In this engaging discussion, cognitive scientist Gary Marcus shares his predictions for AI in 2025. He explores the contentious topic of artificial general intelligence, critiquing past generative AI predictions. Gary highlights the urgent need for regulation, sharing his experience testifying in the U.S. Senate about the risks of unregulated AI. He balances his critical outlook with an optimistic view of technological advancements, especially in large language models, underscoring the importance of realistic expectations as the field evolves.
The debate on artificial general intelligence reveals a stark divide between optimists like Sam Altman and skeptics like Gary Marcus regarding its timeline and feasibility.
Gary Marcus emphasizes the need for a shift in AI development strategies, advocating for the integration of classical approaches with large language models to overcome current limitations.
Deep dives
Contrasting Views on Artificial General Intelligence
The debate surrounding the progress of artificial general intelligence (AGI) highlights a sharp contrast in perspectives within the AI community. Prominent figures like Sam Altman assert that AGI is nearly resolved, suggesting significant advancements in the field, while Gary Marcus argues that we are far from achieving true AGI. Marcus has pointed out persistent issues in AI, particularly concerning the limitations of current models in extrapolating knowledge beyond their training data. This ongoing discourse reveals the complexity and uncertainty surrounding the development of AGI and the varying expectations among experts.
Challenges of Generative AI: Predictions and Reality
Marcus has made several predictions over the years regarding generative AI, indicating a tendency for these systems to hit limitations rather than achieve exponential growth. He specifically highlighted challenges like hallucinations, factual accuracy, and abstract reasoning that continue to plague AI, arguing that despite incremental improvements, fundamental issues remain unsolved. For instance, he predicted that by the end of 2024, there would not be a significant breakthrough like GPT-5. His consistent warnings stem from his early recognition of these underlying problems, emphasizing the importance of deep conceptual understanding over mere data interpolation.
The Necessity for New Approaches in AI Development
Marcus advocates for a paradigm shift in AI development to avoid stagnation and achieve meaningful advancements. He argues for a merger of large language models with classical AI approaches, which could facilitate deeper learning capabilities through neurosymbolic systems. Notably, he points out that while large language models excel in processing vast data, they lack the ability for more abstract, systematic reasoning. Drawing parallels between current AI and previous methodologies, he underscores the need for innovative breakthroughs to move toward reliable and trustworthy AI solutions.
We catch up with cognitive scientist Gary Marcus on his list of predictions for what to watch in AI this year, and beyond.
We Meet:
Author & Cognitive Scientist Gary Marcus
Credits:
This episode of SHIFT was produced by Jennifer Strong with help from Emma Cillekens. It was mixed by Garret Lang, with original music from him and Jacob Gorski. Art by Meg Marco.
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode