How AI can reduce energy footprints in many business sectors
Francis Bach, a researcher at INRIA (the French National Institute for Research in Digital Science and Technology) and specialist in machine learning, looks back at the main successes of artificial intelligence, and discusses the limitations of this technology.
Could you give us an example of an AI application you think is successful or promising?
Currently, the highest profile application of Artificial Intelligence is in personalised advertising and marketing. That particular application may be commercially successful, but it is not necessarily a triumph. It also raises issues around its use of users’ personal data, often without their knowledge. Other successful applications include image recognition and machine translation. AI also offers very real potential in medicine and healthcare. Nevertheless, there is still a significant gap between the technique, which is often well developed, and its use in treating patients, which is still in its infancy. Issues around data access are also central to the widespread use of AI in medical applications, so the key challenge today is finding a way to provide open access to health data without compromising patient confidentiality and privacy.
Conversely, could you give us some examples of AI applications you think are impossible to achieve within a reasonable time frame?
The best example would be self-driving cars. It will be a long time before the promise of autonomous cars sharing the same roads as conventional cars materialises. Right now, AI only works in advanced driver assistance systems, and full autonomy is not a serious expectation in the short- to medium-term. To operate on the basis of machine learning, vehicles would need a much larger volume of data.
What are the major challenges facing AI in terms of energy consumption?
AI is often singled out when it comes to energy consumption, but it is important to put these things in context. This technology is one part of the digital sciences, which are together responsible for around 4 to 5% of greenhouse gas emissions. Therefore, the contribution of AI within that total is currently quite small. In addition to that, work is now being done to reduce the significant costs generated by the AI learning and implementation phases. On a more positive note, AI has the potential to reduce energy footprints in many business sectors by measuring environmental impacts and managing industrial processes like logistics much more efficiently. AI may not be the magic bullet, but it can contribute to combating climate change.
What role do you see AI playing in the transformation of jobs?
AI is often characterised as a general-purpose technology in the same way as electricity, because it impacts every aspect of industry. In the past, these technologies have tended to generate growth. Although jobs have changed, the total number of jobs has not reduced. Some zero-credibility studies have tried to make people believe that AI will replace humans but for the time being Artificial Intelligence operates mainly as a tool for decision-making, and then only in terms of computerised data.