Skip to content

Top 5 trends in AI for 2023

Artificial intelligence (AI) is a fast-moving field showing no signs of slowing down in 2023. As new technologies emerge daily, it is crucial to analyze and be aware of the latest trends as they will influence the tools available to most developers and shape the course of action. According to VentureBeat, AI innovation is predicted to continue to grow as companies are expected to adjust budgets to incorporate this technology into their roadmap or increase the current investment in AI.

For example, deep neural networks (DNNs) were prevalent in 2017 and 2018, but from 2019 onwards, there was more research into other types of AI, such as generative adversarial networks (GANs), which attempt to replicate human intelligence by having two competing algorithms create data without prior knowledge about how the data should be looked at.

In this post, I’m sharing my personal preference and predictions on what I believe to be the 5 key areas that will have the most focus, either from the research and from the industry itself.

 

Data-Centric AI

The AI community has been shifting from the traditional focus on AI models to a focus on the data used to train algorithms, becoming more important to improve the quality of your data than the model itself. To start, understanding, if you have enough data for training ML models, can help save time. Optimizing models can be a waste of time if they are trained on too few examples or use insufficient features or poor data quality. The Data-Centric AI Community defines Data-Centric AI as the process of building and testing AI systems by focusing on data-centric operations (i.e. cleaning, cleansing, pre-processing, balancing, augmentation) rather than model-centric operations (i.e. hyper-parameters selection, architectural changes).

Synthetic Data

The second major trend to keep an eye on is synthetic data. Synthetic data is a way of creating realistic fake data, which can then be used to augment datasets, train AI algorithms and be shared freely. Synthetic data mirrors the attributes and behavior of real data, giving control and granularity of data to data scientists. This powerful tool allows companies to overcome privacy issues, improve data quality, and scale existing data. According to Gartner, by 2030 it is estimated that synthetic data will completely overshadow real data in AI models.

 

Banking transactions generated synthetically (Photo by Mika Baumeister on Unsplash)

Responsible AI

AI is a powerful tool and can help achieve great things for humanity. However, there are some questions raised about fairness, privacy, ethics, governance, and legal. Responsible AI appears as a movement that advocates, raises awareness and creates solutions to mitigate the negative effects of artificial intelligence, as only 35% of global consumers trust how AI is being implemented by organizations. This emerging technology will see a boost of importance in 2023, especially in the ethics and democratization sectors. To safely scale AI solutions, Responsible AI is going to take the central stage for companies that wish to continue to invest in AI solutions. In November of this year, at WebSummit, it was announced the world’s largest consortium on Responsible Artificial Intelligence, where YData is the only company in the consortium entirely focused on data.

Fairness and Bias in Machine Learning

If you are oblivious of AI Fairness, you seriously should update your news record. Bias refers to how our human biases affect the data we collect and use, as well as the algorithms we train on datasets — and how unfair that might seem when applied to new scenarios. For example, imagine you want to develop an algorithm for recommending books based on user preferences; if you use historical data from Amazon customers who live near each other, then your model will recommend books similar in genre or style to one’s people have purchased before, even if they aren’t necessarily as good as other options (because those books have already been chosen by someone else). This leads us into another important concept: fairness metrics for machine learning models (fairness metrics measure whether or not an algorithm treats members of different groups equally). In 2023, companies will be focused on optimizing Machine Learning algorithms, especially to identify, measure, and improve fairness in classifying tasks.

Generative AI

Generative AI is a new field of machine learning that aims to create new data, such as images and music. It’s the umbrella term of Synthetic Data, which was largely popularized with the DALLE and Stable Diffusion models. Large Language Models and Foundational Models are disrupting the content creation space and the internet is loving it.

Applications that use these models are going viral: creating avatars from our pictures, generating images of impossible scenarios, finding tips for home decor, and much more.

Conclusion

Artificial Intelligence is a consolidated driver of innovation worldwide. By 2030, the global artificial intelligence (AI) market size is expected to hit US$ 1,597.1 billion. So, it’s forecasted an increase in the adoption rate of artificial intelligence solutions. In what concerns the applications of AI, becoming more responsible will be the highlight of the next year, and it will be reflected in several AI specializations that will culminate in new tools designed specifically to overcome this negative effect. On the other side, it is expected to see much buzz around new applications for Synthetic Data and Generative models.