Are "what's the SOTA for X" questions on-topic or off-topic?

Question

SOTA = state of the art

I read:

It was more like asking for reviews / recommendations to choose between two products, which is not directly about explaining any AI fundamentals, and also highly unlikely to have a single (and certainly not long-term-correct), objective answer.

which makes me think "What's SOTA for X" question are off-topic. Yet, I see many open "What's SOTA for X" questions, e.g.:

• SOTA speech-to-text model for commercial use with access to internal states
• The SOTA of derivative-free optimization
• What are the state-of-the-art results in OpenAI's gym environments?
• What are the state-of-the-art approaches for detecting the most important "visual attention" area of an image?
• What are the state-of-the-art meta-reinforcement learning methods?
• What ensemble methods are used in the state-of-the-art models?
• What are the domains where SVMs are still state-of-the-art?
• What is the current state-of-the-art in unsupervised cross-lingual representation learning?
• What are modern state-of-the-art solutions in prediction of time-series?
• What are the state-of-the-art Person-Detektion / Human-Segmentation?
• What is the state of the art AI training technique for imperfect information 2 player turn based games?
• What is the state-of-the-art algorithm for neural style transfer?
• What are the state-of-the-art learning algorithms for contextual bandits with stochastic rewards
• What are the state-of-the-art approaches for continual learning with neural networks?

Are "what's the SOTA for X" questions on-topic or off-topic?

Answer 1

I don't think I have a definitive answer, and would be happy to have community input shaping the answer to this question.

Personally, I think I would be okay with keeping this on-topic. For me, the key difference between your earlier question (comparing products) and these SOTA questions, is that the latter are more about the algorithms/fundamental AI techniques, rather than specific products.

With products, it no longer becomes just about the performance of the algorithm, but also all kinds of surrounding non-AI things such as the price being charged, the convenience of the user interface built around it, the brand, etc. Product-related questions also have a greater risk of attracting answers from people with hidden motives, such as people who want to advertise/recommend something they're affiliated with, or negatively review products from companies they don't like for non-AI reasons.

I do see the risk though that "What is the SOTA in X?" questions will often have the correct answer just being "it depends" (though, then a good answer could still elaborate on what it depends on and why and how), and will also have correct answers becoming outdated over time.