A ascensão da IA agêntica, os avanços na IA multimodal, a GenAI se tornando onipresente, a computação com eficiência energética, a IA responsável e ética, etc., estão previstos
Por Kamal Das
As we approach the new year, we highlight our view on the major themes in AI in 2025 based on our interactions with AI subject matter experts.
Rise of Agentic AI
Agentic AI systems have the capability to make decisions autonomously, reason independently, reiterate results, and act toward predefined tasks or goals. Agentic AI uses generative AI, large language models or other advanced AI to perform tasks with little or no human intervention.
In 2025, we expect autonomous Agentic AI to be a key area of focus moving from performing simple automation tasks to more sophisticated tasks. These systems will become more independent and get better at complex planning and taking actions to achieve the allocated tasks.
For instance, in supply chain optimization, agentic AI may optimise delivery time and costs by autonomously planning logistic routes, checking the data on the traffic in proposed routes, and rerouting shipments during disruptions. All this can potentially be completed with minimal or no human intervention.
This will result in enhanced automation, and better operational efficiency leading to significant improvements across multiple industries.
Advances in multimodal AI
Conventionally, AI systems work with a single type of data (such as text, images, or audio) or modality. AI systems are typically fine-tuned to work on the specific data type. We have started to see multimodal AI systems which are capable of processing data from multiple data types in an integrated manner.
Let us take an example of AI to understand a video. If an AI system has only textual understanding capability, it typically extracts the subtitles or captions of the video and answers queries using these. It cannot analyse the images as it lacks the capability to analyse visual data.
With multimodal AI, which can understand text and images, we will get better responses to the queries as it will have insights of both the audio and visual information. As a result, we will get more contextual and accurate information.
We expect to witness significant advancements in multimodal AI in 2025. This will result in simpler interactions as one AI system will be able to assist using a variety of data types. This will also enable a more comprehensive understanding of information by the AI system and they will deliver a more human-like response. In the area of health care, multimodal AI could combine visual scans (X-rays), patient medical histories (text), and voice recordings of patient symptoms to provide holistic diagnostic reports.
GenAI becomes ubiquitous
Generative AI (GenAI) is a subset of AI which has the ability to generate new content such as text, images, music, code, or other forms of media. ChatGPT is the most famous example of GenAI.
Apple has integrated GenAI features into its devices through Apple Intelligence. Other popular applications of GenAI include Google’s Gemini, Meta AI (integrated into WhatsApp), Microsoft Copilot, Claude, and Perplexity.
In 2025, we expect GenAI to continue to grow and get integrated with everyday devices. These become a core companion, seamlessly integrated into our work environment as well as daily lives. Automation of repetitive tasks will save time for many of us. These systems will help us prioritise tasks, make decisions, and draft our responses! Real-time language translations and advanced photo editing are examples of features that will make this part of everyone’s lives.
Cybercrime, fake news, and deep fakes risk increase
According to researchers, cybercriminals can recreate a person’s voice with 85% accuracy using a mere 3-second audio clip! With the advancements of AI, tools to generate fake news and deepfakes are expected to increase in 2025.
Cybercriminals will use AI systems to create and spread incorrect information and fake content. By 2025, the cost of cybercrime may reach $10.5 trillion in part due to AI-enabled threats.
As the saying goes, “poison kills poison.” Similarly, we will witness increasing usage of AI to combat cyber security threats. AI will be used to defend against malicious attacks. AI will aid in detecting fake news and deepfakes and warn people of the risks in real-time.
Energy-efficient computing
Data centres, which host many of the AI models are big energy guzzlers contributing to around 2% of global energy consumption in 2022. This may increase to around 4% by 2030.As a result, we expect to see a focus on energy-efficient AI computing, and algorithms and hardware. The AI industry will focus on more environment-friendly practices and aim to reduce the carbon footprint and increase the sustainability of AI technologies.
Expanding AI applications in social sectors
AI has already made considerable progress in corporate; we expect to see AI used extensively in the social sector, benefiting the common people.
AI will resolve many complex global challenges including healthcare for the masses, agriculture, and education. AI will analyse public health data to predict outbreaks and help hospitals to optimize resource allocation. AI will monitor crop health, optimize irrigation, and reduce waste, improving yields sustainably. AI systems will provide personalized learning experiences to bridge gaps in education access worldwide.AI will increasingly help improve citizen service delivery. AI is expected to play a significant role in the analysis of large and complex datasets and provide insights based on the data for evidence-based policymaking.
Responsible and ethical AI
We expect a stronger focus on ethical AI frameworks in 2025. This will help to ensure that AI systems are more transparent, have less bias and are accountable. There will be greater emphasis on making the systems protect individuals’ privacy, prevent misuse of data, and make the systems more explainable.
Most governments will implement localized AI governance frameworks for responsible and ethical AI. While we expect movement towards a common regulatory environment for AI to begin, it will take a few more years to build consensus on a global framework.
Conclusion
Overall, 2025 holds great promise in AI across multiple domains. With advancements in Agentic AI, multimodal systems, and GenAI, coupled with a strong focus on ethical practices and energy efficiency, AI will become an integral part of everyday life, driving innovation while addressing critical societal challenges.
Fonte: DataQuest
