Imagine a world where we could have meaningful conversations with our beloved pets. A world in which artificial intelligence technology allows us to understand their meows, purrs and hisses. While such an idea may seem far-fetched, two neuroscientists from Tel Aviv University in Israel have set out to make it a reality. In an essay published in Current Biology, Yossi Yovel and Oded Rehavi presented the “Dr. Dolittle Challenge,” an ambitious challenge that aims to enable artificial intelligence to communicate with animals using their own communication signals. The study was published in the journal Current Biology
Three obstacles
To successfully tackle the Dr. Dolittle challenge, AI-based language models must overcome three major hurdles:
1. using animal signals: AI must learn to use the specific signals that animals use to communicate. This means dogs must respond to human commands without having to memorize new signals to ensure authentic communication.
2. Contextual communication: AI should be able to use animal signals in a variety of behavioral contexts, not just limited to specific situations such as courtship or threat. For example, scientists have managed to create a robotic bee that can attract other bees by repeating their wagging dance, but this method is limited to one specific context.
3. Specific Communication: The ultimate goal is for AI to elicit a measurable response from animals as if they were communicating with another animal of the same species. This requires AI to bridge the gap between artificial technology and the natural world.
Dance of the Bees: Stepping Up
Scientists have made strides in deciphering animal communication, such as the famous honeybee dance. By understanding this dance, scientists have developed a robotic bee that can lead other bees to a specific location. This achievement meets the first and third points of the Dr. Dolittle challenge. However, it does not allow the bee’s desires or emotions to be understood.
Limitations of human language
Even if artificial intelligence successfully meets all the requirements of the “Dr. Dolittle” task, there may still be limitations in our ability to communicate with animals at the desired level. Human language may have unique properties that other species do not. Our limited understanding of the world and egocentric perspective, known as umwelt, may hinder our ability to understand animal communication.
Philosopher Ludwig Wittgenstein argued that even if a lion could speak, we would not be able to understand it. According to Yovel and Rehavi, this means that we would never be able to ask a cat how it feels or explain linguistic nuances, such as translating ChatGPT to CatGPT from French.
Cracking the code: Data collection
Deciphering animal communication requires collecting and analyzing a lot of data. For example, the dance of bees contains complex details that researchers have not yet fully understood, including tactile and acoustic cues about the quality of resources. To crack this code, artificial intelligence algorithms will need access to these additional types of data.
When it comes to primates, communication with them may be more accessible due to their close evolutionary relationship with humans. However, training AI models will require long-term observation of primates in the wild to gather the necessary data. The difficulty lies in obtaining this information and measuring the primates’ natural responses indicating their understanding of AI attempts to communicate with them.
The future of animal communication
While the Dr. Dolittle challenge is a daunting task, it pushes the boundaries of what artificial intelligence can do. As technology advances and our understanding of animal communication improves, we may come closer to bridging the gap between humans and animals. However, it must be recognized that our ability to understand and communicate with the amazing creatures with whom we share our planet will always be limited.
