2023 here we come!
We wish you all a Disruptive New Year!
Article 4 - January 2, 2023
Article 4 - January 2, 2023
We are only interested in the future because we will spend the rest of our lives there!
We wish you all a Disruptive New Year!
Thank you for being part on this great AxonJay journey!
Last year was a rollercoaster for us and we enjoyed every second of it. Looking back, the last part of the year was very successful!
Our major milestones from 2022 were:
Since 2012, the field of artificial intelligence has reported remarkable progress in a broad range of capabilities including object recognition, game-playing, and machine translation. Thanks to these advances, AI went from being the cryptic technology of sci-fi movies to one of the most important fields in the modern economy, revolutionizing both companies and everyday life.
The rise of artificial intelligence solutions has also brought some concerns about their impact on the environment. Unlike other fields, the danger for the environment does not come from something concrete and tangible like pollutant agents, but from the very software used by AI. Indeed, most of the recent AI achievements have been obtained by increasingly large and computationally-intensive deep-learning algorithms. These algorithms are based on deep neural networks that learn how to solve very complex
problems and may take days to be trained in those tasks.
Some articles have estimated that the computational effort for training state-of-the-art models is growing exponentially in time without achieving a similar trend in performance. If you are not used to AI, you may be tempted to think that AI projects are just like other IT projects with data-scientists and data-engineers instead of developers. However, the reality is quite different. AI projects are intrinsically different from IT ones because of the costs of the infrastructures needed and because of their R&D nature. Indeed, every AI solution is built after an R & D phase during which experts have to understand which data and algorithms can be used and test several ways to tackle the issue. Besides, data has to be gathered, stored, processed by complex algorithms that may take months or even years to be developed. Analytics Insight claims that in 2021 the overall cost of an AI solution may vary from 20k$ to up to 1M$. Many companies may think about outsourcing the creation of their solution to another company that already has the know-how and the experience or buying an off-the-shelf product, but in this case, the cost may quickly escalate. A pre-built chatbot costs up to 40€k per year (AI Pricing), while a single AI consultant easily charges (Machine Learning Consulting Rates) 200-300 $/hour.
In recent years, the attention has been finally driven towards efficiency and cost reduction with the term Green AI appearing for the first time in an article in 2019. It refers to AI research that “yields novel results without increasing computational cost, and ideally reducing it”. Whereas Red AI has resulted in rapidly escalating computational (and thus carbon) costs, Green AI has the opposite effect since it aims at improving efficiency rather than performance.
Following this principle, we mostly avoid deep-learning algorithms except when we are recycling publicly available pre-trained models like those shared by OpenAI or Hugging Face which are both focusing on efficiency and reusability. Being aware of the cost of gathering and storing data, we have decided to reduce data redundancy in all possible ways to have a lighter infrastructure and reduced costs.
This vision has driven us towards looking for signals rather than data. We do not want to gather all the data by brute force, we rather prefer to detect when changes happen which requires less storage capabilities. In a similar way, we do not simply scrape thousands of webpages from top to bottom, we rather try to extract only the main information from a webpage reducing the cost and time of this process. Furthermore, we mainly gather data only on the clients' demand.
Creating sustainable AI is not just about having smart technical tools, it is also about using AI for environmental purposes, for instance by improving industries efficiency and therefore reducing carbon emissions of companies (article). An article on Green cloud storage points out that cloud storage data centres are designed to store data efficiently reducing the number of duplicates and being, therefore, more efficient and less expensive. A study shows Cloud Computing Can Cut Carbon Emissions by 30% to 90% helping companies save money and reducing their carbon footprint. Within our Self-ML Platform™ we have developed tools to monitor our energy usage so that we can always monitor the efficiency of our solutions. Furthermore, our products are based on reusing data for different solutions and clients which enables us to further improve our data efficiency as well as our scalability.
At AxonJay our mission is to democratize AI by developing AI-based solutions accessible to all the companies. By applying Green AI principles, not only we have developed cost-effective solutions, but also we have ensured the sustainability of them. Our Self-Machine-Learning Platform™ is a PaaS Business model, which was designed to be a fast and scalable application and re-usable model in different industries. Despite this common setting each client will take advantage of a tailored platform thanks to our automatic feedback-loop. Our disruptive business model along with our innovative AI approaches will revolutionize the use of AI in many sectors making it greener than most existing solutions.
2023 looks very promising and it will be even a bigger roller coaster.
And we are starting a Disruptive Seed Round during the summer of 2023.
Keep on moving! Growth and success are around the corner wherever you expect them the least.
Happy Disruptive holidays!