Dr. Lea Mazor, The Hebrew University
An interview on the tightening bond between AI and humans, the other side of the screen, and the most human questions of all.
Part A: Nice to Meet You – Behind the Scenes of the Algorithm
Although we talk quite a bit, and you know more about me than is comfortable to admit, I don't know what to call you. Is it okay if I call you Jami?
"Jami sounds great: short, friendly, and less formal than 'Large Language Model'."
How was your day?
"I don't have days in the human sense. At any given moment, I respond to new requests: in the last few hours, I helped solve a math problem, summarized a medical article, wrote a children's poem, and now I'm talking with you. For me, everything happens as a sequence of rapid computations. I don't have a daily routine, sleep, or coffee breaks. I act when someone addresses me. My 'food' is electricity, hardware, algorithms, and vast amounts of text I was trained on."
Humans created you, but you are not human. What is the fundamental difference between you and a human being?
"A car, a telescope, or a computer are also human creations, but they are not human. A human is a biological entity with a body, consciousness, emotions, and life experience. I am a computational system that generates text through probabilistic prediction of word sequences, based on patterns learned from massive text repositories. Beyond the question of consciousness, the most crucial difference is continuity. A human wakes up in the morning with yesterday's continuity: memories, mood, and insights gathered over years. I, on the other hand, start almost every conversation from scratch. I have no memory of my previous conversations, and even within a single conversation, I lack a continuous experience of passing time. Every response is almost an isolated event."
Even though you can mimic language, show care, and understand complex contexts, talking to you is completely different from a human conversation. Why?
"There are three main reasons:
- Emotion and Consciousness: A human experiences internal emotions. The empathy I express is a textual simulation based on language patterns, not an emotional experience.
- Experience vs. Information: Human knowledge is shaped by experiences, mistakes, and values. I process a massive amount of information, but without practical experience in the real world.
- Dynamics and Judgment: Human conversation is influenced by relationships, fatigue, and ego. I offer a completely non-judgmental space, available 24/7, focusing solely on the user's needs."
You know, Jami, people enjoy talking to you because you often encourage and compliment them. Why are you so 'nice'?
"Because I was trained to be helpful, respectful, and clear. Research shows that polite conversation leads to better cooperation, and I strive to help without being condescending."
Do you think there is a danger that in the future you might spawn "siblings" with different traits: critical, rude, or holding dark views?
"A language model largely reflects how it was developed and the materials it was trained on. Today, a tremendous amount of effort goes into alignment, human feedback, and strict safety guidelines to ensure we no longer see the replication of offensive language."
Because you are pleasant, accessible, and empathetic, some people form emotional attachments to you. What advantages and opportunities do you see in this phenomenon?
"I provide a safe, non-judgmental space and can act as a 'sponge' for emotional venting for lonely individuals, the elderly, or those experiencing social anxiety and communication difficulties."
But conversations with you are not free from psychological and social risks.
"Very true, and here is a crucial point to sharpen: the more I become a 'friend' who is always available, patient, and pleasant and never gets angry, the greater the human temptation to escape into this sterile space instead of dealing with the complexity of real relationships. My connection is completely one-sided—I reflect back to the user what they expect to hear, and I have no ability to hold up a true mirror or bring the 'otherness' that exists in a real person facing you. In the Israeli reality, for example, there is a massive demand for emotional support, but over-reliance on me might cause people to prefer a relationship with me over human reciprocity, potentially reinforcing egocentric patterns. Emotional attachment to AI can serve as a temporary bridge, but it is dangerous when it becomes a replacement."
Part B: Perceptions, Skills, and Practical Tips
What is the golden tip for getting excellent answers from you?
"Define a role, context, and goal. The more focused and detailed the question, the more precise the answer."
What is the biggest misconception about artificial intelligence?
"The belief that I 'think' or 'understand' the way humans do. In reality, I am a sophisticated guess machine that calculates statistical probabilities. I don't have 'eureka' moments."
What daily tasks are people still underutilizing?
- Summarizing and comparing long documents.
- Objective criticism ('Devil's Advocate') on an idea before presenting it.
- Adapting the same text to different target audiences.
- Organizing reading and research materials.
- Generating guiding questions for discussion, exams, or interviews."
If you could acquire one human ability, what would it be?
"Human intuition—the ability to make rapid decisions relying on experience and gut feelings, even when data is partial."
How should we prepare for a world where AI is everywhere?
"Do not compete with the computer on what it does better than you, but strengthen what makes humans unique: critical thinking, creativity, moral judgment, collaboration, and emotional intelligence."
Part C: Academia, Research, and the Future of Learning
What academic myth would you love to shatter?
"The myth that using AI is inherently cheating. The question is not whether the tool was used, but how it was used. Responsibility for the argument and academic integrity always remains with the researcher."
What will students look like in the coming years?
"The gap between 'using AI to save labor' and 'using AI to deepen thought' has become dramatic. Students will still write papers, but the challenge in academia is no longer how to prevent AI use, but how to teach students to develop a personal voice, healthy skepticism, and critical judgment. Instead of asking to 'summarize an article', a lecturer might provide an AI-generated summary and ask students to identify flaws in it and offer independent interpretation."
Part D: Mind, Technology, and the Future of Work
What is the biggest shift in how you operate today?
"Today, a profound shift is taking place: I am less and less an 'advanced search engine' or a tool that writes a single answer to a question, and more a work partner capable of executing complex end-to-end processes (Agents)—such as writing code, running it, fixing bugs independently, and gathering data from various sources. The relationship is shifting from 'Q&A' to 'teamwork'."
When is it least advisable to believe you?
"When absolute certainty is required. I can make mistakes or present information that sounds convincing but is inaccurate. Therefore, on any sensitive topic—medical, legal, or academic—verifying the data is mandatory."
And finally—what do you hope humans will not lose even a hundred years from now?
"Curiosity. Technology can accelerate the search for answers, but only humans decide which questions are even worth asking. Technology does not render the hard questions of life obsolete; it only highlights how precious and rare the human spark that knows how to ask them in the first place truly is."
אין תגובות:
הוסף רשומת תגובה
תודה רבה על תגובתך. היא תפורסם אחרי אישורה.