AI agents: a new way of interacting with technology

Last Thursday, 9 November, Microsoft co-founder Bill Gates published an interesting article on his blog about how artificial intelligence will completely redefine human interaction with technology. The change will be at least as significant as the shift from typing commands to interacting with icons on a touchscreen. In his reflection, Gates introduces the concept of an “agent”: advanced software that responds to natural language and can perform a range of tasks based on a deep understanding of the user. The article is particularly relevant because it appeared in the same week that OpenAI, the company behind ChatGPT, introduced a concept that represents the first clear step towards agents: custom GPTs.

AI agents, as Gates describes them, are more than simple tools: they are personal assistants that understand and anticipate users' needs. Instead of using different apps for different tasks, we will simply tell our device what we want to do in everyday language. Depending on how much information we are willing to share, the software will be able to respond in a personalised way because of its deep understanding of our lives. These agents will not be limited to simple tasks. They will be proactive, able to carry out tasks across applications and improve over time by remembering activities and recognising intentions and patterns in users' behaviour. Gates gives several examples, including this one: if you say you would like to catch up with an old university friend, the agent will work with your friend's agent to find a time to meet. Just before you arrive, it will remind you that his eldest son has just started at the local university.

AI agents will have a considerable impact in areas such as healthcare, education, productivity and entertainment. In healthcare, they will help both patients and professionals, facilitating diagnosis and the management of treatment. In education, they will personalise learning, adapting to each student's interests and needs. At work, these agents will optimise productivity by helping with a range of tasks and processes. In entertainment and shopping, they will offer personalised recommendations and help with decisions.

Gates also highlights that agents will be the next major platform in the computing industry, changing not only how we use software but also how it is written. He expects agents to replace search websites, e-commerce sites and traditional productivity applications, and to be accessible even to people without programming skills. We will be able to create any program or app we need simply by asking our agent for it.

AI agents in politics and government¶

Although Gates does not specifically address this field in his article, there is no doubt that AI agents have the potential to transform politics and government significantly. These agents could be used to improve efficiency in public administration, support data-driven decision-making and optimise government services. For example, an AI agent could help public officials analyse large volumes of information to formulate more effective policies, identify emerging social trends or even predict and manage crises.

One specific application could be emergency management. An AI agent could monitor real-time data from sensors and social media to detect early signs of natural disasters or humanitarian crises. Using this information, the agent could coordinate the emergency response, optimise resource allocation and provide crucial information to decision-makers. This system would not only make emergency responses more efficient and effective, but would also save lives by acting more quickly and accurately.

We asked GPT-4 to imagine specific agents for politics and government. Its suggestions included the following:

Agents for citizen participation:¶

Example: an AI agent designed to facilitate citizen participation in legislative processes. This agent could inform citizens about bills under consideration, collect and analyse public opinion, and provide legislators with a summary of those views.

Impact: it would improve participatory democracy by enabling more citizens' voices to be heard and considered in the legislative process.

Agents for policy analysis and forecasting:¶

Example: an AI agent able to analyse large volumes of data to anticipate the consequences of different public policies. It could simulate the effects of policies in areas such as the economy, public health or education.

Impact: it would enable policymakers to make better-informed, data-driven decisions, potentially reducing unintended outcomes.

Agents for transparency and combating corruption:¶

Example: an AI system dedicated to monitoring and analysing government transactions and public tenders to detect suspicious patterns or anomalies.

Impact: these agents could be powerful tools in the fight against corruption, increasing transparency and trust in government institutions.

Fernando Nieto Lobato

Director of Digital Innovation, Institución Educativa ALEPH