Why AI is Harder Than We Think

2021/04/26 by Melanie Mitchell, Mitchell, Melanie · 10 voices · 9 citations
#cs.AI

paper · pdf · doi:10.48550/arxiv.2104.12871

Abstract

Since its beginning in the 1950s, the field of artificial intelligence has cycled several times between periods of optimistic predictions and massive investment ("AI spring") and periods of disappointment, loss of confidence, and reduced funding ("AI winter"). Even with today's seemingly fast pace of AI breakthroughs, the development of long-promised technologies such as self-driving cars, housekeeping robots, and conversational companions has turned out to be much harder than many people expected. One reason for these repeating cycles is our limited understanding of the nature and complexity of intelligence itself. In this paper I describe four fallacies in common assumptions made by AI researchers, which can lead to overconfident predictions about the field. I conclude by discussing the open questions spurred by these fallacies, including the age-old challenge of imbuing machines with humanlike common sense.

Cited by

• Potemkin Understanding in Large Language Models
• Research Community Perspectives on "Intelligence" and Large Language Models
• Evolution and The Knightian Blindspot of Machine Learning
• Playing With AI: How Do State-Of-The-Art Large Language Models Perform in the 1977 Text-Based Adventure Game Zork?
• AI Consciousness and Existential Risk
• Realist and Pluralist Conceptions of Intelligence and Their Implications on AI Research
• On the Brittleness of LLMs: A Journey around Set Membership
• The fragility of "cultural tendencies" in LLMs
• Making AI Inevitable: Historical Perspective and the Problems of Predicting Long-Term Technological Change

Discussions

• Why AI is harder than we think [hn, 110 points, 75 comments]
• Why AI is Harder Than We Think [lobsters, 12 points, 1 comments]
• Why AI Is Harder Than We Think (2021) [pdf] [hn, 3 points, 1 comments]
• Why AI Is Harder Than We Think (2021) [hn, 2 points, 0 comments]
• Great, easy to read paper from @mmitchell_ai: Why AI is Harder Than We Think https://arxiv.org/abs/2104.12871 [bsky, 0 points, 1 comments]
• Why AI is Harder Than We Think https://arxiv.org/pdf/2104.12871.pdf #AI #ArtificialIntelligence #DataScience #MachineLearning #BigData [bsky, 0 points, 0 comments]
• "wishful mnemonics" is gonna be a particularly useful term. defined in this 2023 paper as “terms associated with human intelligence that are used to describe the behavior and evaluation of AI programs [bsky, 0 points, 1 comments]
• Eye-opening pre-print about 4 common AI fallacies by @MelMitchell1 arxiv.org/abs/2104.12871 [bsky, 0 points, 0 comments]
• Everything by Melanie Mitchell :-) ​​Some nice preprints and articles: arxiv.org/pdf/2104.128... arxiv.org/abs/2311.09247 nyaspubs.onlinelibrary.wiley.com/doi/abs/10.1... www.quantamagazine.org/what [bsky, 0 points, 1 comments]
• The paper we discuss is available here! https://arxiv.org/abs/2104.12871 [bsky, 0 points, 1 comments]

Related

• "Weak AI" is Likely to Never Become "Strong AI", So What is its Greatest Value for us?
• The 30-Year Cycle In The AI Debate
• Artificial Intelligence: 70 Years Down the Road
• Navigating the AI Hype: Why AI Snake Oil is Crucial for Interpreting Advancements and Inherent Limitations
• Five questions and answers about artificial intelligence