Artificial intelligence is many things but “natural” is probably one word that isn’t associated with it.
Apple wants to change that as it works to make AI-powered services like future versions of Siri “more natural” in the way they interact with users.
This includes contextual reasoning or “reading between the lines,” PetaPixel reports. This is a method whereby AI can “intuit” based on screen inputs what the ideal output should be.
The research white paper released on the initiative notes that language itself contains various ambiguities that typically can only be inferred from context. That is, you need to know what someone was saying prior to the statement in isolation for that statement to make sense. Examples cited include “they” and “that” but there are many others and this isn’t something exclusive to English but present to varying degrees in every language in the world.
You might recall some of our previous articles on this subject, particularly with regard to generative AI and photography. In those stories, there was a different kind of “unnatural” that developers were trying to target; namely, the addition of limbs, fingers, and otherwise bizarre output that generative AI imagines when outputting images based on humans and animals.
In fact, AI’s somewhat “uncanny” appearance is really one of its distinguishing features here in the early days so the gradual elimination of this aspect is not only to be expected but it is more than warranted (particularly when it assigns someone 20 fingers).
But the challenge with actual speech is much greater and when you combine the vagaries of speech with generative output, you’re likely to end up with some crazy stuff.
Any recommendations on AI platforms or thoughts you might have on this fast-moving space are welcome in the comments.
We have some more photography headlines in our news section if you want to check them out.
