Don’t Feed the Amoeba

Don’t feed the bears…? Nope. Now it’s don’t feed the amoeba.

It’s a good idea to know what (big data) you are feeding your machine learning.

It’s also a good idea to know what it’s eating that you didn’t plan on feeding it.

Bears may be omnivores, but there are things bears will not eat. You don’t feed the bears because bears will learn that you have easy food and will keep returning. Most of us don’t want to collect and use what comes out of a bear.

Amoeba don’t care. Amoeba will ingest anything. Amoeba don’t stay in easy food areas. Amoeba are too primitive to care.

Deepfake and chat artificial intelligence are too primitive to care. They will ingest what you feed them. They will ingest whatever they find on their own. Why do we want to enthusiastically collect and use what comes out of deepfake and chat artificial intelligence?

A woman received a message from a friend “Hey, is this you?” The face was indeed hers, but used for a camping stove advertisement that she had never agreed to, and didn’t know existed.

How? Sleuthing discovered: 74 reposts of the picture in online marketplaces, selling a variety of products from a camp stove to a portable gas cooker 1 linchpin photo with a different woman camping 1 original photo for a camping tent from 2018 and woman model

The more the image was posted, the more the image changed to look like the woman. The algorithm keeps learning, moving toward the desired features and look.

Vint Cerf, inventor of Internet architecture and many foundational protocols, warns not to jump into using chat artificial intelligence just because it’s the hot new thing . Vint, the Internet Evangelist for Google and many, many other organization boards and standards committees, would certainly know. He advises “Be thoughtful about how we use these technologies.” They don’t know the difference between eloquent and accurate responses. They may produce convincing but fictitious answers.

If you can control its diet, it becomes a much more useful tool. Right now there are ethical questions about both how we use the tool, and the output of the tool, that need alot of careful thought.