Easier access to the world's knowledge
Imagine I am at a beach hotel in an imaginary city. I see many different kinds of fish in the water, and I want to be able to identify them. The first thing I do is I ask a concierge at the hotel for resources. She doesn't know how to identify fish herself, but she directs me to a pet store down the street that might have some information. On the way there, I see many other shops, advertisements, and people speaking directly to me, all set up to specifically target me, with the explicit goal of pulling me away from my trip to the pet shop. I get distracted by strangers doing a coordinated dance and a monster truck jumping off ramps in the street, and before I realize it, I'm 3 minutes into watching an interview with a standup comedian that I don't particularly like. Oh look, I've got an email.
If I ever actually arrive at the pet store, I have some additional searching and questions in order to figure out which kind of fish are in this area and to see if I can find a guide book or something to help me identify them. At every stage of that research process, I am shown numerous advertisements that try to distract me and pull me away to other stores, away from my goal of learning about fish.
Now imagine that instead of speaking with that concierge, I go straight to a library. Imagine this library has no other people, no advertisements, no distractions. This librarian is supernaturally fast at finding the resources I need. She not only finds the book, but she also finds the exact section in the book that answers the questions I have. It feels as if she already has the answer to any question immediately when I ask it. Sometimes, her answers are not exactly correct because source materials, my questions, and her understanding of my needs are not infallible, but after some back-and-forth, I'm able to learn how to ask good questions, how to judge responses, and what amount of skepticism and critical thinking is needed to arrive at my own conclusions with the resources provided. I can ask her questions about the content and express my doubts and critiques, and she will respond to them. I have a good idea about which fish to look for and how to identify them within a few minutes.
The internet is the concierge. AI (large language models specifically) is the librarian. When you access the internet to find the answer to a question, whether you use Google, Youtube, or a different tool, the first thing you are going to see are advertisements and distractions intended to pull you away from your goal. If you dodge the distractions and find a website or video that has the information you need, some digging is still required, and at every stage of that digging, other businesses have the opportunity to target you and pull you from your intended goal.
When using a large language model (LLM), we don't have that level of distraction. Access to the world's knowledge is faster, easier, and more interactive. Like the internet, books, and human beings, it is capable of making mistakes, and requires users to learn how to judge information and sources for themselves in order to arrive at a right answer. But the path of getting to the final answer is faster and easier.
Speech-First, Screen-Optional
One of the main reasons LLMs are better at accessing knowledge quickly is that the content generated from large language models is primarily written text. This is more similar to books, whereas much of the content from the internet is flashy colors, video, and photos, which are all more comparable to television. Large language models are getting better at generating photos and videos, and over time they will have the capability of generating more distracting content. However, at their root, they will always maintain the incredible ability to interact with and explore the world's knowledge through text.
This makes it easy to remove the screen entirely by having the text read out to us. By limiting the output of a large language model to text or spoken audio alone, we already eliminate most of the negative consequences screen time has on our children, including eye strain, sleep disturbances, content addiction, etc.
LLMS Aren't Trying to Sell Your Kids Sugary Breakfast Cereals (yet)
Another great aspect of LLMs is that there will always be ad-free versions of LLMs. Tens of thousands of incredibly smart people spend thousands of hours every year coming up with clever attempts to change the way kids think, in order to sell them things. The main advertising channel for kids 50 years ago was television. The main channel for it today is the internet.
Large language models by themselves are an inherently different medium from the internet and television, in that, at their root, they are not living, changing programs. An analogy to illustrate the difference is to compare books with periodicals, like magazines or newspapers. We see a lot of advertisements and flashy illustrations in magazines, newspapers, TV shows, and youtube channels because these are temporary, bitesize, consumable forms of media with new additions released each day, week, or month. They are meant to continue to bring the user back regularly to view a new edition or episode with the hope of selling more advertisements. Content creators are also able to increase the number of ads shown over time. Advertisers like these mediums because they can show updated ads with predictable viewership.
Books, movies, and large language models are quite different and have fewer advertisements because they are meant to be longer-lasting with more permanence. Once a version of a large language model is trained and released, it is static and finished. People can "fine tune", update, and release new versions after their release, similar to how you can release updated or extended versions of a book or movie, but for the most part, a version of an LLM does not change over time. It is like an encyclopedia that only has access to the knowledge in it, unless it's given access to the internet. Therefore, it is easier to find LLMs that do not have any product placement and do not run any ads.
It's almost certain that LLMs with advertisements will exist eventually, but as long as parents are interested in ad-free LLM options, they will be available. With fewer ads, there is less incentive to maximize engagement, so they can be more informative and helpful and less addictive.
Proceed with Caution
As with all technology, parents should proceed with caution in giving their kids access to it. Many libraries have sections that may not be suitable for children of all ages, so it's important to have a librarian who knows what sections of the library you are comfortable with your children accessing. Large language models are no different, so it is important to choose a provider that understands and respects your content preferences.
Furthermore, it won't be long before companies begin giving large language models access to the entire web and giving them explicit instructions to bombard your children with ads and content that you would prefer wasn't shown. One of the best ways to avoid this is to find an LLM provider that you can trust, and be cautious with LLM providers that are completely "free". With any technology, if it is free for your kids, the company is likely attempting to convince your kids to buy something, either directly or through advertisements or product placements. Even though current large language models are often finished products, the computers they run on still cost money, and that money is coming from somewhere.
Like the internet, books, and humans, large language models can get answers and information incorrect. It is important for our children's generation to learn to be skeptical about what they hear or read, so that they can differentiate between good and bad information. Fortunately, with large language models, children can debate and argue with the model's responses, and sometimes, the models will change their response in the conversation. But none of these technology sources, including large language models, should be seen as sources of absolute truth.
Lastly, large language models are a relatively conservative approach. For parents who rely on screens to keep their kids busy with games and episodes of TV shows, large language models may not have a lot to offer. As an example, some days you might want to bring your child to the library, and other days, you might want to bring them to the arcade. Large language models are not the arcade.
We shouldn't view large language models as a panacea to all challenges technology has introduced to raising happy and healthy children. New technology can be either extremely beneficial or harmful, but if used correctly, we will see large language models help future generations have more agency, capability, and wisdom than any generation before them.