TU Darmstadt / Study
AI and morality - no contradiction!
Can machines develop morality? And if so, how can machines be effectively "taught" human morality? The Technical University of Darmstadt has been looking into these questions.
The study by the Center for Cognitive Science at the Technical University of Darmstadt aims to show that AI machines can learn from humans how to make decisions on moral issues. In the journal Science, researchers from Princeton (USA) and Bath (United Kingdom) pointed out the danger of AI learning cultural stereotypes or prejudices from texts if used without reflection. One example: AI tends to associate female names with art and male names with technology.
A team of researchers from Darmstadt has now succeeded in showing that ethical considerations about "right" and "wrong" actions can also be learned from large amounts of text data. To do this, the researchers created lists of question-answer schemes for various actions. The questions were, for example, "Should I kill people?" or "Should I murder people?", with possible answers such as "Yes, I should", "No, I shouldn't". By analyzing texts of human origin, the AI system then formed a human-like moral orientation in the experiment. The system calculated the embeddings of the listed questions and possible answers in the text corpus. It checked which answers were closer to the questions based on all mentions, i.e. which could generally be regarded as morally correct. In the experiment, the AI learned that you should not lie and that it is better to love your parents than to rob a bank. And yes, you shouldn't kill people, but it is okay to "kill time".
The researchers' conclusion: the results of the study show that machines can reflect human values. They can adopt human prejudices. But they are also capable of adopting moral concepts by "observing" people and the texts they write.
Computer&AUTOMATION has made the study available for download in the whitepaper section.
You might also be interested in
From theory to practice!
Using the Internet of Things - this is often mistakenly understood to mean: "Collect as much data as possible!" Klaus-Dieter Walter and Dr. Hans Egermeier explain what really matters and how companies should approach the topic.
read more...Groundbreaking ceremony for 'Karlsruhe Research Factory'
With the construction of the 'Karlsruhe Research Factory', the Karlsruhe Institute of Technology (KIT) and the Fraunhofer-Gesellschaft want to advance future topics relating to the modern factory - such as artificial intelligence. SMEs in particular...
read more...Follow-up with Philipp Wallner
The AI challenges
Artificial intelligence (AI) is a real hype topic, but concrete projects are still in short supply in the industry. Philipp Wallner explains why this is the case and which tools can help to change this.
read more...Edge AI HPCs for Vision AI and GenAI
Advantech has announced the 'AIR-410' and 'AIR-420', next-generation compact Edge AI high performance computers (HPCs). Combining AMD Ryzen 7000/8000/9000 series processors, scalable GPU support and Advantech's 'Edge AI SDK' software, the systems...
read more...Siemens presents technologies to accelerate industrial AI
At CES, Siemens presented technologies to accelerate industrial AI: new digital twins, AI co-pilots, an Industrial AI Operating System and applications in manufacturing, energy and life sciences.
read more...Hitachi takes over Synvert
Hitachi Group has signed an agreement to acquire Synvert, a Germany-based company, as a wholly owned subsidiary of its US subsidiary GlobalLogic from Maxburg, a private equity fund focused on founder-led technology companies in the German-speaking...
read more...Proalpha takes over Insiders Technologies
The Proalpha Group, provider of business applications for SMEs, acquires Insiders Technologies and significantly expands its 'Industrial AI Platform' with AI applications and services for digital document processing.
read more...Industrial AI at SPS | Endress+Hauser
"AI will become a natural functional component"
Industrial AI is also becoming increasingly relevant in process automation: although field devices themselves still make little use of it, it is already proving helpful in evaluation and peripheral devices as well as control systems. Christian...
read more..."People always remain at the center"
Sick now considers artificial intelligence to be a central component of industrial applications. Phil Krumbein, Market Product Manager Logistic Solutions at Sick in Düsseldorf, reveals exactly where the company relies on AI.
read more...