the ways we live, work, and consume. In the coming years, these powerful tools will surely form a crucial part in the global effort against catastrophic climate change. 

The post How Machine Learning and AI Can Help in the Fight Against Climate Change appeared first on Artificial Intelligence.

Elon Musk recently warned that an international artificial intelligence race is more likely to cause World War III than a 20th Century-style arms race. We lawyers have our own “end-of-the-world” concern: that AI will ultimately end most lawyers’ careers. But before we start speculating about what havoc HAL 9000, J.D. could wreak on the legal profession, we should keep in mind a key distinction between two different forms of artificial intelligence: what is known as “weak AI” and “artificial superintelligence.”

The latter is the flavor of AI you are most likely to encounter in a sci-fi movie; that is, an artificially intelligent machine that has the capacity to learn and master a diversity of topics, much like a human can, but at a rate that far surpasses human ability. You may be picturing a robot that can do everything a lawyer does, in a fraction of the time, and at a fraction of the cost — every lawyer and managing partner’s worst nightmare. And indeed, the application of artificial superintelligence to the profession would logically end with robots taking over, and it would make lawyers replaceable.

The reality is that we are many years away from that happening, especially in the legal industry. The AI applications we see today, from driverless cars to image recognition, are “weak” AI. This is the same sort of technology on display in Google’s AlphaGo, an AI program that beat the world champion of the game Go. “Weak” AI systems can process large sets of information, but are only able to do a single task really well (meaning that your self-driving car won’t teach itself how to beat you at Go).

Realistically, only services powered by “weak” AI, rather than artificial superintelligence, will be deployed in the legal profession in our lifetimes. It will be several more decades before AI can “learn” how to synthesize diverse skillsets. As attorneys, we deeply analyze complex subjective issues, we make strategic decisions, and we hone our power of persuasion. That’s why our clients turn to us. Teaching a machine any one of these skills will be a huge challenge; teaching a single machine to apply all of these skills simultaneously is immeasurably more difficult.

This doesn’t mean that today’s “weak” AI isn’t actually extremely powerful. Even though current AI technologies will not take the place of a litigator’s strategic judgment, they’ve already fueled game-changing innovations in the legal market. Like self-driving cars, AI-backed legal technology has the power to automate certain key tasks which we previously thought had to be done by humans, giving users more time to focus on the tasks that do require complex analysis and creative thinking.

Take Kira Systems, a contract analysis platform. Through natural language processing and machine learning, Kira’s AI can process thousands of examples of contract language relevant to the user, providing valuable insights into those contracts. Casetext’s CARA, modern legal research technology, can analyze a document (like an opponent’s brief), and identify the cases and previously filed legal briefs most relevant to the issues in the brief. Both forms of AI save lawyers time on the more mechanical tasks of legal practice, which ultimately leads to more satisfied clients, less time spent on non-billable work, and, very often, more business.

Legal practice often boils down to pattern recognition, and even “weak” AI can uncover patterns that humans aren’t capable of recognizing. For example, the old legal research model presumed that you needed a massive human effort to summarize cases to provide the information litigators need. But new technology applications can find the patterns already present deep in the law. For instance, it turns out that when judges summarize another case in an opinion, they often repeat similar language. Casetext uses those patterns to uncover judge-written summaries of judicial opinions, which most attorneys trust more than any other source. Legal professionals can look forward to even wider-reaching innovations based on AI systems’ capacity for machine learning and natural language processing that will help attorneys do their best work.

The truth is, though, that even though AI can uncover insights about legal writing, there are decades of work to be done before AI can craft arguments at the depth of a human. It’s going to be a while before robots are arguing cases at the Supreme Court. Until then, we can dedicate the time saved by using AI tools to thinking about how to avoid WWIII.

The post Artificial Confusion: The (Overblown) Threat Of Artificial Intelligence appeared first on Artificial Intelligence.