Robotic builders frequently make it seem to be housebots are imminent, even when they don't seem for many years

The walking, talking and dancing Optimus robots caused quite a stir at Tesla's recent demonstration. But this resulted in disappointment when it turned out that much of the action was actually controlled remotely by humans.

While this is still a fascinating glimpse into the future, it's not the first time robots have seemed a little too good to be true.

Take, for example, Sophia, the robot that Texas-based Hanson Robotics developed in 2016. The company portrayed them as essentially intelligent beings, leading numerous tech specialists to describe this as something beyond our capabilities at the time.

Similarly, we've seen carefully choreographed videos of pre-written action sequences like Boston Dynamics' Atlas gymnastics, the English-made Ameca robot “waking up” and, most recently, Tesla's Optimus in the factory. Of course, these are still impressive in different ways, but they are far from the complete sentient package. Let Optimus or Atlas walk around any home and you'll see something completely different.

A humanoid robot that can work in our homes must be able to perform many different tasks, use our tools, navigate our environment, and communicate with us like a human. If you thought it would only last another year or two, you'll be disappointed.

Building robots that can interact and perform complex tasks in our homes and on the streets is still a major challenge. It is phenomenally difficult to design them well for even one specific task, such as opening a door.

There are so many door handles with different shapes, weights and materials, not to mention the complexity of dealing with unforeseen circumstances like a locked door or objects blocking the way. In fact, developers have now developed a door-opening robot, but robots that can handle hundreds of everyday tasks are still a thing of the future.

Behind the curtain

The Tesla demonstration's “Wizard of Oz” remote control technique is a commonly used control method in the field, providing researchers with a benchmark against which to test their actual progress. This so-called telemetric control has been around for some time and is becoming increasingly sophisticated.

One of the authors of this article, Carl Strathearn, was at a conference in Japan earlier this year where a keynote speaker from one of the leading robotics labs demonstrated an advanced telemetry system. It allowed a single human to simultaneously control many humanoid robots semi-autonomously using predetermined movements, conversational prompts and computer-assisted speech.

This is clearly a very useful technology. Telemetric systems are used to control robots that work in hazardous environments, in disability care, and even in space. The reason a human still takes the helm is that even the most advanced humanoid robots like Atlas are not yet reliable enough to operate completely independently in the real world.

Another big problem is what we can call social AI. Leading generative AI programs such as DeepMind's Gemini and OpenAI's GPT-4 Vision could provide a foundation for creative autonomous AI systems for humanoid robots in the future. But we shouldn't be fooled into thinking that such models mean that a robot is now capable of functioning well in the real world.

Interpreting information and solving problems like a human requires much more than just recognizing words, classifying objects, and generating language. It requires a deeper contextual understanding of people, objects and environments – in other words, common sense.

To find out what is currently possible, we recently completed a research project called Common Sense Enhanced Language and Vision (CiViL). We equipped a robot called Euclid with common sense as part of a generative AI vision and language system to help people prepare recipes. To do this, we had to create common sense databases that relied on real student problem-solving examples.

Euclid could explain complicated steps in recipes, make suggestions when something went wrong, and even point people to places in the kitchen where utensils and tools would normally be found. However, there were still problems, such as what to do if someone had a bad allergic reaction while cooking. The problem is that it is almost impossible to deal with every possible scenario, and yet it requires common sense.

This fundamental aspect of AI has been somewhat lost over the years in humanoid robots. Generated speech, realistic facial expressions, telemetric controls and even the ability to play games like Rock, Paper, Scissors are impressive. But the novelty quickly wears off when the robots are actually unable to do anything useful themselves.

That's not to say there isn't significant progress toward autonomous humanoid robots. Impressive work is being done in the area of ​​the nervous system of robots, for example to give robots more senses for learning. It doesn't usually get the same attention in the press as the big revelations.

The data deficit

Another major challenge is the lack of real-world data to train AI systems, as online data does not always accurately reflect the real-world conditions required to train our robots. We still need to find an effective way to collect this real-world data in large enough quantities to produce good results. However, this could soon change if we can access it through technologies like Alexa and Meta Ray-Bans.

Still, the reality is that we may still be decades away from developing multimodal humanoid robots with advanced social AI capable of helping around the house. Perhaps in the meantime we will be offered robots controlled remotely from a command center. But do we want them?

In the meantime, it is also more important that we focus our efforts on developing robots that can support people who urgently need help now. Examples would be healthcare, where there are long waiting lists and understaffed hospitals; and education, to offer overly anxious or seriously ill children the opportunity to attend classes remotely. We also need more transparency, legislation and publicly available testing so that everyone can distinguish fact from fiction and help build public trust when the robots finally arrive.

Carl Strathearn, Research Associate, Computer Science, Edinburgh Napier University and Dimitra Gkatzia, Associate Professor of Computer Science, Edinburgh Napier University

This article is republished from The Conversation under a Creative Commons license. Read the original article.