I am a VERY amateur woodworker. I have made a coffee table and a built-in for our refrigerator. It sounds more impressive than it is. The one thing they stress in woodworking, especially furniture making, is making sure that all your equipment is square, giving you a perfect 90-degree angle.
A cut that is not 90 degrees is not a huge problem. You can overcome it. But continued cuts will keep compounding to make a disaster. And usually, once you notice it is usually too late.
Ask me how I know…
Look up the 5-cut method for making sure your table saw is square if you are curious.
How Can Artificial Intelligence Be Biased?
Artificial intelligence (AI) is only as good as the data we feed it. Though machines themselves don’t have biases, the data we feed them can.
For example, a big issue in research is what is called external validity. External validity means how well you can use the findings of the study to apply to a broader population or situation.
For example, a large study conducted in the United States regarding heart disease where the primary population is elderly white men has great external validity for elderly white men in Canada maybe even Europe but does not have good external validity for elderly Asian men living in America.
So if we are feeding this data to an AI platform then you could see how biases can be introduced to the system. When we are seeing individual patients these biases don’t always cause issues at mass scales but with AI it is now automated and out of our consciousness.
The best way we can combat this is by getting as diverse of a data set as we can. That is not a small task and is extremely difficult. Many scientific studies don’t have great external validity to the minority population. There are many reasons for this. Some of it has to do with the lack of access to clinical trials. There are some interesting things being done in the remote clinical trial space to combat some of this inequity.
Artificial Intelligence In Healthcare
With all that being said there are so many places where AI can be utilized.
AI could be used to:
Automating administrative tasks such as billing, scheduling, and writing notes
Be used to make decisions that are clearly black and white
Radiology and Pathology (being used in some capacity already)
The possibilities are endless with AI
This is not to say that we should not work on AI models. It just means we have to be cognizant of what kind of data we are building these models upon. Because the last thing we need is biases being executed at scale.
Every week I share a random fact about myself. This week’s fact is…
Random fact about me: The Chicago Bulls are my favorite sports team
Thank you so much for your support! Please do not hesitate to reach out if you have any questions or leave a comment.
If you enjoyed what you read please consider subscribing and sharing!
I hope you have a great day!