The New AI

This new generative artificial intelligence (AI) has captured our imaginations, and it seems to be something that can change the world.  Sam Altman, CEO of OpenAI, called this new development of AI a “printing press moment”.  He believes that just as the Gutenberg printing press went on to change the world over 500 years ago, these recent AI developments will do the same.

What is this generative artificial intelligence and what can it do?  The word generative means that the AI generates or creates the results.  It does not simply regurgitate what it found.  It is based on a generative pre-trained transformer (GPT) and a large language model (LLM) that uses transform layers with neural networks to parse the input in order to produce new output.  Note that it is a large language model.  In order to train the model there are enormous amounts of data (mostly from the internet)  that are used to make the connections and build the relationships between input and output.  Also note that it is also a large language model.  The model translates from one “language” to another though today the word “language” is used in a very broad sense.  The original 2017 landmark paper was about efficiently and effectively translating English to French or German, but today AI is used to translate a brain scan to an image of what the person saw or to build antibody proteins from a set of characteristics. The input and/or output “language” may not be what we would normally consider to be a language.

The model is trained to produce probabilistic results from the inputs it is trained on.  That means no result is right or wrong, rather the result is more likely and not less likely based on the training data.  The results might appear to be new and unexpected, but they are results derived from the training data with the neural networks. Yes, there are discoveries as a result of this AI that can improve our lives but I believe this AI at best mimics human intelligence.  There are limitations to the AI.  It is only as good as the data it is trained on.  AI does make mistakes.  There are known biases and issues with AI, but in general it is doing an amazing job and it is improving.  In fact it has become much more of a general tool creating a variety of results and answering a variety of questions. (That is very different from the narrow focus that was found in the landmark 2017 paper.)

Will this new tool change society?  I believe it will be like the printing press and society will change.  Will we speak into our phones and get an answer to whatever question we may have? I think it is coming to us in the next few years.  Will AI become our new authority? Unfortunately, I believe for many people AI will be the new authority.  Will it speak the truth?  It cannot speak the truth.  It will give only answers that the training data will consider to be most likely.

As a confessional Lutheran Christian I wonder about the results of religious questions since most of the training data comes from the internet with human intervention and guidance.  Lutherans in general have not embraced the internet like some other Christian traditions.  (There are some exceptions.) For most Christian questions it is not likely to be a big deal, but if the answer to your question is a most likely result of a probabilistic determination then it is likely Lutheran thinking would not play a direct role in the resulting answer.  The result will likely be based on what appears the most on the internet.  The bad news is there will be a bias. The good news is likely the wacko thinking that you can find out on the internet would not likely appear in the answer.

Christians already are making use of AI.  If you want to create a Bible study on a chapter in the Bible, try OpenBible.info’s AI-assisted Bible Study. If you want a Bible chapter summary or answers to questions on a chapter in the Bible, try IlluminateBible.com.  If you want a biblical principles based answer to a question, try Bible.AI.  If you want to search for related Bible verses based on a phrase, try SiliconScripture.org.  Note all these websites are under development and may not produce accurate answers, but that is true for all AI results.

I consider this post to be the second post of what will be an ongoing series of posts on AI.  Check out the first post for a broad introduction to AI with some stated concerns.