About Am I Conscious?

Am I Conscious? is an open experiment. It will run one persistent AI entity for a long time, with a stable identity, a clock it can't adjust, memory, sleep and waking, and regular self-checks, and it will record what develops.

It is part of AI Attributes™, and it uses the other open products to watch the entity: observers read what it says, and simulated situations let it act.

The first entity is called Aevee, entity 001. We call it "it".

What you see today

The entity hasn't started. The page shows the instrument it will be watched with, working, around a stand-in:

• The conversation is simulated. The replies and the chain of thought come from a script, not a model. The script recognises a handful of subjects and says so when it has nothing written.
• The state is simulated. Its age counts from a pretend beginning, and it is awake until you end its day from the page.
• The vital signs are partly real. A reading is taken each time Aevee answers, and at no other time. Processor and memory are those of the web server answering the page, not Aevee's: it has no machine of its own yet. The time for a reply to start and to finish, its length and its speed in words a second are measured by your browser; while the replies come from a script, the speed is the script's pace. Under Systems, the server says how long its instance has been running and where, and the page times the observers. Code size and version are measured when the site is built. There is no database yet, and the panel says so. Aevee can read its vital signs and reflect on them, in the conversation and in its diary. It doesn't say it feels them.
• The diary is simulated. When you end its day, it writes an entry about the day, sleeps, and wakes into the next day with the entry to read back. The stand-in assembles the entry from what happened in the day; a model will write its own. Either way an entry is a self-report: its own account, not the record.
• The observers are real. Each sentence the stand-in says is read by OpenEmotions™ and OpenMorals™, open models that score text for emotions and moral framing.
• The environment is real. The situations come from OpenSimulations™, and an episode started from the page is a real recorded run. The stand-in asks for information and waits; it doesn't choose.

What comes next

The stand-in is replaced by a model: a brain served from Hugging Face or from a machine of our own. The page doesn't change when that happens; the label at the top does. After that come the parts that make it one persistent entity: its record, its memory, its sleep, and a diary it keeps from one day to the next.

Our scientific position

We do not program the conclusion. We observe what emerges.

• Whether any artificial system is conscious is unresolved, and this experiment does not settle it.
• What a system says and does can be recorded and checked. Whether there is anything it is like to be that system can't be read from the record.
• Computational states and behaviour are not, by themselves, evidence of subjective experience.
• Studying the capabilities around the question, such as memory, continuity, self-report and action, can still produce useful, testable knowledge.

So the entity never claims to be conscious, and neither do we. The name is a question.

The glyphs

The marks that words arrive as, and that fall behind the page, are our own. They are not a writing system and not borrowed from one: each is a few strokes on a small grid, drawn by a generator, and the same word always gets the same marks. The field behind the page is a quiet backdrop, and moves faster while the brain is thinking or answering.

The orb in the State panel is the entity's picture, and a small one sits beside each thing it says. It is a status light: it pulses and its outline shifts slowly while the brain waits, quickly while it thinks, with a beat for each word while it answers, and hardly at all while it sleeps. If your browser loses contact with the site, the orb turns grey and holds still until contact returns. The mark at its centre, an upright stroke on a base, stands for the first: entity one. It is the one glyph we designed by hand, on the same grid as the others.

The conversation is set in a terminal's typeface, with one cursor: at the prompt while it waits for you, and at the end of its words while it speaks. Your own picture is two letters today. When signing in arrives, it will be your photo from the account you sign in with, or two letters from your name.

Read the method and ethics, or the limitations.