Articles

Ubiquity

Volume 2025, Number July (2025), Pages 1-5

Ubiquity Symposium: Artificial Intelligence: Introducing the ACM Ubiquity Symposium on Artificial Intelligence
Jeff Johnson, Peter J. Denning
DOI: 10.1145/3747352

Artificial intelligence is changing our world. The arrival of ChatGPT in 2022 brought AI into the public spotlight with a dramatic new capability. A user can how engage in a fluent conversation in ordinary language with a computer. Suddenly many people want to know what AI is, whether it is safe, and what benefits it might bring. Artificial intelligence is defined to be a collection of machines and algorithms that perform tasks normally considered to require human cognition. Opinions and claims about the possible benefits and risks of AI are all over the map. In response to the confusion, the Ubiquity editors are launching a symposium on AI. It will be a series of about two dozen articles on the many aspects of AI and its applications. The series will be synthesized into the "Ubiquity Report on Artificial Intelligence," which is intended to be an authoritative and trustworthy summary of present and future AI.

Artificial intelligence (AI) is changing our world. The dramatic arrival of ChatGPT in November 2022 suddenly made AI widely accessible. Within five days of its release, more than 100 million people had accessed it. Soon there were other offerings of the technology, which is now called generative AI (GenAI) or large language models (LLMs). Anyone can get a free account with ChatGPT (Open AI), Gemini (Google), Claude (Anthropic), or one of many other LLMs. Users with no prior experience of programming or knowledge of AI can today hold fluent conversations in ordinary language with a computer.

LLMs are only the leading edge of a new wave of AI technologies. AI promoters believe these new technologies will impact on virtually everything we do. But even as numerous new consumer-level AI products appear, many leaders in the field believe AI is very far from achieving human-like intelligence and its capabilities and potential impact are greatly exaggerated. Every day the media contains articles, texts, podcasts, and videos with stories and speculations about AI. Opinions and claims about the possible benefits and risks of AI are all over the map.

In response to the confusion, the Ubiquity editors seek to bring some order to the cacophony. Our symposium on AI is a series of about two dozen articles investigating the many aspects of AI and their impacts. When the symposium is complete, we plan to synthesize the best of the papers into "The ACM Ubiquity Report on Artificial Intelligence," which we intend to be an authoritative and trustworthy summary of present and future AI.

A major aspiration of AI is artificial general intelligence (AGI). The premise of AGI is for machines to understand and learn any task a human being can and develop all the cognitive abilities of the human brain. At present, no existing machines display anything like intelligence, understanding, or care. They are sophisticated data processors. No one knows whether AGI machines are even possible.

The quest for AGI is complicated by the lack of an accepted definition of artificial intelligence. In his 2024 book Artificial General Intelligence, Julian Togelius (a professor in the Department of Computer Science and Engineering at New York University) enumerates and explains the many interpretations afoot, and the expectations each one generates. These range from a bright view that AI is fundamentally beneficent and will produce an "AI utopia," to a dark view that AI machines will develop superhuman intelligence that loathes humanity. For the purposes of this symposium, we will use the following definition:

Artificial intelligence is a collection of machines and algorithms that perform tasks normally considered to require human cognition.

The machines and algorithms constituting AI are not stand-alone. They are composed of many existing technologies:

• Search; looking for solutions to problems and answers to queries in a large space of possible answers
• Reasoning; applying rules and facts to make logical deductions
• Artificial neural networks (ANNs); interconnected artificial neurons that mimic neurons and their connections in the brain
• Natural language processing (NLP); systems that analyze text and discern its meanings
• Large language models (LLMs) and generative AI (GenAI); systems trained on huge text corpuses that respond to "prompts" with highly likely texts and images
• Multimedia; systems designed to make pictures, drawings, videos and sound
• Programming; almost all AI systems use conventional software engineering to assemble reasoning and machine learning modules into useful systems
• Human-computer interaction; design principles for interfaces that make the user experience intuitive, smooth, efficient and enjoyable
• Communications; networked communication including mobile phones and the internet
• Hardware; highly parallel chips that bring supercomputing power to the simulations, programming, and training referred to above

Most AI systems available to the public use combinations of these technologies. These and other technologies supporting AI will be explained in the symposium.

Although today AI systems can do some very impressive things, they are still very far from AGI. And, frankly, we are not even sure that AGI should be the goal. A much better goal would be AI tools that are safe and useful for everyone.

The quest for AI has raised many interesting questions, including:

• Will AI ever achieve a "singularity" with machines more intelligent than humans?
• Will AI replace most jobs?
• Can AI augment work without displacing workers?
• Will AI make production of food and basic necessities so cheap that no one will need a job?
• Will GenAI based on LLMs improve, make fewer mistakes, and become trustworthy?
• Are LLMs a breakthrough or a dead end?
• Will GenAI transform the ways we live and work?
• Will AI revolutionize education?
• Will AI automate cheating and frauds?
• Could AI instill a love of mathematics in my granddaughter?
• Can we manage the penchant for abuse, conspiracies, falsehoods, fake voices and videos?
• How should AI be regulated to keep people safe?
• Will regulated AI stifle innovation?
• Will a team of humanoid robots beat the world champion soccer players by 2050?
• Will AI exacerbate international conflicts and geopolitics through disinformation and faulty simulations?
• Will killer machines be able to autonomously choose human targets, e.g. swarms of drones?
• Can AI be prevented from interfering with democratic elections?
• Will AI supercharge cyberattacks on businesses and governments?
• Can AI be used to manipulate people's emotions and values?
• Can AI detect falsehoods?
• Is AI combined with social media safe for children?
• Can AI protect children?
• Will copyright and intellectual property be respected when gathering training data?
• Will synthetic data degrade the performance of LLMs?
• Can the knowledge of all humanity be captured in LLMs, leading to an ultra-wise and all-knowing LLM?
• Will user communities gravitate toward ethical use of AI?
• Will ANNs or LLMs produce breakthroughs in science?
• Will AI machines be able to design and implemented better versions of themselves?
• What is the ultimate value of AGI and the quest for it?

The authors in our symposium will explore these and other questions in the pragmatic uses and dangers of AI tools. We will avoid hype and cite relevant science to support our conclusions. Points-of-view will be expressed in terms specific enough to be testable by operational methods within a time horizon of 10 to 20 years. We invite your opinions, views, and practical experiences with AI. We welcome your suggestions for articles on topics that we should address.

(Please respond to symposium editor Jeff Johnson using the email address [email protected] and he will reply within a few days.)

—The Editors¹

Footnotes

1. Rob Akscyn, Espen Andersen, Kemal Delic, Peter Denning, Jeff Johnson, Andrew Odlyzko, Michael Quinn, Walter Tichy, Martin Walker, and Phil Yaffe

2025 Copyright held by the Owner/Author.

The Digital Library is published by the Association for Computing Machinery. Copyright © 2025 ACM, Inc.