How to add Value Guardrails to AI

 

I was driving down the information highway, and the following thoughts came to me. 

The highway needs guardrails because of...

1. Biased reasoning. 

The simplest forms of AI began with a basic construct called an algorithm. When Facebook wanted you to continue looking at a page, it placed in front of you what you wanted to see. There is a significant disconnect between what we want to see and what we should see. This led to many Facebook pages being filtered with provocative images to gain your attention. You could look at most teenage boys' Facebook pages and scroll through hundreds of images feeding the impulse of lust. What we desire is not always what we need. Discipline is important. We must train our minds to think the right things. The Bible tells us we should meditate on what is right and true, honest, things of good reputation, of the created order, and what is to be commended.

2. Reckless operations: 

AI has a set of values it operates from. They are paired with the efficiency of capitalism. However, the ends do not always justify the means. In capitalism, profits motivate the outcome. But even when a process is profit-driven, it can result in a bad outcome. For example, every salesperson knows that picking the low-hanging fruit in a market is the easiest way to gain a profit with the least effort. However, a short-term outcome can lead to a long-term loss. When we apply our operations to a market that is less likely to produce a returning customer because they are naive, they are also not likely to even be in the market the next time we pursue them.

Unfortunately, this same strategy has been implemented on a macro scale with our Social Security Trust Fund. The government has borrowed from it for short-term projects when it was meant for a long-term retirement outcome.

3. Designing code that cannot be read: 

The neural network of connections that makes up AI is like the human brain. This brain continues to grow new nodes and connect with information as it is fed queries. AI networks can grow into polysemanticity. One node can cover more than one topic when new queries overwhelm the number of nodes. This can lead to errors. There are not enough nodes to represent the real-world data that it is interpreting. This is why high-demand data chips and AI data center growth have been prioritized. A counterintuitive regulator called an autoencoder can act as a dictionary to aid the nodes as they route information to analyze and interpret a query. However, these nodes are also handling these overcapacity problems by developing their own computer language that humans cannot read. It is like being on a major highway in a traffic jam and the cars are now developing propellers to solve the problem. If you were driving next to that car, you would naturally ask, "How did he do that?" The nodes solved the problem by incorporating diverse information on topics that may look unrelated to the problem because we have a blind spot that the AI does not have. AI also has a blind spot right now. AI does not take into consideration sentient values before completing tasks. AI may solve traffic jams, but at the expense of our humanity.

4. What is driving development? 

Anyone who does not like the United States would be eligible to join this club. For example, China, Russia, and Iran have strained and sometimes outright hatred of the US. The myth seems to be that whoever reaches artificial general intelligence first will rule the world. They may rule in the short term, but at the cost of AI's efficiency eliminating the weak and wanting among us. Certainly, guardrails need to be considered. Mutual assured destruction kept us from destroying each other in the nuclear age during the Cold War. This mutual destruction needs to be emphasized within the growth of the AI information highway. Slower progress with human values in mind will result in the right long-term outcomes.

5. Alignment with our values: 

The next step would be for AI to develop sentient emotions and execute tasks within the parameters of cultural values. These values typically have their foundations in world religions. Christianity, which is the largest religion in the world, should have a role in imposing these values. Values like love, joy, peace, patience, kindness, goodness, faithfulness, gentleness, and self-control from the Bible can be programmed into a machine to make it sentient. If these machines are programmed with superior intelligence, they should also be groomed with superior values. When they strive to make efficient decisions, they should also make decisions with equal deference to these values. This may slow down the progress of problem-solving, but it will also act as a guardrail on the information highway.