The Boundaries Of AI: What Artificial Intelligence Can And Cannot Do – Followed By Q&A Session

The Boundaries Of AI: What Artificial Intelligence Can And Cannot Do - Followed By Q&A Session
The Boundaries Of AI: What Artificial Intelligence Can And Cannot Do – Followed By Q&A Session

The last decade has seen exciting developments in AI – and AI researchers are tackling some fundamental challenges to develop it further

AI research seeks to understand what intelligence is, and then recreate this through computer systems that can automatically perform tasks that require some level of reasoning or intelligence in humans.

In the past, AI research has concentrated on creating detailed rules for how to carry out a task and then developing computer systems that could carry out these rules; researchers and programmers would develop a solution to a problem, codify this into a computer program, and in so doing create a system to implement the solution.

By pre-programming computer systems with rules in this way, computer systems can perform as well as experts in certain, specific tasks. However, while these systems have been successful in highly specific domains – such as playing chess – they have a significant weakness: inflexibility. If something unexpected happens or inputs do not follow a standard pattern, these systems cannot adapt their rules or adjust their approach.

In the last decade, new methods that use learning algorithms have helped create computer systems that are more flexible and adaptive, and Demis Hassabis FRS (co-founder, DeepMind) has been at the forefront of many of these developments.

AI technologies today are a topic of widespread interest, and there is a lot of speculation (and hype) about what AI might or might not achieve in the coming years. However, there remain significant limitations to what AI technologies can do (see Current challenges in AI research). For some, these limitations make AI particularly exciting, as they generate new lines of research that address profound questions about the nature of intelligence.

About Speaker: Dr Demis Hassabis | FREng FRS | Co-founder, DeepMind
In a lecture exploring the frontiers of AI, Demis Hassabis explained how AI algorithms can learn how to carry out a task, by analysing data and experiencing their environment. Instead of relying on designers to specify a solution to a problem, these algorithms use data to figure out how to solve a problem.

About Series: You And AI: Conversations About AI Technologies And Their Implications For Society
Artificial Intelligence (AI) is the science of making computer systems smart, and an umbrella term for a range of technologies that carry out functions that typically require intelligence in humans. AI technologies already support many everyday products and services, and the power and reach of these technologies are advancing at pace.

The Royal Society is working to support an environment of careful stewardship of AI technologies, so that their benefits can be brought into being safely and rapidly, and shared across society. In support of this aim, the Society’s You and AI series brought together leading AI researchers to contribute to a public conversation about advances in AI and their implications for society.