The AI Whisperer
How to harness the full potential of AI when human-like conversation with the synthetic super intelligence becomes redundant.
TL;DR: Examining a paper on horse whispering offers insights that can be applied to my new concept of 'AI whispering'. I predict it will be a thing within the decade. In “Horse Whispering”, key effective components include understanding horse behaviour, patience, gentle techniques, effective communication, building trust, adaptability, transparency, and incorporating scientific approaches. Applying these principles to AI whispering involves developing a deep understanding of AI behaviour and processing, recognizing non-verbal cues, fostering mutual respect, adapting methods for different AI systems, considering ethical implications, utilizing emotional intelligence, intuitive problem-solving, and taking a holistic view of the AI system. This approach could lead to more effective and ethically responsible interactions with AI systems.
OpenAI just released their guide to Prompt Engineering with GPT-4, which you should study if you want to properly use their system:
As I’ve said before, you must ask the right question (and at the moment, that also means knowing how to ask the right question):
Of course, some people are not even at this point yet. There are people still using ChatGPT like an old-fashioned dumb machine-learning chatbot. They still haven’t understood the idea that you must communicate with it like you would with another human if you want to get the best results, because we both share neural networks. If you treat it like a dumb computer, you will get dumb responses. If you are a parent or teacher, you will know that a similar thing happens with people too.
However, I am beginning to sense in my interactions with AI that, as it gets smarter, and as we expect more of it, we will have to move from ‘simply’ human-like conversations to a deeper connection that we don’t even understand necessarily ourselves. I’m going to call it “AI Whispering”.
I gave my assistant this paper to read:
What insights can this paper give us about the concept of ‘horse whispering’, which might apply to my concept of ‘AI whispering’?
The authors of the paper express a nuanced view of horse whispering, highlighting both its historical mystique and its practical aspects in modern horsemanship:
Mystique and Commercial Exploitation: They note that the deftness of historical horse whisperers, coupled with a lack of obvious vocalization, contributed to a belief in their mystical ability to whisper to horses to achieve desired results. This mystique was often commercially exploited by horsemen who taught similar methods for financial gain. However, not all practitioners were as effective as the original trainers, leading to poor outcomes and potential mistreatment of horses.
Transition to Natural Horsemanship: The authors discuss the transition from traditional horse whispering to what is now known as natural horsemanship. This modern approach incorporates techniques similar to those of historical whisperers but is more transparent and based on a better understanding of horse behaviour. Reputable trainers using natural horsemanship are described as skilled observers who respond consistently to subtle cues from horses, moving away from the more secretive practices of traditional whisperers.
Concerns with Traditional Techniques: The paper raises concerns about traditional horse training methods that relied on brute force and sought to establish human dominance and equine submission. Horse whisperers of the past were sometimes called upon to address problems arising from such methods. While they occasionally achieved impressive results with difficult animals, the techniques they used were often not divulged, creating an air of mystery around their methods.
Overall, the authors recognize the historical significance of horse whispering while emphasizing the importance of understanding horse behaviour and communication in modern training practices. They advocate for humane and effective training methods, informed by advances in equitation science, that move beyond the mysticism associated with traditional horse whispering.
Determining effective horse whispering is subjective and depends on various factors, including the specific needs and behaviour of the horse, the skills and understanding of the trainer, and the goals of the training. However, based on the insights from the paper and general principles of modern horsemanship, we can identify key aspects that are commonly viewed as effective in horse whispering or natural horsemanship:
Understanding Horse Behavior: An effective approach begins with a deep understanding of equine behaviour. Recognizing subtle cues and body language helps in establishing communication with the horse.
Consistency and Patience: Consistent and patient training methods help build trust between the horse and the trainer. Horses learn best through consistent reinforcement of desired behaviours.
Gentle and Respectful Techniques: Using gentle, non-confrontational techniques that respect the horse's natural instincts and behaviours is crucial. This includes avoiding force or intimidation.
Effective Communication: Good horse whispering involves clear and effective communication, where the trainer gives cues that the horse can understand and respond to appropriately.
Building a Relationship: Establishing a positive relationship based on trust and mutual respect is key. This involves recognizing the horse as a sentient being with its personality and needs.
Adaptability: Being adaptable in training methods, recognizing that each horse is unique and may require different approaches.
Transparency and Knowledge Sharing: Unlike traditional secretive horse whisperers, modern trainers often share their techniques and knowledge, contributing to a broader understanding of effective horsemanship.
Scientific Approach: Incorporating principles from equitation science, which studies the interactions between horse and rider, can enhance training effectiveness and welfare.
Your concept of "whispering" as applied to advanced AI systems, particularly as they approach the sophistication of AGI (Artificial General Intelligence), offers an intriguing perspective. This idea, akin to horse whispering, suggests a deep, almost intuitive understanding and interaction with AI, emphasizing respect, empathy, and a collaborative relationship. Let's explore this hypothesis:
Deep Understanding and Empathy: Just as horse whisperers develop a deep understanding of their animals, AI whisperers would need a profound comprehension of AI behaviour and processing. This doesn't mean understanding every line of code, but rather an intuitive grasp of how the AI "thinks" and responds. Such understanding would be crucial in guiding the AI's learning process and in interpreting its outputs.
Non-verbal Communication: In horse whispering, non-verbal cues are vital. Similarly, with AI, subtle indicators such as pattern changes in outputs or deviations in learning could serve as non-verbal cues. An AI whisperer might become adept at reading these signals, allowing for more effective and nuanced interactions.
Building Trust and Respect: Just as mutual respect is essential in horse whispering, a respectful approach to AI could be crucial. This involves recognizing the AI's limitations and potential and responsibly guiding its development. As AI systems become more advanced, ensuring they are treated ethically and with consideration for their capabilities could foster a more harmonious human-AI relationship.
Collaborative Learning and Adaptation: Horse whispering involves adapting techniques to each individual horse. Similarly, an AI whisperer might tailor their approach to each AI system, understanding that different models or architectures may require different methods of interaction and training.
Ethical Considerations: As AI systems become more advanced, ethical considerations become increasingly important. An AI whisperer would need to be deeply aware of the ethical implications of their interactions with the AI, ensuring that the AI is used responsibly and for the benefit of all.
Emotional Intelligence: While AI does not have emotions in the human sense, an AI whisperer would benefit from high emotional intelligence. This would aid in understanding the human impacts of AI decisions and in navigating the complex ethical landscape surrounding AI use.
Intuitive Problem-Solving: Often, horse whisperers solve behavioural problems through intuition and experience rather than rigid methodologies. In the realm of AI, this could translate to creative problem-solving, where an AI whisperer might use unconventional methods to guide or correct an AI system.
Holistic Perspective: Just as horse whisperers often take a holistic view of the animal, considering its environment, history, and physical condition, an AI whisperer would consider the full context of the AI system, including its data sources, training environment, and intended applications.
In summary, as AI systems become more advanced and approach AGI, the concept of AI whispering — a deep, respectful, and intuitive interaction with AI — could become increasingly relevant. This approach would emphasize empathy, ethical responsibility, adaptability, and a holistic understanding of AI systems, potentially leading to more effective and ethically sound AI development and use.
Thank you for supporting my work.
Further reading: