We’ve come a long way. The first article described the Turning Test and when this test was passed. We moved onto the experiment performed at MIT then onto a general overview of learning, human vs AI. I thought it important to establish some baseline in the learning process prior to jumping into inference. If you remember from that experiment at MIT, it was inference that put AI over the hump to it’s discovery of Halicen which the experts admitted would have taken decades, if ever, for them to have discovered.
Inference:
Definition from Oxford Languages: A conclusion reached on the basis of evidence and reasoning.
There are three basic types of inference. Let’s put humans head to head with AI.
| Type of Inference | Who Wins? | Why? |
Data-Driven (e.g., predicting a stock dip) | AI | It processes more variables than a human brain can hold. |
Logic-Driven (e.g., “If I move this chair, will it block the door?”) | Human | Humans understand physical space; the AI only understands the description of space. * This is now debatable given the MIT experiment |
Social-Driven (e.g., “Is my friend mad at me?”) | Human | Humans can infer “tone” and “history”; the AI only sees the text on the screen. |
Inference: Human vs AI
Let’s go ahead and add another table showing the differences in inference between humans and AI.
| Feature | Human Inference | AI Inference |
Foundation | Mental Model: “I know how the world works.” | Statistical Model: “I know how words/pixels usually cluster.” |
Incomplete Data | Humans use logic to bridge the gap. | It uses probability to bridge the gap. |
New Situations | High. Humans can guess what happens in a brand-new environment. | Low. It struggles when a situation is “Out of Distribution” (not in its data). |
Hidden Patterns | Low. Humans often miss subtle correlations in big data. | High. It excels at finding “invisible” links in massive datasets. |
Conclusion
Given the information above, I see a tremendous opportunity for AI to work together and along side humans. At our jobs we enjoy having a subject matter expert with us. A person, or now and AI, that brings different skill sets to the table. We see above that AI wins when inference is applied to processing big data and finding hidden patterns. Humans win when inference is applied to social and new situations. Lastly, I’m going to call it a toss up between AI and humans when inference is applied to logic. For this particular area, I think it depends on the situation.
