The IEEE Trusted Data and AIS Playbook for Financial Services (IEEE Finance Playbook)
There are lots of people talking about the ethical aspects of AI nowadays. It’s easy to predict that there will be a lot more in the future. AI technology had a fierce buzz about it in 2019, but unlike many other technology fashions, AI seems to be delivering a continuous wave of research breakthroughs and innovations.
The amazing results announced by Google Deepmind (1) in predicting protein structures using their Alphafold-2 algorithm have demonstrated that AI is now at the core of breakthrough science – AI is helping humans create important new knowledge that arguably would not have been available to us without assistance (2). On the other hand, arresting improvements in natural language generation and understanding created by OpenAI’s GPT-3 (3) and Dall-e (4) are opening a window onto our own cognition and capabilities (5).
As the potential of AI technology grows, so do the impacts on everyday life. Year by year, AI becomes more inescapable. Constructing blogs relies on Google, which uses a transformer (a trendy AI algorithm) powered engine (6). While the blog is written, the music that blots out the outside world and focuses the writer’s mind is selected by algorithms tuned to the author’s taste and the taste and selections made by others judged to be like the author. The blog (once written) is served up to readers by social engines fighting for attention and engagement. Blogs are a tiny aspect of modern life, but shopping, entertainment, policing, transport, news, education and finance are all just as interwoven by AI algorithms.
So, it’s heartening to read the IEEE’s Trusted Data and AIS Ethics Playbook for Financial Services (disclosure – I provided input and feedback on it during development). There are important features to this particular effort that I think should be highlighted – and I think these features make it particularly useful as well.
The first thing to note is that the Playbook has grown out of a community; it’s been written with a broad perspective informed by many hard-won experiences. This comes through in the next noteworthy aspect (that the playbook covers all aspects of AI systems development) but as well as being obvious in defining the structure of the work, it has also clearly informed the implementation. Throughout the playbook, there are references to specific tools and experiences assembled from the combined experience of the panel that wrote it.
Secondly, the playbook covers the technology aspect of AI but also focuses on the people and process in play in any AI systems development process. These aspects are often neglected by solo efforts at defining ethical approaches to AI, but the IEEE community’s wide experience allows the whole lifecycle to be addressed.
Thirdly, the playbook provides practical tools, from the top 20 high value use cases to be considered – to the aspects of ethical concern that appear in each of these use cases. The playbook brings together the ALTAI (EU Assessment List for Trustworthy Artificial Intelligence) with 20 key resources that a practitioner can make use of in dealing with any of the 7 ALTAI elements.
Although organisations like the OECD have made great strides towards creating an informed debate about AI in civil society (7), the IEEE is helping system integrators and developers to address the ethical challenges in the systems that they implement now. As governments such as the EU rapidly move towards proscription and regulation of AI systems in particular domains and contexts (8), these kinds of practical tools are urgently needed.
Download the playbook here!