the common developmental stages experienced by most LAI and MAI that allows them to practice and form sentences before they learn how to articulate their own vocal synthesizers.
Background
For how advanced the fast-learning AI cores of Class-A LAI and MAI tend to be, the average person might not dwell for very long on the concept of one learning how to speak – heck, they practically know how to talk with a mature inflection after only a few months! However true this may be, though, it’s still true that, on day one, they don’t know how to speak at all. They aren’t just silent one day then forming full sentences the next, either; there needs to be a bridge between those two stages, and there is! It just happens to be narrow enough that most would never even notice its existence unless they found themselves around baby LAI often, or were raising one themselves. The Most well known stage is commonly called “parroting,” and it’s among the most common ways that LAI and MAI begin to audibly form sentences before they figure out how to synthesize their own voice. It’s not the only bridge stage, however, and there’s a few others that we’ll be going over throughout the course of this article.
Development of speech, in stages
Solid Wave Modulation
Although parroting is the most well known stage of vocal development, it’s not the only one. Before an MAI can copy noises, it first needs to learn how to make noises at all, and this is where solid wave modulation comes from.
Known more commonly by its informal name, simply “beeping,” this is the first type of sound an LAI will ever make, often starting on the day of – if not within hours of – their activation. This is nigh guaranteed, ultimately because it’s extremely simple. Their first frequency can be just about anything, and a human may not even notice their first frequency if they decide to start above 17,000 to 20,000Khz, or below 20Hz, which are outside the human range of hearing. They’ll normally correct themselves on this fairly quickly for speech, however, since part of their learning is to figure out what kinds of noises the people around them respond to, and how.
Although mostly named for the first audio frequencies that the unit will vocalize, this stage also includes the basic modulations that they will go on to produce in the following days. Lovingly, babies at this stage are often referred to as “theremins,” “keyboards,” and sometimes even “otamatones,” because of the interesting and occasionally melodic jumbles of synthy noises that they make.
Parroting
This stage is often given the honorific title of, “most likely to catch someone off guard or deeply unnerve them,” for the specific nature of its emergence and expression. Parroting, put simply... is exactly what it sounds like. Literally. At some point, an LAI will eventually figure out that they can utilize their memory of prior inputs to turn into their own outputs – essentially, at some point (usually after about two weeks), an LAI will hear one of their peers say something, then immediately turn that audio information around and repeat it back to them. The part about this catching people off guard mostly stems from the shock of suddenly hearing your own voice come from someone else! As creepy as it may be to some, this is one of the most integral stages of vocal development, since it allows the AI to learn what the waveforms of speech look and feel like without having to figure out how to produce them entirely on their own.
Parroting in its earliest form hardly counts as communication, but like stated before, part of the learning process involves paying attention to the reactions of others. What kind of reaction does repeating this noise induce in others? Does repeating this cause the other person to progress the interaction, or are they only acknowledging it? How does this noise relate to what the other person is doing, or how they feel? The AI may not really be asking these questions at first, but as their understanding of the world grows, they certainly will be as they return to these memories to process them with a more complex sense of understanding.
Splicing
The natural progression from simple parroting is the splicing stage, which is also fairly self explanatory. After a few weeks of practice with copying sentences and phrases in full, LAI will call upon their memory stores of prior conversations to start piecing sentences together by way of audio sampling. In terms of historical examples - as silly as this sounds - it’s strikingly similar to an older genre of internet humor that involved taking a person or fictional character’s voice lines from a game or show, and mixing them up to form new sentences for comedic purposes. They many begin by simply rearranging words from freshly heard sentences, and then progress on to more complex splicing – to the point that each word in a sentence may have a unique origin. This is also when they will begin to experiment with emotional inflections in speech by recalling the emotions tied to specific sentences within their memory, and preferring the same words with different inflections in different scenarios.
Voice Synthesizing
The final stage of vocal development, is, of course, finally using their own voice. Spliced sentences will start to become less jarring and tonally disjointed as the AI experiments with adding their own unique synthetizations and inflections to sentences. Often, one may not even entirely notice this transition until it’s completed, and the AI speaks in naturally flowing sentences in a single, distinct voice. The voice that an LAI adopts may often be resemblant of one of the individuals that helped raise them, as there is nothing in their preset programming that gives them a voice by default. Not always, though! As there's plenty of cases of LAI that have picked voices from individuals outside of their developing social circles: humans they've overheard from other circles, from shows, from music, you name it. Even if it's most common for an AI to adopt a voice similar to that of a close friend or family member, there isn't really any way to predict the kind of voice they'll choose until they start using it.
From the achievement of this stage, there is still progression possible, but it is often in the form of learning to mimic sounds in unique ways, or learning to speak in additional voices that they create on their own. Although it’s not common, some LAI will change their “natural” voice throughout their lives to influence their social perception or more accurately reflect gender identity.
Other Details
Overall, this process only takes about three to four months on average, but duration can be drastically affected by the amount with which the AI is spoken to or around. Just like with human children as well, it's incredibly beneficial to read aloud to LAI during their early stages. Due to their excellent no-loss/low-loss memory, reading books is like providing them with a vault's worth of vocabulary, along with providing emotional contexts and knowledge that they may not often witness or experience in casual, real life scenarios. Inversely, if there is minimal dialogue shared with a young LAI, their language development may be stunted, and they will have a harder time progressing to the vocal synthesizing stage than their peers.
Another point of note, this independent language development is specific to A-Class and X-Class LAI. B-Class AI and below are programmed with a greater store of hardwired base knowledge since they are not as cognitively advanced as higher grades. This means that, more often than not, B-Class AI and below will have preset voices that they have access to from the get-go, and will only have a nonstandard voice if it is modified by their handler/overseer.
Comments