If you asked 100 people what artificial intelligence (AI) is, you would probably get 110 answers. Computer scientists, mathematicians or philosophers would give very different answers. No wonder. The term is vague and difficult to define because it already lacks a precise definition of “intelligence”.
Popularly, we usually understand AI as software or algorithms that do something (meaningful) without the instructions for it being hard coded. For example, self-learning software that recognizes tumors on X-ray images, translates texts into foreign languages or autonomously controls vehicles.
Strictly speaking, this is usually machine learning, a sub-field of artificial intelligence in which algorithms learn to recognize patterns based on training data.
We are still a long way from a universal artificial intelligence that can do almost anything. But specific AI applications, or machine learning, have long been part of countless digital everyday products: for example in Google search, in language assistants (Alexa, Siri) or automated recommendations for music (Spotify) and series (Netflix).
Such highly specialized AI applications, each trained for specific problems, have made tremendous strides in recent years and are turning our society upside down. The following examples should clarify this.
When it comes to AI, questionable applications for the military (autonomous war drones), police (algorithms predict crimes), or technology companies (Facebook facial recognition) often make headlines. However, much of AI research revolves around making people’s lives longer, easier and healthier. For example, AI helps diagnose diseases more efficiently, develop medicines faster, personalize treatments or edit genes.
AI will not replace doctors in the foreseeable future, but AI models trained for medical applications support doctors in cancer screening, for example. Algorithms are more reliable than human eyes at finding breast cancer nodules in black and white ultrasound. Seven of the ten largest cantonal and university hospitals already use AI as standard.
The AI learns on its own and gets better with each patient analyzed. The danger here: In the event of an unnoticed incorrect diagnosis, the software learns incorrect information. One misdiagnosis would be followed by more. The use of such adaptive AI in medicine has therefore not yet been approved in Switzerland. Instead, rather rigid, pre-trained AI products are used that neither get smarter nor dumber.
Thanks in part to recent developments in artificial intelligence, it has been possible to develop several vaccines against the coronavirus in record time with an effectiveness that was previously thought impossible. Using an AI model from DeepMind – owned by Google since 2014 – the shape of proteins could be predicted with great accuracy. This is key to quickly studying viruses and developing effective vaccines.
AI supports scientists, governments, but also the economy to limit climate damage by, for example, helping to reduce CO₂ emissions.
A few concrete examples:
Overall, AI can support climate and environmental policy by enabling better predictions and more sensible climate protection measures with increasingly accurate models. AI should also help to use the Earth’s resources more efficiently. This could help us maintain our standard of living without destroying the planet.
Conversely, there is a risk that oil companies will use AI to extract resources more efficiently. In addition: AI applications themselves are currently still very energy-consuming. So there are also critical voices that put the benefits of AI for the environment and climate protection into perspective.
Modern cars are equipped with various driver assistance systems that make driving safer, more comfortable and more environmentally friendly. The computer stays in lane, initiates emergency braking, and uses cameras and sensors to monitor those areas the driver can’t see – apart from that, the AI doesn’t get tired either.
For this purpose, the driving assistants use image and object recognition, ie trained AI models, to recognize the environment and distinguish between, for example, cars, pedestrians and other objects.
Self-learning driver assistance systems and, in the long term, autonomous driving cars should drastically reduce the number of road casualties. AI-powered cars also reduce fuel and energy consumption because they drive more regularly and more predictably than humans.
Since 2017, the astonishingly good online translation service Deepl.com from Germany has been proving that AI is a practical helper in everyday life. Previous online translators produced little more than gibberish. Nowadays, online translators work with neural networks, which means that the correct wording is calculated.
DeepL also uses a trained AI model for this. The problem: The AI can now very well calculate the correct name of the sentence to be translated, but does not understand the language well. So she doesn’t understand the context outside the sentence.
Yet language barriers are currently being broken down at an incredible rate by self-learning, ever-improving translation algorithms. You can also see this on social media, when posts in a foreign language are automatically shown in your own language.
Google’s speech recognition AI has been automatically generating subtitles on YouTube for some time, possibly including translation. In the future we will increasingly see such live subtitles elsewhere – a blessing for the deaf, even if the subtitles generated in real time are often not yet sufficient.
There has been an arms race between credit card providers and online criminals for years. Artificial intelligence helps to identify new fraud trends at an early stage.
For example, the algorithms look for deviations that people normally do not do, such as paying several times in a row at gas stations. For this, the AI is fed with the solved fraud cases of the credit card provider. It aims to discover fraud patterns, identify fraudulent transactions with a higher probability and minimize erroneous card blocking.
However, AI cannot replace humans for the time being. When a customer becomes a victim of fraud, most want to speak to a real employee, not a chatbot.
Modern smartphones learn in everyday life. For example, the AI learns which apps are used often and when, or which display brightness is ideal in which situations. Based on usage behavior, apps not used by the user are closed in the background and processor and display brightness are controlled.
The AI tries to predict the user’s usage behavior, which ultimately saves battery.
A well-trained AI recognizes the subject we want to shoot and magically chooses the right camera setting. But AI often supports us without us noticing. When the smartphone converts the shaky holiday video into a silent, almost perfect film, it does so thanks to AI. And if the smartphone automatically brightens a photo that is much too dark and at the same time optimizes colors and contrast, the AI is also at work.
AI assists IT security teams in detecting cyber-attacks and waves of phishing and malware. For example, email providers use AI models trained on spam and phishing.
The catch: The criminals turn the tables and also use AI to better disguise and professionalize their attacks. For example, with AI you can generate new and more credible phishing emails much faster, which more victims fall for.
AI can therefore be used in the fight against spam, phishing or malware, but can also be misused for exactly such attacks.
AI-based image recognition methods can detect extrasolar planets (exoplanets) that remain invisible with conventional approaches.
By feeding the AI algorithm with countless photos, it learns to recognize for each pixel of a constellation what object it represents – for example, a solar eclipse by a planet. With this method, the two exoplanets Kepler-1705b and Kepler-1705c were discovered.
AI-based image recognition methods can therefore be used in many ways: from facial recognition in mobile phones to traffic recognition for driver assistance systems in cars to detecting exoplanets. However, unlike a human, the AI still needs an extremely large amount of training images to distinguish, for example, a cat from a dog.
Artificial intelligence can make our lives easier, more beautiful and at the same time more difficult or make it a nightmare: AI can already compose a classical string quartet or a self-invented jazz melody. An AI model is fed with information about notes, rhythm and timbres. The AI recognizes patterns and starts creating new works independently.
However, AI can also be used in warfare, for ubiquitous surveillance or for the oppression of minorities. In China, for example, the Uyghur Muslim minority is monitored using intelligent software that can recognize faces and movements.
AI is also at the center of research and development in China’s current five-year plan. And Russian President Vladimir Putin said in 2017: “Whoever has the upper hand in this area will become the ruler of the world.”
AI is also about control, efficiency and profit. This shows:
AI is not good or bad in itself, for now it is what we make of it. From a purely technical point of view, AI is about simulating intelligent behavior using mathematics and computer science. That is why it is also referred to as “weak” artificial intelligence, since it only masters one very specific task:
For example, Tesla’s AI drives statistically safer than people on highways, but the autopilot cannot compose music. Conversely, the music-trained AI model cannot drive a car. Previous AI models can only do one very specific thing, while humans can do countless things.
An all-encompassing or “strong” AI, which like humans shows flexible intelligence for very different problems, will probably remain a dream of the future for decades to come. And maybe that’s for the better: an AI that is not only specifically, but generally superior to humans, could one day be our downfall.
Resolution of the skin spot image: malignant melanoma (black skin cancer) on the left, an innocent birthmark on the right.
Source: Blick
I am Ross William, a passionate and experienced news writer with more than four years of experience in the writing industry. I have been working as an author for 24 Instant News Reporters covering the Trending section. With a keen eye for detail, I am able to find stories that capture people’s interest and help them stay informed.
