Smart Software: Bureaucrats Race the 20 Somethings
May 11, 2022
One of the ArnoldIT team is delivering a talk about smart software. One of our points is that AI is moving along the innovation curve and bright young sprouts are installing “smart” software in interesting applications.
The European Union is nervous about smart software. Johnny on the spot is that bureaucratic outfit.
One of the most common tropes in science-fiction is a computer or a robot achieving sentience. In the stories, technology gains sentience in a variety of ways: lightning strikes, digital evolution, alien intervention, etc. While some of the stories end on a positive note, many end with a warning message to humanity: Don’t play God. Tech Radar explains that the European Union is taking preventive measures: “New EU Rules Would Allow It To Shut Down Ai Before It Got Dangerous.”
The European Union has worked on an AI regulation framework since March 2018 as part of its Digital Decade regulations. Work on AI regulations has been slow because the EU has focused on the Digital Services Act and the Digital Markets Act that manage how much power American tech companies can have.
The EU AI Act is also undergoing a review and critique phase through the Ada Lovelace Institute, an independent research facility that works on data policy. The Ada Lovelace Institute scrutinizes the AI Act:
“The full report (via TechCrunch) includes a lot of detail on the pros and cons of the regulation, which is a global first, with the main takeaway is that the EU is setting itself up to have some pretty powerful tools at its disposal. The EU plans to create and empower oversight bodies that can, theoretically, order the withdrawal of an AI system that is deemed high risk, before requiring the model be retrained. The draft AI Act has been under a lot of scrutiny – and has received a fair amount of criticism – and will likely still fall short of the EU’s most expansive goals: creating the conditions for “trustworthy” and “human-centric” AI.”
The current EU AI Act needs to be revised, but that does not mean it is a failure. The act is a good beginning to creating a viable framework to govern AI.
Our fearless but quite aged leader (Stephen E Arnold) believes that it may be difficult to regulate smart software, especially in the United States where big tech companies are influential in the economy and politics. AI often thrives in powerful black boxes that are inordinately programmed with ethnic and socioeconomic biases. Developers have yet to remove these biases because at the Google there may be zero biases or the datasets are synthetic. (Yep, that means what you think it means: Statistical confections and close enough for horseshoes outputs.)
Can the EU set the standard for how AI is regulated across the globe, a bit like kicking the Russian oil and natural gas habits? Worth watching… from a distance.
Whitney Grace, May 11, 2022
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
