By framing your question like this you are essentially asking "are bricks and cement a modern house?", but because "Artificial Neural Network" sounds like it's on the same level of fanciness as sentience you do not notice how ridiculous the question is.

It makes you unable to think of the answer too, namely: ANNs are a house, and for it to become a modern house it needs some extra components (,or maybe the ANNs are a different component to make the metaphor work better; it's a component among many).

Both "Sentient" and "Artificial neural network" are useless concepts for this question.

"sentient" has become a term with overloaded meaning and as such is not a useful category for this question.

with "sentient" do you mean:

• human-like; in which case: is just a brain without a body sentient? what about it missing some subset of input signals? does the zombie-walk to the coffee machine after waking up count as sentient?
• able to respond to situations it perceives: plants can release toxins through their system once leaves are bitten/harmed.
• tell us their experience: are less linguistically able people less sentient?
• hold a conversation: what about those introverted friends who hardly ever contribute half a sentence if they're out of their comfort zone?

An ANN is nothing without data, training and action space. compare some ANN to classify MNIST digits to ACT-1 or GPT with a python interpreter at it's disposal.

The former is much more purely an ANN, whereas the latter two are given the programmatic equivalent/precursors of bodies, especially ACT-1. Still relatively limited (with domain limited by links on pages themselves, rather than direct url query, but with lots of room to expand, given the ubiquity of software) and prompt driven (but I would say people underestimate how they themselves are also very much prompt-driven and with planning coming into the picture following external commands becoming a smaller ratio of it's set of behavior).

I'd say the most serious lack w.r.t. sentience is response adjustment out of training phase although this seems more an engineering challenge than an ANN challenge (when do you accept that the environment tells you that you're incorrect and should adjust: not always sometimes it's a fluke and sometimes the person telling you you're wring is actually the one making the mistake, not to speak of malicious actors).

It is also considered a proverb that "insanity is doing the same thing twice and expecting different results", yet many people do not adjust their actions. As such I'm not sure this should be a requirement for sentience unless you'd want to exclude people ad I'm not sure this response adjustment out of training phase is something people in general are good at either.