In addition to the future risks underlying general artificial intelligence (AI) and superintelligent AIs, there are current risks that are little discussed, and deserve to be.
Artificial intelligence (AI) companies have been accused of infringement of intellectual property rights. There is widespread unauthorized use of data. Additionally, leading AI companies have been sued for using copyrighted material to train their AI models without permission or compensation (see lawsuits from the Toronto Star and the Canadian Broadcasting Corporation). There are also related security risks; Generative AI models can expose companies to security risks, such as unauthorized access to sensitive data and manipulation of open and broad language models (LLM). There are still legal challenges – current legislation does not explicitly contemplate the authorship of works created using AI.
However, the most urgent thing is to adopt policies that make AI companies respect intellectual property rights. In the meantime, it is important to increase users’ knowledge of generative AI chatbots (like ChatGPT) and alternative chatbots that place an emphasis on ethics, transparency and consent (e.g.: Bloom).
Another of the current risks of AI is the significant increase in carbon glued that AI companies are creating, due to the high energy consumption required to power data centers and AI systems (c. 2% of global energy and 2.5% to 3.7% of global greenhouse gas (GHG) emissions which, in most cases, still comes from the burning of fossil fuels. What is the size of the carbon footprint of AI companies? According to MetaAI, training an AI model can emit up to 284 tons of CO₂. It also states that Microsoft has recorded a 30% increase in GHG emissions since 2020, while Google has recorded a 50% increase in 2023 compared to 2019 (resulting from the development of AI). Interestingly, it omits data regarding the Meta group… The annual demand for electricity generated by NVIDIA’s AI servers around the world is expected to increase from 85.4 to 134 terawatt hours by 2027.