How to stop AI going the social media way | Your car is tattling on you
CyberInsights #134 - A visionary points out the challenges of monetising AI | Driving risk scores are derived from people's data being shared directly by the manufacturer
CyberInsights will not be available for the next 2 weeks. I am travelling and plan to be off the grid (if that is at all possible)! Will probably update you with happenings and stories one I return. Expect the next edition on the 10th of April.
What happens when AI, like social media, wants to run ads for monetisation?
It has the potential to make all the same mistakes as social media
This is a post by Bruce Schneier. Regular readers will know that I quote Bruce’s blog often. This one is a deep dive on the problems of social media and how they might repeat themselves with the AI industry when it goes on a monetisation spree. [LINK]
Bruce identifies 5 areas of social media that have led to its current state where it may hard society. Here is a summary of his post. I urge you to read the entire post.
Advertising - targeted advertising to be precise. Currently it is easy to skip the ads on websites. AI will be able to generate ads on the fly that will have their copy (the text of the advertising) in line with what is being searched. Scary, eh? What if LLMs like ChatGPT are monetised? How would we know if an answer is the correct answer, or is it an advertisement?
Surveillance - A survey shows that each user has over 2200 companies tracking her. Imagine being tracked for the questions you ask LLMs and chatbots!
Virality - Bad news travels fast. AI can be used to generate lots of stories - platitudes, semi-truths, falsehoods, etc. Spin Doctors can use AI to spin any story and get it viral. Internet connected LLMs will learn this soon and you will not know if the AI is hallucinating, or is merely learning from the wrong sources.
Lock-In - Image that we have created an AI powered digital assistants on an AI platform. We have spent years training that AI to work for us and be the most efficient. It does things exactly as we want. What if we we now want to change our provider? Can we?
Monopolisation - If big tech and big VC funding is used to build something of scale, will it be open to all? Or will the goal be, again, to have the largest user base that cannot switch?
Take Action:
Bruce has compared AI with the problems of social media.
If your job involves AI ethics - this article is something that you should read, understand and implement controls to prevent any of this from happening.
If you are a cybersecurity professional, and if you have not already, start building your skills on managing AI risks.
Your insurer knows how you drive…
...and it’s your car manufacturer selling.
Six months ago, I wrote that your car is selling your data:
Now we know to whom.
This article in the New York Times [LINK] talks of how car manufacturers are selling user driving data with risk scorers like Lexis Nexis.
General Motors, Kia, Subaru, Mitsubishi have shared data with Lexis Nexis for generating a risk score that can be used by insurers. Ford, Hyundai and Honda also seem to be selling!
It’s reasonable for insurers to know more of your behaviour to give you insurance. If an insurer wants to know that I walk 10k steps a day and wants to offer me a discount, I would grab it. If an insurer wants to know that I am a safe driver and wants to offer me a discount, I would grab it. However, I would never want the insurer monitoring me surreptitiously. And I would never want, in either case, the insurer to know the locations that I have been to or the latitude and longitude.
Take Action:
If you are an auto manufacturer, then please get your head screwed on the right way and stop selling driving data. If you want a few million on the side, please have a very clear consent form with the vehicle owner and mention exactly what data you collect. Limit the data to only the data that is required by the insurer to make an underwriting decision.
If you are an insurer, please ensure that the data that is used to give you easy risk scores is sourced from a fair and consented source.
If you are the owner of a connected car, read the terms and conditions! If you do not agree, turn off the relevant feature in your car. Visit the website mentioned in the article to check if your car is selling your data for risk scoring.
Someone shared this with you? Share it forward!