How Can A.I. Become Aware?
What does it mean for A.I. to be aware of what is going on around it? How is that different from simply seeing and hearing things?
Awareness
Take a moment and turn your attention to what’s going on around you. What noises do you hear? Is there a distant hum of a fan or traffic that you hadn’t paid attention to until now? Are you gradually becoming aware of these sounds?
There is a difference between simple hearing or seeing and being aware. You are always hearing the noises around you; they are always in your ears. But only when you turned your attention to them did you become aware of them.
Is it absolutely necessary for A.I. to have awareness? You and I often get by just fine without it. You can drive a car or wash the dishes while your mind wanders, not fully aware of what you are doing. You can see, hear, and even react to things around you on unconscious reflex. Even while reading these words, you aren’t paying attention to every letter (although you probably are now). So why does awareness matter at all?
Awareness Forms Memories
One thing you might miss if you don’t pay attention is the ability to learn new things. It seems self-evident that in order to learn a new piece of knowledge you have to first be aware of it. You will only remember something if you were aware of it at the time. The act of awareness and the act of forming memories are tightly connected.
For example, while doing a routine activity like locking your front door, your attention may wander. Later you might ask yourself whether you in fact did lock the door. If you weren’t paying attention, and therefore weren’t aware, you’ll have no memory to fall back on.
Can you remember the ambient noises you became aware of at the start of this article? If you do, then you formed a new memory at the time. That memory was created the moment you become aware of the sounds around you.
It seems the biggest difference between being aware of something and simply seeing or hearing it is that you remember things you are aware of, whereas you don’t remember things you aren’t aware of. Such memories rarely last a lifetime. Most are forgotten within minutes or days.
Without being aware, A.I. couldn’t learn much about how the world works, or make predictions based on the past. The ability to make predictions, to make plans, to have expectations, even ambitions and dreams, all require learning about the world. If we’re to replicate awareness in A.I. we need to understand the exact steps involved in how awareness creates memories.
Forming a Memory
Paying attention is the first step to being aware. If an event is over before you have time to pay attention, you won’t remember it.
You can even pay attention to and become aware of your own thoughts and feelings. ‘Looking inside yourself’ is similar to looking at the world, in that you can only remember your thoughts and feelings if you pay attention to them.
Once you start paying attention, you start learning. Like turning on a recording device, you remember some of what happens next, for some period of time. If you look at your own memories, you’ll find that most of them have a duration of less than a few seconds. None of your memories is infinitely long.
This implies that you are only actively aware in short segments of time. At the end of this period a new memory is formed. The sights and sounds you experienced between when you started paying attention and when you ‘closed the loop’ become your new memory. Afterwards, you can look ‘inside your mind’ and see the memory you just formed.
This happens, for example, when you are reading a book. You turn your attention to a phrase, and understand and register its content. You then turn your attention to the next phrase, remember something about that, and so on. Each time you form a small, usually temporary memory that sets the stage for the next section.
In many cases you refer to this process as noticing something. To notice something means to be aware of and to remember it. Such a memory can last for a short time, or sometimes longer. Awareness itself is the sum of all these small moments. This is true even when you look inside your mind. You can ‘notice’ your own thoughts and feelings.
It seems the process of creating memories is triggered by the act of paying attention, at the end of which a memory is formed. This presents two crucial questions:
- How and when should an A.I. decide to pay attention?
- What should it remember when it does?
Awareness has an Agenda
Paying attention isn’t an arbitrary act. You must be motivated to pay attention, and have a reason to do it. You wouldn’t pay attention if you didn’t care to. So when you pay attention you are, in a sense, looking for something. It is a kind of search. You may not know exactly what you’re looking for, but you notice it when you see it.
To put it another way, every time you pay attention you are implicitly asking a question. The question may be as basic as “what’s going on?” These questions rarely get posed as actual sentences. They are more akin to a need, a demand, a request. This need looks for an answer from the world, or sometimes from your own thoughts.
Take a simple question like “what’s in front of me?” If I asked you to pay attention to what’s around you right now, you might notice things like desks, or pens, or people. You are unlikely to notice “love” or “bravery” since these are abstract concepts which don’t answer the question posed. You are looking for something specific, and only certain information will qualify as an answer. On the other hand, if you’re talking with someone and you’re wondering “what is he feeling?” you might notice “love”.
A.I. will pay attention when it has a question it needs to answer, or more generally, a problem to solve. And this is triggered by the situation.
You may not realize this agenda was present when you later look back on the moment, because only the memory of what you heard or saw will remain. But it is present. For example:
(Pay attention) “What is she saying?”
(Answer) “She said she bought eggs”
After the fact, you may not realize what drove you to pay attention in the first place. It may have gone something like this.
(Problem) “I might be offending her by not listening.”
(Pay attention) “What is she saying?”
(Answer) “She said she bought eggs.”
Other examples are:
(Problem) “I might trip when I start walking.”
(Pay attention) “What’s in front of me?”
(Answer) “There’s a curb and a puddle.”
or
(Problem) “This stranger may be dangerous.”
(Pay attention) “What is he doing?”
(Answer) “He’s walking across the street.”
The initial problem drives the whole process. Paying attention is only a link between the initial problem and the answer.
In order for A.I. to be aware, it must recognize certain situations as problems, as well as what qualifies as a solution. You can read about how this happens in another article.
Once the problem is triggered, the A.I. need only wait for something matching the solution to present itself. A solution is any set of conditions that you add to the initial problem to neutralize it. For instance, once you spot a stranger you might feel ill at ease. Knowing what it is that he is doing might make you feel more at ease.
The Answer Becomes a New Memory
The loop closes and the memory forms as soon as you have an answer to your original problem. The answer that you get is what you remember. To learn anything after that you must have a new problem, a new request, a new reason to pay attention.
There are too many sights and sounds in the world to remember them all. Like humans, A.I. must be selective in its questions and demands. What it learns must be useful, by which I mean it must answer a need. At any given moment an A.I.’s motivation must drive what it learns from its surroundings. Learning is an active, self-directed process.
In summary, awareness is motivated seeing and hearing, motivated learning. When you and I learn about the world we always do so with an agenda in mind. This motivation is the key difference between simple seeing and awareness.
Are you also interested in applying Artificial Intelligence to human creativity, human understanding, even human values? Do you feel that our current goals with A.I. are limited? I’m looking to connect with others who have a similarly ambitious vision of the future of A.I., whose goal is to tap the full creative potential of human intelligence through software.
