In the wake of leaders from prominent AI labs like OpenAI and Google DeepMind sounding an alarm on the potential risks of AI, the hype around AI communication has never been higher. The Atlantic even called communicating with AI The Most Important Job Skill of This Century So, what's all the fuss about?
If you're part of the 80% who haven't who haven't yet dipped their toes into the AI pool by using generative AI tools like Chat GPT, Midjourney, or Google Bard, fear not. I'm here to demystify the science of human-to-AI communication.
How Humans Communicate With Each Other
Humans have a complex communication process involving various parts of the brain. It's not just about the words we say, but also our tone, body language, and facial expressions. However, in a chat with AI, there's less emotional involvement. The interaction focuses more on problem-solving and task completion.
Similarly, social cues and cultural norms heavily influence our communication. When communicating with AI, these social factors don't apply, so it becomes more functional. Therefore, when talking to AI, be more explicit, provide context, and be careful about your language choice.
How Humans Communicate With AI
The way that humans communicate with each other is different from the way that humans need to communicate with artificial intelligence. This is because AI does not have the same understanding of the world as humans do. AI systems are trained on large amounts of data, but they do not have the same ability to understand context, nuance or to infer meaning from nonverbal cues.
In terms of human-machine communication, the interaction activates different neural pathways. There's less emotional involvement, and the areas of the brain associated with social and emotional processing are less active. The communication is more centered around problem-solving and task completion.
As a result, humans need to be more explicit when they communicate with AI. They need to provide more context and they need to be more direct with their use of language.
Why Context Matters
Let's start with the idea of context. In human conversations, we often rely on shared experiences and prior knowledge to communicate effectively. Picture this: I’m enjoying a peaceful moment of solitude on a Sunday morning, sipping a coffee while engrossed in The Sunday Times. Suddenly, I hear a distant shout from my wife upstairs. "Towels!" she yells.
What does she mean? Am I in hot water because I left wet towels on the bathroom floor again? Darn it! Or, is she in the shower, and forgot to take her towel with her? Cha-ching! Or, plot twist, maybe she's reminding me to pack the beach towels for our trip to the beach with the kids? Without context, I’m lost in a sea of fluffy, terrycloth confusion.
Similarly, in AI conversations, context is king. An AI, doesn't have personal experiences or the ability to infer meaning from the surrounding circumstances. Instead, it relies on the information it's given. The clearer and more detailed the context, the better the AI can respond.
Human Understanding of Context
Experience-Based: Human understanding of context is grounded in personal experience, sensory perception, and emotional resonance. We draw from our lifetime of experiences, emotions, and learned knowledge to understand the context of a situation.
Holistic Understanding: Humans take into account not just the words spoken, but also non-verbal cues, tone, mood, setting, cultural norms, societal structures, and much more. This multi-modal processing allows for a much more nuanced and complete understanding of context.
Empathy and Emotion: Humans can empathize with others, which allows us to understand the context from someone else's perspective. We can infer emotional states and intentions based on subtle cues.
Adaptation and Learning: Humans can adapt and learn from new situations, enabling us to update our contextual understanding over time.
Generative AI's Understanding of Context
Data-Driven: Generative AI's understanding of context is based on patterns it learned from training data. It has no personal experiences or emotions to draw upon. It doesn't truly understand context but uses statistical patterns to generate responses that appear contextually appropriate.
Text-Based: AI models like GPT-3 only understand text. They do not process non-verbal cues, tone, mood, or any other non-textual context. They also don't understand the world in the way humans do because they don't have a sensory experience of it.
Lack of Empathy and Emotion: AI does not have feelings, consciousness, or judgement and therefore cannot empathize with humans or understand emotional context in the same way humans do.
Static Training: GPT’s knowledge was cutoff in September 2021, so models like GPT don't learn or adapt based on their interactions after being trained. Their understanding of context is static and doesn't update over time.
When interacting with AI, try to be as descriptive and clear as possible. Imagine you're setting the scene for your AI friend, not just yelling "Towels!" but rather saying, "Darling, I'm in the shower and I forgot my towel. Could you please bring me one?" More context equals less confusion and a better, more productive conversation!
Structuring Your Prompts Logically
Along with context, the way we structure our prompts can significantly affect the responses we get from an AI. The goal is to be clear and logical. Try to think of your conversation with an AI as a highway. If you're clear in your directions and the path is well-laid out, the AI can smoothly follow along and get to the desired destination.
Consider the question, "What's the weather like in San Francisco today and what was it last Tuesday?" While it seems logical to us humans, it's quite complex for an AI to process. A better approach would be to break it down into two separate questions: "What's the weather like in San Francisco today?" and then, "What was the weather like in San Francisco last Tuesday?"
Understanding AI's Limitations
Finally, it's essential to understand that, as advanced as AI has become, it still has its limitations. Let's break down a few of them:
Hallucinations: No, your AI isn't seeing pink elephants! In the world of AI, "hallucination" refers to the tendency of AI to generate information that wasn't in the input or its training data. This can range from making up details in a story to providing incorrect or non-existent facts. Always cross-check information given by AI, especially when it matters. You don’t want to be like that lawyer last week who only relied on Chat GPT’s output and got in hot water.
Mathematical Challenges: While Chat GPT can handle simple arithmetic, complex calculations are not its strong suit. So, if you need help with your Beautiful Mind calculus, you might want to use the Wolfram plugin.
Humor: Humor is a uniquely human trait, deeply tied to our cultures, experiences, and even our personal tastes. AI, unfortunately, doesn't quite get the joke. It can mimic patterns of humor it's been trained on, but don't expect it to come up with a knee-slapper on the fly.
Despite these limitations, AI is a powerful tool that can help us in myriad ways, from answering questions to sparking creativity. The key is to understand its strengths and limitations, to communicate effectively, and to appreciate its unique capabilities.
In conclusion, while there's a learning curve involved in communicating with AI, it's well worth the effort. The more you understand and practice, the more you'll be able to harness the power of AI.
Remember, communicating with AI isn't about trying to make the machine human. It's about understanding the machine's limitations, capabilities, and language, and adapting our communication accordingly.
Comments