By Dave DeFusco
Artificial intelligence is everywhere right now. It writes emails, summarizes documents, creates images and answers questions in seconds. For many people, that sudden visibility raises big questions: Will AI replace my job? Can I trust what it tells me? Is this real progress or just hype?
Those questions were at the heart of a recent program on 鈥淪eparating Science from Fiction: What AI Means for Jobs, Growth and the Economy,鈥 at the YU Museum in November. Sponsored by the Sy Syms School of Business, Katz School of Science and Health and YU Office of Alumni Affairs, the event featured a wide-ranging conversation between two experts who have spent decades studying, building and questioning artificial intelligence.
Dr. David Magerman, a renowned computer scientist with a Ph.D. from Stanford University and more than 25 years of experience in AI and data science, began by addressing a common misconception. 鈥淚鈥檝e been doing AI for a long time,鈥 he said. 鈥淧eople today think AI was started five years ago, but that鈥檚 just not true.鈥
He explained that artificial intelligence as a field dates back to the 1950s and 1960s, when early computer scientists began asking whether machines could mimic human thinking. One famous early example was a simple program called ELIZA, which imitated a therapist by repeating and reframing users鈥 words. Though the technology was basic, people were amazed.
鈥淚t looked like it was actually human,鈥 said Magerman. 鈥淧eople were ready to say, 鈥榃e鈥檝e got artificial intelligence like a human.鈥 Obviously, it was oversold.鈥
That cycle鈥攂ig excitement followed by disappointment鈥攈as repeated itself throughout AI鈥檚 history. Over time, many successful AI tools stopped being called AI. 鈥淕oogle search is probably the most valuable piece of AI ever created,鈥 said Magerman. 鈥淲e don鈥檛 call it AI anymore. It just works.鈥
Dr. Pablo Hernandez-Lagos, director of the Sy Syms MBA program and a leading scholar at the intersection of AI and the global economy, built on that idea. He emphasized that what the public sees today鈥攖ools like ChatGPT鈥攊s only a small part of a very large field.
鈥淭his is the result of thousands of scientists writing thousands of papers,鈥 said Hernandez-Lagos. 鈥淲hat we are seeing now is just the tip of the iceberg.鈥
He explained that recent breakthroughs, including neural networks and a technique called the "transformer," made AI tools faster and more commercially appealing, but there is a psychological reason this moment feels different.
鈥淲hen AI imitates humans, we get the chills,鈥 said Hernandez-Lagos. 鈥淏ut imitation does not mean intelligence.鈥
Magerman warned that AI companies often impress the public with flashy demonstrations that don鈥檛 reflect real-world usefulness. 鈥淭here鈥檚 a reason every robotics demo you ever see is playing soccer or dancing,鈥 he said. 鈥淭hose things look amazing, but they鈥檙e not very useful in industry.鈥
Companies, however, continue to promote these systems as nearly finished products. According to Magerman, this gap between marketing and reality fuels public confusion. Audience members asked whether today鈥檚 AI boom resembles the early days of the internet. Both speakers agreed there is a bubble, but not the same kind.
鈥淭he internet was infrastructure,鈥 said Magerman. 鈥淎I is an application.鈥
While the internet created a foundation that transformed society even after the dot-com crash, Magerman argued that today鈥檚 AI hype is driven by tools that are expensive, inefficient and often misused. 鈥淲e鈥檙e using a sledgehammer to push in a thumbtack,鈥 he said.
Hernandez-鈥婰补鈥媑辞蝉 added that massive corporate spending makes AI seem unstoppable, but much of that money, he said, circulates within a small group of tech firms, inflating stock prices without clearly benefiting the broader economy. "One company today spends more on AI than entire countries spend on research and development," he said.
The conversation then turned to the issue many in the audience cared about most: employment. Hernandez-L鈥媋鈥媑辞蝉 summarized current research on AI and jobs. 鈥淪o far, the overall effect is very small,鈥 he said. 鈥淪ome entry-level positions, such as accounting, software development and clerical roles, are more exposed to automation, but the impact is limited.鈥
The key factor, he said, is expertise. 鈥淭hose who are experts actually receive a compensation bump,鈥 said Hernandez-L鈥a鈥嬧赌嬧赌gos. 鈥淚f you are an expert, AI becomes a tool that helps you. If you are not, it may not help you at all.鈥
Magerman added that recent tech layoffs were often blamed on AI, but the reality was more mundane. He also warned that replacing junior workers with AI could create long-term problems. 鈥淐ompanies overhired and AI became a convenient story to mask cost-cutting,鈥 he said. 鈥淚f entry-level employees don鈥檛 do the work, how do you know who to promote?鈥
Hernandez-Lagos shared research showing that widespread AI use can actually make hiring harder. When everyone uses AI to polish resumes and applications, it becomes more difficult to tell who truly has expertise. For highly skilled candidates whose first language is not English, however, AI can help real expertise stand out rather than be hidden by language barriers.
As the evening drew to a close, the conversation widened to ethics, responsibility and values. Magerman spoke about the importance of constrained capitalism and building technology that serves humanity, not just profit. Hernandez-Lagos echoed that sentiment, acknowledging the tension between innovation and responsibility.
"We live in a world driven by incentives," he said, "but character and trust still matter."
