Stay updated with the latest in AI by joining our daily and weekly newsletters for exclusive industry content.
About a year ago, I discussed how 2023 would be the year dominated by large language models (LLMs). Models such as Llama 2, Claude, and Cohere have emerged, challenging OpenAI and spurring innovation despite various challenges. After such an explosive year, what can we expect for AI in 2024?
The coming year promises advanced AI being applied in new and creative ways, driving significant progress across various industries. However, there are also clear signs that AI may be misused by malicious actors. While the future is uncertain, it’s evident that AI advancements in 2024 will heavily influence both our professional and personal lives.
Focus on AI Agents
AI development is increasingly centered around AI agents, as highlighted during OpenAI’s recent DevDay event. These sophisticated tools are beginning to impact multiple industries, but the advancements so far are just the beginning.
Earlier this year, the ReAct paper demonstrated how LLMs could effectively learn to use tools, inspiring further research. Companies like OpenAI and Anthropic have refined their models to integrate these techniques, with OpenAI’s Function Calling and Anthropic’s Claude XML support being notable examples. Institutions like Berkeley have developed specialized LLMs, and open-source libraries such as Langchain and Rivet have made these techniques more accessible.
AI agents are now easier and more affordable to develop, making them likely to become widespread. They enhance human creativity and efficiency by connecting deeply with important data. We may look back on 2024 as the start of the “age of agents,” a new era in software and technology interaction.
Enhanced User Expectations
The rise of intelligent agents will bring a significant change in user expectations. Customers will begin to demand higher levels of responsiveness and interaction from their technology, transforming it from a tool to a collaborator.
Technological shifts often lead to changes in user expectations, as seen when Apple released the first iPhone and people began expecting more intuitive mobile device controls. Similarly, as consumer-focused cloud apps became popular, enterprise users demanded simpler and more user-friendly work tools.
As more people become accustomed to AI assistants, they will expect similar smart responses in other areas of their lives. These agents are not just minor improvements; they offer new capabilities, allowing users to achieve more.
Assistants like Microsoft Copilot and Google Duet can draft documents, summarize emails, create presentations, and perform other tasks. As these agents become more common, companies that don’t adopt them may risk losing customers.
Breakthroughs in AI Vision
ChatGPT’s ability to understand and express natural language was a major breakthrough, but AI vision could be even more impactful. LLMs can now train on both text and visual data, making them multimodal. OpenAI’s GPT-4 and Google’s Gemini exemplify this, with more models likely to follow.
Images and illustrations can communicate complex ideas more effectively than words alone. Wearable devices are being developed to assist in daily life by providing background information on people, visual cues for work, or real-time task suggestions.
Predicting the exact trajectory of innovation is challenging, but interpreting images and videos in real-time adds a crucial dimension to how intelligent AI agents can assist humans.
AI-Powered Manipulation
Imagine receiving a link from a friend that leads you to a social network group with various users interacting. It feels real, but everything could be fake. Misinformation has always been a concern, but social proof (“if others trust this, it must be trustworthy”) is no longer reliable because we can’t be sure who is real and who is an AI bot.
Technology has never before been so capable and accessible for influencing and manipulating people on a large scale. AI has blurred the lines between real and machine-produced interactions and content. Images and videos can be easily generated to show almost anything.
This advanced technology is now accessible to almost anyone, not just sophisticated hackers or nation-states. The coming year might witness the consequences of AI-powered manipulation, ranging from automated blackmail and fraud to spreading conspiracy theories.
AI will bring many incredible advancements in the next year, but it will also present new challenges. I believe that society can harness the positive aspects of this technology while adapting to its risks. Though the adaptation process may be rocky next year, I am confident we will overcome these challenges.
Cai GoGwilt, cofounder and chief architect of Ironclad.