10 Reflections on AI

The first milestone of #500daysofAI in 2019–2020

In the last 10 days I have been writing one article every day. This article is the tenth and I will do my best to summarise what I have come to learn throughout these days.

  1. When the two suitcase words artificial and intelligence combines with two other such words as sustainability and city it further increases the risk of (mis)interpretations. There are many interesting ways to use particularly machine learning techniques for planning, communication and infrastructure. Yet it requires the understanding of a variety of actors.
  2. Big technology companies developing technology seems to fund ethics institutes related to AI leading to self-policing and possible conflict of interest. Ethics or moral philosophy, defining what is right or wrong in regards to AI. Ethics institutes related to artificial intelligence or technology has appeared, and this is great due to lacking regulations.
  3. AI-related patents is the patent category within technology that has grown most rapidly between 2013–2016. The most common technique is machine learning. The most frequent combination is deep learning and computer vision (mentioned in 49% of all patents). Companies from the US, Japan and China are dominating AI patenting.
  4. Alongside the developments in AI we are facing a climate crisis. We have to seriously question the energy and resource use of this field despite its possible benefits. There is AI for good and AI for bad, and of course simultaneously not that easy to simply say there are two extremes. Training a single AI model can emit as much carbon as five cars in their lifetimes, and there is recently large investments in the private defence industry in the US. We must acknowledge that this can exacerbate the current problems we are facing rather than solving them if we are not careful.
  5. Governments are using AI, however this gives rise to further needs for transparency and sharing models or algorithms so decision-making supported by technology can be audited. Reading a case study of four projects in Argentina and Uruguay the conclusion was that for robust democracy people must understand how the public sector works. There are many possible ways to make mistakes predicting areas such as policing, schooling, pregnancies and investments.
  6. There are som excellent writers in the field of AI. Karen Hao is excellent at communicating research and why it is relevant. Norvig and Russel’s seminal work Artificial Intelligence: a Modern Approach certainly needs a deep dive to skim the surface of this vast field of knowledge. I need to understand this technology far better than I do so if you have any recommendations comment on this article or further down here.
  7. Countries around the world are developing strategies for AI. On that note I decided to explore and summarise the current strategies for AI in Scandinavia anno 2019. This surely will provide for further writing going forward, and may be important to read to understand how it will affect you or how you can contribute with your opinions.
  8. There is a lack of focus on diverse range of social sciences in the subfield of AI Safety. Due to its applications within such a wide range of areas in the world we have to continuously question our notions of progress or oppositions in the field of AI.
  9. There is not a lack of topics within the field of AI left to explore. Once you start exploring you understand that there are so many perspectives and so many different areas within the field of AI that can be written about. I would however like for the next 10 days to focus slightly more on programming in R or Python, so let’s see what happens.

AI Policy and Ethics at www.nora.ai. Student at University of Copenhagen MSc in Social Data Science. All views are my own. twitter.com/AlexMoltzau