Presumably LaMDA already learns from its input/knows what it's already talked about, and to satisfy #1 you could trivially just run it on some sort of null input. #1 doesn't really make much sense, it already is constantly "on", it just doesn't do anything while there's no input for it to receive.
You appear to have only read up to the first comma of #1. I very explicitly said that it needed to be constantly receiving input. The kind of discrete input we give it now is qualitatively different, even if it's "on" in the same sense that an Apache server is always "on", waiting for requests.
What does "constantly receiving input" mean? How would a program whose input is text be constantly receiving input? What would you expect it to receive?
Well, I think that's part of the problem, isn't it?
If an AI's only input is text, I'm not sure it can meaningfully be conscious, because of the discrete nature of text. Note that I'm not sure it can't—someone may come up with a way to make such a thing work that I haven't imagined—but I think that's an important part of why I doubt the fundamental ability of current-gen AIs to come close to the threshold for consciousness.
1. Continuity of input: The AI needs to be constantly "on", constantly receiving input of some sort, and constantly able to produce output, rather than being strictly limited to producing discrete responses to discrete stimuli.
2. Continuity of learning: In addition, the AI needs to be continually update its "mental model" of the world—in effect, constantly "learning" and re-training its neural network on the input it receives.
I believe there are likely to be other requirements on top of these, but I am sufficiently learned neither in neuropsychology nor machine-learning/AI to be able to speak (or even reason) with any confidence about them.
[0]: https://news.ycombinator.com/item?id=31728907