I watched a TED talk by Janelle Shane on the dangers of artificial intelligence (AI). But not the typical dangers we think of when discussing AI. Not AI stealing our jobs or taking over the world. It is more about the weirdness, almost humorous nature of AI when given a task or objective. As we’re shown from different experiments, AI gets the job done, no doubt. But not exactly how we thought it would, or ever could have imagined. For example, when told to get from point A to point B, one would assume the AI would develop legs and walk to the end point. But the AI with a limited mind of its own, instead gave itself a tall leg tower and simply fell over to reach point B. It clearly completed its objective. And that is the “danger” of AI. We have to be a little more specific with what we want it to do because then, it’ll create an absurd solution.
Another example shown was Amazon’s attempt to create a resume sorting algorithm based on applications of people they have hired in the past. Based on these instructions, the AI learned to discard resumes that included the word “women” in them. It’s mishaps like these that remind us that AI has a long way to go before matching the human mind.
The presentation overall was brief and entertaining. The AI’s hijinks presented were pretty comical. The audience seemed to enjoy it. The speaker spoke confidently and knowledgeably. I would enjoy learning more about AI.