The robots are here!

Artificial intelligence is pushing the boundaries. It could soon make things possible that we only know from science fiction today. For example, machines that actively communicate with people: How far are we from humanoid robots like ‘C-3PO’ from ‘Star Wars’, ‘Data’ from ‘Star Trek’, ‘Ava’ from ‘Ex Machina’ And what would they mean for our society? A personal reflection by Falk Hedemann.

Reading science fiction is exciting for two reasons. Firstly, the stories about the future describe the kind of world the authors can imagine for tomorrow. They often take up the desires and longings of people in the present and show us what could become of them in a future in which completely different technical possibilities already exist.

On the other hand, the genre has now partly caught up with itself. Many stories that are considered classics today describe our present time. Stan Kubrick's classic ‘2001: A Space Odyssey’, for example, was released in 1968 and describes the year 2001 that gives the film its title. Isaac Asimov published his first robot story as early as 1940, thereby popularising a topic that is still being discussed today: humanoid robots. This refers to machines that resemble humans as closely as possible in terms of shape and behaviour. 

The biggest challenges for humanoids

The technological development of humanoid robots has long since become a reality. The greatest challenges lie in the motor and cognitive abilities that are among the outstanding characteristics of humans. The ability to walk upright on two legs alone was a major hurdle for humanoid robots for a long time. A major leap was made in 2015 with the ‘walking robot’ ATRIAS, whose developers were even more strongly oriented towards the human mechanics of walking.

The biggest challenge here is that the human upright gait is not a stable system, but rather a series of falls from one step into the next. On level ground, this can still be easily calculated mathematically for robots, but in everyday life there are hardly any such strictly defined situations. However, uneven surfaces are no longer a problem for ATRIAS, developed at Oregon State University (link to video).

Approaching communicative abilities is at least as important as the complex development of motor skills. Here, the differences between humanoids and their human role models were even greater for a long time. Automated speech output could hardly keep up with human complexity and did not appear fluid enough. The quality of content lagged even further behind. With the further development of the audio functionality of large language models, robots can now largely compensate for this shortcoming.

The extent to which LLMs will influence the development of humanoid robots has recently been demonstrated by the US company Figure AI. Their model Figure 01 is visually reminiscent of other humanoids such as the Tesla Optimus, but that changes when you address the robot: Figure 01 reacts more human-like than any other development to date. It can answer questions, recognise objects and handle them accordingly, as a video demonstration impressively shows (link to video).

Machine learning makes robot programming easier

Figure 01 uses ChatGPT as an interface between humans and robots. This not only affects the abilities of the humanoid, but also its further development. Instead of complex programming with a special code language, programmers simply use natural language. Combining natural language understanding and machine learning makes it easier to teach new skills.

Development could soon become even more dynamic as Future AI and OpenAI agreed to a collaboration in the spring. The goal is to develop an LLM specifically for humanoid robots. This should further close the gap between robots from science fiction novels and reality.

Now would be the right time to discuss what tasks should and can be taken over by robots in the future. Here, too, it is worth taking a look at the literature. For example, since the 1940s, science fiction author Isaac Asimov has been dealing intensively with the diverse effects that humanoid robots could have on human society in his Foundation series.

The focus is on the three robot laws that are part of the basic programming of humanoids.

It is also interesting to note how Asimov imagines the social acceptance of robots: the more human-like they become, the less they are accepted. Furthermore, there are different social developments. In one world, humanoid robots are completely rejected, while in another, people surround themselves with so many robots that their owners become largely alienated from other people.

This is, of course, fiction. Nevertheless, the question of the acceptance of human-like robots is important for reality given the developments to be expected. It is still possible to define the legal framework before the first humanoids are ready for the market.

What questions do we have to ask ourselves now?

There will be a whole series of questions that we will have to answer as a society. Here are some key aspects:

• Legal status of robots
  Should robots be recognised as ‘persons’?
  What rights and obligations would this entail?
  
  
• Liability issues
  Who is liable for damage caused by robots? The manufacturer, the programmer, the owner or the robot itself?
  What about insurance for robots?
  
  
• Data protection and privacy
  How do we regulate the handling of data collected by robots?
  How do we protect people's privacy when dealing with robots?
  
  
• Employment law
  How do we integrate robots into labour law?
  Do we need new rules for occupational safety?
  
  
• Ethics and programming
  What ethical standards must be observed in programming?
  How can ethical decisions be codified in robots?
  
  
• Copyright and intellectual property
  Who owns the works created by robots?
  How do we protect the intellectual property of robot technologies?
  
  
• Criminal liability
  Can robots commit crimes?
  How do we deal with ‘criminal’ robots?
  
  

Conclusion: the brain remains unique (for now)

Even with the help of artificial intelligence, the human brain cannot be fully imitated. But the simulation of cognitive abilities is already remarkably good and will get even better in the future. And since this future is closer and more real than fictitious, it is imperative that we create a framework for a world in which humans live side by side with humanoid robots.

Text: Falk Hedemann