One of the most powerful things about AI is what it is teaching us about ourselves. If one dives in deep enough, one finds people discussing concepts that relate to consciousness.
What are we learning about ourselves?
The answer to this question obviously extends beyond one article. Therefore, we will focus upon something a bit lighter.
In this article we will dig into how AI learns and relate it to learning a foreign language.
When it comes to AI, most people are familiar with large language models (LLM). Naturally, there are many different areas of AI. Here we will focus upon LLM and the process.
Models are derived from repetition. Data is fed in, being used to train the model. We see the learning process take place over many different iterations. The net result is a model that basically guesses at the next word based upon statistical probability.
That is what is derived from the training. By reading how people write (and how they speak), the model picks up on things such as grammar. Each language has a certain grammatical make up, something linguists know very well. In fact, those who are adept at learning languages quickly cite this as one of the keys. By "wiring" the brain to the grammatical structure, the process is made easier.
AI does something similar.
When we think about it, we are always dealing with probabilities.
If I write, the girl went to the __________, what are the odds the next word could be "are"? It is low since we know a verb does not fit into this sentence. The same is true for drive, see, or was.
Context is driven by more exposure along with repetition. This is an important point. When learning a new language, it is important to get reference points via repetition. The mind, however, gets bored easily, so if no new information (data) is added, progress is stopped.
This is why models need ever more data.
Consider how we learned to speak out native language. From our time of birth, we heard people speaking. There was a time we have no idea what they were saying but we still learned. Our subconscious was learning the entire time. We were also framing the grammar of the language.
As time went by, some of the vocabulary started to make sense. We did not sit down and study work cards for years. We simply started to add words as time passed. It was not concentrated learning like memorization.
Using AI as a guide, we can model this.
Many will state the best way to learn a language is immersion. Drop one off in a foreign country and necessity mandates rapid learning (at least knowing how to ask where the bathroom is). This is a key. Immerse oneself as much as possible.
This brings about repetition, something AI models undergo with training. It will run through the same data dozens of times until it gets it down.
Another component is to add more data. With languages, this means learning new words. This means expanding the content we consume.
A final piece is to use many forms of media. Again, models now can do this since they are multimodal. They learn from text, audio, images, and video. For a foreign language, learning the alphabet, watching videos, listening to music, and reading articles can help to break up the learning process.
It also enhances the type of information (vocabulary) we are exposed to.
Over the last 6 months, I started to connect some dots between AI and humanity. Many fear what AI will do to humans. Actually, I think we should focus upon what the technology is teaching us about ourselves.
Human expansion is tied to the emergence of this. Ironic that it might be more reflective than something outside of ourselves.
Come to think of it, that isn't that surprising. Technology tends to be extensions of ourselves. The power behind AI means it will uncover deeper revelations about ourselves.
