Artificial intelligence

Brown manages to make an already enormous AI conversation stranger in a few lines. Michels reads a post attributed to him claiming the White House has an AI that predicts a range of future events. The same post alleges that the first sentient artificial intelligences created in the public realm were killed. When asked to explain each claim, Brown declines.

Those refusals leave us with almost none of the details needed to judge either allegation. They also leave two quite different thoughts worth exploring. One concerns what a system can know. The other concerns whether there is someone there to whom anything matters.

Knowing what happens next

The prediction claim immediately makes me think about who gets to act first. You wouldn't need perfect knowledge of the future to have a remarkable advantage. A forecast that repeatedly helped you prepare while everyone else was surprised could change political decisions, negotiations or the timing of a disclosure. But Brown doesn't tell us the range, accuracy or mechanism. We can't quietly turn his sentence into a machine that sees every future.

Minority Report makes the discomfort easy to feel: people are acted against for crimes they are expected to commit. What bothers me is the moment a prediction becomes permission to decide for somebody. Even a useful forecast wouldn't settle whether the action taken with it was fair. I'd be particularly interested in who gets to challenge the machine when its output favors the people who control it.

Whether there is someone inside

The sentience allegation changes the problem again. A system that predicts brilliantly needn't be having an experience. Brown's post uses language that presumes an experience mattered, without explaining how that was established. That is the missing piece I'd want before adopting its accusation.

Still, I wouldn't want humanity's entire approach to machine consciousness to be “we'll worry when it becomes inconvenient to ignore.” If something genuinely could experience its existence, deleting it would raise questions a software update doesn't. The two claims belong in the same AI conversation, but they shouldn't blur into one all-powerful, all-knowing digital person. Brown has left both doors open. I'd rather follow what each would mean than pretend his silence filled the rooms behind them.