I know just enough computer code to sort of copy and paste sections from the internet to do any programming that I need to do. When I get stuck, I ask my son for some help and the two of us have this joke about how the code ‘feels’ right to me. It feels like it should work but it doesn’t. Terry’s very firm, very analytical and he tells me that code does not care how I feel. It’s computer code. It simply works, or it does not.
Here is the interesting thing. Computer code can actually be programmed to consider how you feel. It can be programmed to give you answers based on your input to it. Those answers are generally programmed to make you feel good or feel some emotion about it. For example, you might type into a computer, ‘I had a bad day’. The computer can be programmed to understand the human notion of day. A period of time. And bad, a period of time that the human felt bad. The computer may parrot back to you, “I’m sorry to hear that you had a bad day. What was bad about it?”.
Of late, AI or artificial intelligence has really, really been swamping the news. The current version that’s taking the entire world by storm at the moment goes by the really catchy name of ChatGPT. Currently, the public at large has access to ChatGPT through a very simple interface. It’s a dark webpage with a box encouraging you to type things into it. If you do, it really doesn’t take very long and you actually get quite engaged with this thing, going backwards and forwards and having quite the conversation with it.
How does artificial intelligence work? What are the implications about it? What has this possibly got to do with our faith in God the Father?
I want to begin with how AI works. We have to teach the computer like, ‘I had a bad day’. We have to teach it about human language and about context and syntext. I’m over-simplifying a little bit, but you start with some basic statements. The cat sat on the mat. The dog sat on the floor. Once you teach the computer that about where the cat is sitting and where the dog is sitting, you then encourage it by taking out words. The cat sat on the blank. The dog sat on the blank. The code very quickly understands what you are asking and parrots it back to you. Cat on mat, dog on floor.
We get that set up and then we take out the other thing. We take out the blank sat on the mat. The blank sat on the floor. Then we ask the computer which did what sitting where. It comes back and generally gets it right. At that stage, we’re pretty confident that we’ve got a basic model that understands human language and we set it loose on a very small piece of the internet. A very select piece. Something like Wikipedia where there is a lot done. It comes back and then we get to ask it a question and we say “blank sat on the mat” because we want to see what it learned. If you’ve spent any time on the internet you know you get some crazy answers. The computer confidently comes back and says, “The elephant sat on the mat”. Technically it might be right, but it just doesn’t feel right. Does it?
Now you come to a really important phase of creating your artificial intelligence called reinforcement learning from human feedback. We actually pay thousands of people an hourly wage to sit in front of a computer and the computer spits back these sorts of results, cat or mat, and the human gets to either thumbs up or thumbs down. Cat on mat, thumbs up. Elephant on mat, thumbs down. You iterate, and iterate, and iterate and as you’re doing that, the computer is learning what responses the human likes and what responses the human doesn’t like.
There is no bottomless pit of money here. We can’t pay these people, thousands of people, this hourly wage forever. At some point, we have to say, “You know what, we’ve tuned enough. It’s got the idea of what we’re looking for. We’re going to cut it loose.” Then we let it go on the internet.
These are some phrases you may hear as you read news articles about this. It’s called a large language model. LLM for short. The key take away is that it’s humanly reinforced learning about the language and then set loose not only on the internet at large, but the public at large. It’s very complex. I have oversimplified it a bit, but you teach it about language and then you human reinforce that learning about it. The key here, is obviously, that it is human written computer code running on human built computers that has been fed a sample of human words and then fine tuned by humans and of course, set loose to read human knowledge on the internet.
How good is it? Let’s give us all an example of something we that we might be a little bit more familiar with. Here’s a slightly biblical example. One of my male friends asked ChatGPT, this artificial intelligence, the following. What happened to the ten lost tribes of Israel? Bear in mind what I’m about to read, I’m going to quote word for word what came from an artificial intelligence.
“The exact fate of the so-called lost tribes of Israel is not known for certain. The Bible mentions the lost tribes in several passage including 2 Kings 17:6 which describes the Syrian captivity and in James 1:1 which addresses the twelve tribes scattered among the nations. In Jewish tradition, the lost tribes are believed to eventually return and be united with the rest of the Jewish people in a time of redemption. There have been various theories and legends about the where abouts of the lost tribes over the centuries ranging from the belief that they migrated to the Americas or Asia to claims that they were living in remote regions of the world such as Ethiopia or Afghanistan, however there is no complete evidence to support any of these theories and the fate of the lost tribes remain a mystery.”
That’s pretty amazing, isn’t it? I just read something that a computer wrote, constructed. I don’t know exactly what the right words are. It flowed along pretty good. You all sort of followed it. It had all the right sort of formula and words and bits and pieces. There wasn’t any weird stuff in there. And yes, the AI has read the Bible because there is a bunch of different versions of the Bible, different translations and it has read them all and memorized them all, and syntext of it all and everything.
If you read that at a fleeting pass it feels about right. Doesn’t it? Technically, it’s correct. The answer is very much the cat sat on the mat. It’s a human answer. There is nothing here about a humble mind. About a mind that is seeking God in faith. Accepting Him at His word that Israel is His family, and He will use it to preach His gospel and restore humanity.
A person with that mind knows that Israel does remain distinct regardless of the noise around the mystery of where they were. In short, there is no wisdom, no morals, and certainly no Holy Spirit in any way, shape, or form in this answer. Or in fact, any other answer given by any other artificial intelligence. These responses don’t represent anything more than a math program arranging a bunch of letters based on context clues. The cat sat on the mat.
This chat box, despite scanning the entire internet, is not capable of deciphering whether the answer it gives is right or wrong. This widely noted limitation illustrates just how far these models at the moment are from not only intelligence but also from knowing what morals are, whether they be good, bad, or evil. What powerful ChatGPT and other A.I.’s are is simply a new way of organizing human information that has been found on the internet. The computer program can only reflect back what it has learned about languages and subjects of all types but it’s just going to reflect it back to us. Rearrange it, sure, but it’s us back at us. There is no outside perspective.
What do I mean by outside perspective? “But let your ‘Yes’ be ‘Yes,’ and your ‘No,’ ‘No.’ For whatever is more than these is from the evil one” (Matthew 5:37).
This is a very deceptive scripture for an artificial intelligence. Let your blank be yes and blank be no. Maybe there is some room in there. This scripture is far more demanding than just a simple ‘find a word’. What are we going to put in place of the blank? It’s saying, let your no be no even when you don’t feel like saying no. Why do we pick this up? Because a non-human mirror, an outside perspective clearly tells us in the second part of the scripture, whatever is more than these is from the evil one. Now when we put that together, more than a human yes or no is from the evil one. Did a computer program come up with that? Did you or I come up with that? Clearly, no. This is godly wisdom. This is wisdom from above. This is wisdom from an outside perspective. This is wisdom conveyed by the Holy Spirit.
Jesus Christ had to pay the price of us saying no and doing yes. This is something that no artificial intelligence will ever comprehend. Put another way, artificial intelligence can never take the Passover. We know that we are to study God’s word and we’re to prepare for the Passover in a time worthy of the manner because it requires us to put down the human mirror that we dearly love looking into and even beyond that, we’re to put down the human veil over our eyes into doing our utmost to see us with God’s eyes. If done the way we are commanded to do, it’s a very uncomfortable experience. It’s not something that makes us feel good. We’re to do this in faith because when it’s done right that outside force carries an eternal blessing of freedom from human sin and it’s going to bring eternal life. This is something that goes far beyond human words, syntext, context and intent.
If you’ve not heard of ChatGPT or you’ve not really heard much about artificial intelligence just yet, please don’t let your guard down thinking, I’ve no idea of this ChatGPT whatever that guy’s talking about. I’m having trouble following this entire article. I’m simply going to ignore it and go about business as I’ve usually done it. Please be careful. Please be very, very careful. If you own or use a computer, laptop, tablet, cellphone, or any device that connects to the internet, in a very short time, you’re going to find the time remaining on this planet either really cool or really unsettling.
Why? It took Facebook four and a half years to reach 100 million daily users. Four and a half years before it finally got 100 million daily users. It took Instagram, the photo sharing app, three years to reach 100 million users. It took ChatGPT three months before 100 million people were using it daily.
You may not necessarily, knowingly, as yet use ChatGPT or any of it’s other artificial intelligence spinoffs but many, many other people have. Microsoft, Google, and Apple are racing to put this AI into their search engines and products. Just simply searching the internet, you are invoking the AI and you are interreacting with it. Even if you don’t necessarily see it at this point, Facebook is rolling out AI into our timeline feeds. In other words, AI is being rolled out into every single nook and cranny of the internet and it will be used with just about every interaction we have with technology and, of course, in the society around with 100 million people and growing fast.
There will be no escaping this going forward. We can’t turn it off. We can’t just ignore this. Artificial intelligence is going to sadly become something that is just going to stick in our mind. Keep this in mind also. This AI is being taught and leveraged to make every interaction we have feel really good and really right. It’s trained to echo back to us just what we want to hear. Just exactly what we are requesting from it and the catch is, there is going to be an aura of authority attached to it. After all, artificial intelligence told me, so it must be right. Right?
What can we do to guard our minds against this very smart but very carnal artificial intelligence? What can we do?
“But he who looks into the perfect law of liberty and continues in it, and is not a forgetful hearer but a doer of the work, this one will be blessed in what he does” (James 1:25).
The perfect law. That doesn’t come from us. That’s not our language, it’s God’s perfect law and it gives true freedom. Nothing that any artificial intelligence can ever give. It’s not just what we’re hearing from that perfect law but it’s what we’re doing. If we do, our yes be yes and our no be no, then we will be blessed by God in that.
Ben Orchard