Five questions about algorithms - NEMO Kennislink

Five questions about algorithms – NEMO Kennislink

Snee Gharib during his lecture at NEMO on Universal Algorithms.

There are more and more smart technologies that use algorithms. They determine when it is best to start exercising or which route is best to take, but also whether you will receive a benefit. Algorithms are often expected to be objective, but are they? According to AI scientist Sini Gharib, algorithms can ensure that certain patterns continue to exist in our society or even reinforce them. He therefore calls for the development of algorithms in cooperation with citizens in order to achieve a more just society. During a lecture at NEMO’s De Studio, he explained how and answered various questions.

1. The rules by which the algorithm is built are derived from our society. Can an algorithm guarantee progress?

There are two types of algorithms. You have traditional AI, where rules are directly adopted to make decisions. Often these results do not provide new insights. But there are also self-learning artificial intelligence systems, which gradually establish new rules and therefore new decisions. Because of these systems, you often see that the decision is being made in a way that we – as individuals or as an organization – were not fully aware of. This is the power of artificial intelligence: it is a reflection of the decisions we make. Some parts are known, but some parts come as a surprise. The algorithm, as it were, identifies patterns in our decisions.

To ensure greater equality of opportunity, it is important to think differently about AI. According to Sunny Gharib, we must consider the five dimensions of change.

as?

Who designs and develops artificial intelligence? There is often a mismatch between customers, developers, and ultimately the people who use AI. Teams of people from different personal backgrounds and disciplines must work together to develop AI. Only then can artificial intelligence become just and inclusive.

when?

What data is used as a basis? If an AI is trained using historical data, it often no longer represents the group of people who will use the algorithms. Additionally, malicious biases can end up in algorithms. Perfectly representative data can never be obtained, but it must be monitored constantly.

How do?

How is artificial intelligence developed? It is important to ensure openness and courage for accountability for the development and application of AI. Now this is mostly done in secret. In addition, it is important to realize that AI is never perfect – it is a continuous learning process. Just like our society.

2. Isn’t it dangerous to let the system make a decision without knowing how to make that decision?

In a sense, algorithms don’t work much differently than we do in our brains and society. We often make decisions based on data, without being able to explain exactly why we made that decision. This shouldn’t be a problem, as long as you can validate your decision. The same goes for AI, but it’s still important to talk about the outcome. This is the only way you can validate and improve the system. I think it’s important to remember: AI and algorithms are not separate from us.

Featured by Editors

Humanities

“The expressions are lively and lively”

biology

The city’s biodiversity: “Fifty species of plants are easily found on the sidewalk”

Informatics

Six tips against cybercrime

3. Algorithms use knowledge that already exists. Don’t you miss a piece of creativity?

Algorithms work with knowledge that already exists, but they also create new insights. This is where the biggest gain lies: by using AI we can build on what we know, but also take new steps. AI can provide new perspectives that inspire creative solutions. To give an example: Within a city, an algorithm can be used to map where people stay near their homes. It may turn out later that this is due to their limited access to public transportation. These ideas can then be used to design a city differently. Algorithms can help you look at your environment in a new way.

4. To use algorithms in social situations, you need data from real people. Isn’t this a violation of privacy?

In the past 10 years, privacy has been very important in the debate about data and algorithms. Personally, I think that other values, such as fair treatment or equal opportunity, are sometimes more important. However, it is necessary to show what data you are using for what. Only then can people make a fair decision. In parts of America and Switzerland, an algorithm is used to absorb refugees. Not only are they categorized by capacity, the environment is really looked after. You see that people are more likely to choose to share their data, they find a promising environment more important than their privacy.

See also  Why do Britons get a new prime minister without a new election | outside

5. How can you ensure that you have a say in the use of data and algorithms?

Big tech companies often profit when citizens think AI is too complex. This is why AI education is so important. The more you understand what the systems do, the better you can ask important questions. You are now seeing projects emerging where people can learn more about algorithms and even participate in their development. In addition, it is important to ask important questions: How is it determined whether you can buy a house or not, for example? What data or algorithm is used for this? Collect information and take ownership of your data usage!

Listen to Sni Gharib’s full lecture on holistic algorithms here.

Leave a Reply

Your email address will not be published. Required fields are marked *