How AI became my sparring partner

A few months ago, I was attempting to comprehend why one high-protein formulation expanded during low-moisture extrusion while another did not. I provided ChatGPT with the formulation specifics, expansion results, rheological measurements, and trial observations. I then began querying it about potential explanations. One line of reasoning hinged on critical strain, which refers to the extent of deformation a material can endure before its structure begins to disintegrate.

A plausible hypothesis suggested that greater deformation tolerance might facilitate material expansion. However, an issue arose. One of the materials with the highest critical strain had underperformed in expansion. Consequently, I challenged that explanation: if it were accurate, how could the experiment yield the opposite outcome?

We reassessed the interpretation, incorporated other observations, and explored alternative hypotheses. Ultimately, the critical lesson was not that a single rheological measurement could predict puffing. Instead, it was that multiple properties and processing events needed to be considered together. This exchange led me to realize that I was no longer merely utilizing AI to retrieve information.

Rather, I was leveraging it as a sparring partner. It did not merely return the ball, as I previously engaged in table tennis, and provided a competent sparring partner would. Instead, one shot went to my forehand, another to my backhand. At times, the ball even reached a corner I would have preferred to avoid. This is how I now engage with AI.

I can propose an explanation; ChatGPT offers an alternative possibility. I question it, it challenges an assumption, and I introduce additional experimental data. Sometimes the original explanation persists; sometimes it does not. And sometimes the correct conclusion is simply that we lack sufficient evidence to make a definitive judgment.

The value lies not in AI providing the answer. Rather, the exchange assists me in formulating better questions. There is a comprehensible apprehension that readily accessible AI-generated answers might impede critical thinking. However, my experience has often been the opposite. When employed in this manner, AI prompts me to scrutinize my own thinking more rigorously.

Furthermore, this has proven particularly advantageous in my advisory work. When assessing research and innovation projects spanning disciplines outside my expertise, I must first comprehend the subject matter sufficiently to evaluate the proposal's scientific foundation. Previously, entering an unfamiliar technical domain could be time-consuming.

I would conduct searches for background material, commence reading, encounter terminology or mechanisms I did not grasp, and then conduct further searches, gradually constructing a mental map of the field. AI has significantly expedited that initial stage. I can commence with elementary questions. If an explanation introduces another unfamiliar concept, I can inquire immediately.

I can compare mechanisms, question which interpretation holds more weight, and persist in probing until I grasp what I should be reading and questioning. It does not evaluate the project for me. I must still examine the original material and evidence and make the judgment myself. Nonetheless, it significantly reduces the cost of being inquisitive.

AI also streamlines time-saving processes in more apparent ways. I employ it to organize lengthy documents, compare versions, restructure technical content, and work through scientific manuscripts. However, efficiency raises another concern: trust. Recently, I meticulously worked through numerous versions of a lengthy document with ChatGPT.

This was not a situation where I surrendered the task and accepted whatever emerged. I questioned arguments, requested specific alterations, and persistently refined the direction. After each round, ChatGPT summarized the changes it had made. Gradually, I recognized that I had begun trusting those summaries. If it informed me that a particular point had been rectified, I no longer always reopened the entire document and verified every amendment before proceeding to the next issue.

The collaboration had evolved to the extent that I trusted the account of what had been done. Additionally, I find myself pondering a distinct question from when I initially began using AI. Not merely: What tasks can AI accomplish for me? But: What kind of thinking does working with AI enable me to perform more effectively? Occasionally, it aids me in entering unfamiliar fields more swiftly.

It assists in managing an overwhelming volume of information. And at times, it poses a question from the other side of the table that uncovers an assumption I had not previously recognized. This is why I persist in returning to the table-tennis analogy. A proficient sparring partner does not play the match for you. It does not eliminate the need for technique, practice, or judgment.

Instead, it enhances your ability to cover more of the table. The most effective use I have found for AI is not outsourcing my judgment. Rather, it provides my judgment with a superior sparring partner.