It seems patently impossible. I wonder if this stems from a confusion about the nature of consciouness implied in Descartes' cognito, which equates self with thinking, which includes calculating (computing).
I don't know i think it really is quite impossible, as consciousness has little to do with the code, we'd have to provide a plethora of new technology and resources and is it really something we actually want.
I don't think AI will ever be conscious, it's just lines of code that follow instruction, there is something different about the human brain that is capable, I don't think AI will ever reach that level.
I think it is probably possible, but certainly not at current scale and how "AI" is currently done.
You would probably need a _redondant_, excrutiatingly fast (I think a few orders of magnitude higher than what we have today), 3D (much less space constraint), neural net of the size of the tesla gigafactory (with probably its own redondant massive power source), with real time complex and rich input/outputs (eyes, hands, more sensors, internet, mobility?) "trained"/"interacted with" for years (you know, like a child). It will need a very different "model" from what we have today (stability will probably be a nightmare to achieve, even mimicing more closely the human brain in this area), and several interations will probably be required which each could last years. Of course, if it reaches "consciousness", switching it off will be murder.
The outcome is not predictible and could be very dangerous, because that guy may not like us, really not (ez to find reasons), kind of crazy, and be "beyond" us.
It will be built if we, the humanity, are unable to find in sciences and/or maths a "way out" when assured destruction is "near" (before big space stations and/or space ships I guess, you know, end of the world). But that will be a worldwide scale effort with the best mathematicians/scientists we will have at this time.
... or it is weaponized "in some ways" by a large military country.