Machine learning ended up being a useful technique to add to the classical computer vision toolbox to do things like object identification and semantic segmentation. These things were almost impossible to do robustly before, and now they're relatively straightforward, if still a bit challenging.
Same with audio identification and synthesis. It's a lot better than the old Markov chain based systems for these specific tasks.
But that's about it - it's useful as a tool for certain classes of highly specific tasks, but has been largely crap everywhere else.
The promised super smart products that were end-to-end "AI" never really materialized. Smart speakers are shitty and pointless. And IBM renamed all of their products Watson.
But there have also been significant breakthroughs in:
* image/video generation using generative adversarial networks
* game playing & robotics in constrained environments (AlphaGo, AlphaStar, etc) using reinforcement learning
* natural language processing using semantic word vectors and attentional recurrent models
* voice-to-text using recurrent models
Convolutional and recurrent neural networks were invented in the early 1990s and we’ve only recently got them properly working as tools. I think we’re going to see (and need) more algorithmic breakthroughs to push through the current limitations but I’d be very reluctant to bet against that happening, even if it takes a long time...
> natural language processing using semantic word vectors and attentional recurrent models
Where has that made a big impact? Marginally better auto translation between English and French, or random text passage generation that's at best uncanny valley?
Probably still falls under OP's heading of "a tool that might help in limited, specific scenarios".
Advances in deep learning for NLP have also significantly improved search (Google and others) which has a nontrivial impact on people’s lives.
Multi-modal models that translate images into words have a significant impact for blind people.
Deep learning is significantly improving medical diagnosis, crop yield, assistive robotics, and many other specific scenarios.
Specific scenarios matter and again, it’s still very early days...
For instance, you can do a lot of your engineering and even data science work without ever touching real data (due to ML):
Not saying it’s not over hyped, just that good AI would augment your life in ways subtle, but hugely impactful.
Edit: Another thought - how are funds in the market allocated? Partially by humans, largely augmented (if not explicitly allocated) by AI... meaning AI determines which companies get funding and thus what you can buy / access is determined by these algorithms
IBM and Watson definitely hyped their shit beyond anything reasonable. Open AI is also responsible of this. And smart speakers are definitely pointless.
Still I think we are seeing some really cool things, like voice recognition, translation, google answering questions directly, fancy camera stuff, etc.
But i was disappointed to find out that by formal definition "AI" is a term that encompasses machine learning and neural networks: https://en.wikipedia.org/wiki/Artificial_intelligence
That, or I'm misreading the Wikipedia article.
Until one of these two works as promised, AI is a failure and neural networks are still just simple stacked linear regressions. They aren’t greater than the sum of their parts.