From 9 to 11 February 2023, HPA participated in the World Official Intelligence Cannes Festival (WAICF23). Here are a collection of thoughts and moments from the red carpet.
Small pieces of AI are everywhere.
Perhaps we are not always aware of Artificial Intelligence, but it’s already improving the quality of our life and will become more pervasive day after day. For example, health, security, finance, production and logistics, traffic control systems, entertainment, and many more sectors use daily forecasting, recommendation engines, computer vision algorithms, and natural language processing models to transform data into useful information. As users, we must be aware and critical to objectively evaluate the progress of AI and understand its impact on our daily lives. As companies, we must commit ourselves to develop and bring to the market solutions based on AI that is, in addition to efficient and performing, transparent, and in line with human ethical values.
Increasingly only a few tech companies with huge investments can create disruptive innovation in the AI field, starting from the research papers. This is because the best ML algorithms are useless without massive datasets, tons of parameters, and extensive training, requiring billions of investments. The good news is that — even if only the more prominent players have the right money to lead the research — small and medium AI firms can create today faster and cheaper than ever innovative services and valuable products for their clients.
Is AI sustainable?
AI Artificial Intelligence can do a lot for sustainability, but how much is AI sustainable? For example, huge ML models create a growing large carbon footprint, so it’s necessary to develop new practices to reduce energy consumption by selecting efficient ML model architectures, using processors optimized for ML training, and increasing the efficiency of data centers.
Transformers are shaking up AI
We are not talking about Optimus Prime and Bumblebee but about one of the main highlights of advances in deep learning and deep neural networks becoming famous thanks to the worldwide success and fast adoption of ChatGPT by the general public.
Transformers are a type of neural network architecture introduced in 2017 in a paper titled “Attention is All You Need” by Google researchers and has become very popular in Natural Language Processing (NLP), leading to numerous improvements in applications such as machine translation, text generation, and question answering.
They use a mechanism called “self-attention” to process input information. Instead of traditional convolutional or recurrent neural networks, transformers use a series of self-attention layers, where each layer processes input information autonomously, allowing to process of information in a very efficient and parallel way and making it possible to manage sequences of variable length with much less computational effort than traditional neural networks.
On the other hand, Transformers are huge and complex and require significant computational resources to train and operate. Therefore, transformers also require large amounts of training data to function correctly. If insufficient data is unavailable, their ability to generalize may be impaired.
Because of their complexity, transformers can be difficult to interpret so it could be challenging to understand precisely how they achieved their results, which could be a problem in some contexts. In addition, like any machine learning model, Transformers can be affected by bias in the training data leading to unfair or discriminatory results, especially if the training set is not representative of the population to which the model will be applied.
Currently, Transformer models can understand the meaning of words and phrases in a given context but are unaware of physical reality or events outside the input data they receive.
Companies need AI to improve efficiency and reduce costs, and people are eager to use AI applications. Still, both are increasingly aware of the transparency and responsibility of AI models, so it’s necessary to improve techniques that can explain in the best way possible how a specific decision was taken, showing the main steps followed by the ML model.
Let’s talk about business!
Adopting Artificial Intelligence in your company can create a massive competitive advantage, but investing in cutting-edge technologies and state-of-the-art algorithms is not enough. AI firms have to align their language to the business language of the companies cause everything has to be evaluated in ROI terms.
There is so much to do, and we can’t wait to make ourselves available to companies to support them. Whether you are ready to start your AI project or need help defining your AI strategy, we can help you.
👉 Ask us: info [at] hpa.ai
HPA — High Performance Analytics is an accredited spinoff of the University of Verona. Since 2017 has been designing and developing custom Artificial Intelligence solutions in five areas of expertise: predictive analysis, anomaly detection, constrained optimization, text analysis (NLP), and image recognition. With significant experience and successful case studies in numerous industrial sectors (energy, manufacturing, transport, real estate, CRM), HPA offers SMEs and large companies effective and efficient solutions based on a deep knowledge of mathematical and statistical models gained over 20 years of academic research. Since 2021 HPA has been A.I. Terranova Software Competence Center.