Winter is coming, again.
If you could name three of them I'd be really grateful. Serious question; everything surrounding ML seems to be only good for (non-monetizable) art projects.
As art it is amazing, not going to lie, but "commercial use" seems like a huge stretch.
Translation (Google translate, DeepL)
Automatically generated product descriptions, sometimes also edited by humans (Alibaba)
Image Tagging (Facebook photos)
These applications seem to firmly fall into the "I'm willing to compromise on quality if I don't have to pay a living person a wage" niche, so they're value-destroying, not value-creating.
Are there examples of value-creating applications for ML? (From a business point of view; obviously the "shitty translations but at no cost" proposition creates value for the average Internet user.)
In either case it is only 'value-destroying' if the business has unlimited resources.
I'm talking about commercial applications, like the OP said. That is, things I could potentially pitch to management and substantiate with something concrete that isn't "you can fire your lowest-paid contractors now".
At any rate here are way more than 3 other uses of DL today off the top of my head:
* Autocompletion (be it in search engines toolbars or in Gmail/Word/...)
* Superresolution GAN, the most interesting example to me being NVIDIA DLSS, you render a game at ~720p or less and then upscale it to the target resolution of 1080p or 4k, allowing to get quality that the machine would not have been able to support at the target resolution directly.
* Image recognition/tagging: Most of this is used in the security domain, but there is also a lot of stuff around inventory management, safety etc.
* Semantic search
* Protein folding (AlphaFold)
* In astrophysics: detection of supernovaes, FRBs and probably a bunch of other stuff I'm not aware of
* Self-driving cars: Even assuming self-driving technology does not evolve anymore from now on, the current state of the art is still a selling point.
* Predictive maintenance: Used for plane engines and other things
Today, I've received a package from Amazon containing router bits (for wood working not IT). It contains a so called "User Manual" which is obviously so badly translated, I assume automatically, that it will only fool a spell checker, that it is actually written in German.
I often hear and read good things about Google Translate but every time I read something from it, e.g. when a browser or webpage helpfully decides that I would prefer a butchered salad of German words instead of an English web page, I am repulsed.
Most of these were developed woth actual business partners and are being used right now.
Latent Dirichlet Allocation, message-passing in Hidden Markov models, and Naive Bayes for spam filtering. Outside my subfield, there's always the basic handwriting recognition employed in ATMs.
These projects have an impact in the real world [predictive maintenance on infrastructure that serves people, for example].
I see the sentiment you've expressed a lot, and feel it speaks to a massive disconnect amongst developers. Most people are interacting with ML systems dozens if not hundreds of times a day.
The stuff you listed doesn't, it's just part of a moat for already established products that don't depend on ML for their market share. (It's not like Netflix will lose market share if they switch from ML to some other approach for their recommendation system.)
Industrial robotics, driver assist, drug discovery, computational photography, speech translation, and so many other examples illustrate a clear commercial applicability of ML methods scaled in the last 10 years specifically.
Facial recognition. iphone face unlock, photo tagging, etc.
Behavior prediction for advertising. Ad quality scores basically.
> Siri/Cortana/Google Home/Alexa, powered by DeepSpeech+language models
> Google Search, powered by BERT
> Tesla, powered by variants of YOLO
> Facial recognition powered by MTCNN+FaceNet
> AirBnB product search+recommendations
> Amazon product recommendations
GANs are a bit artsy, but JFC they're not even a decade old - we've gone from shitty MNIST clones to fully synthetic faces in the span of 5 years!
Predictions: 1. I suspect we're going to see DeepFakes in Hollywood - famous people might license their faces to movies that they might not have the time to star in
2. People are going to start building even more powerful versions of search, like combinations of CLIP
3. Neural networks still aren't optimized for edge devices - we're going to see a deluge of cheap drones with cutting edge computer vision by default
It is also used quite a bit in graphics and imaging; DLSS is a consumer-facing application, but it is also used in other domains, like OCR.
Machine translation is another ubiquitous use case. As is any kind of language processing, like text-to-speech.
Also, in the industry, it is heavily used for anomaly and defect detection. Also, Google reportedly uses it for a lot of search/recommendation stuff.
ML is definitely monetizable, but not every company needs it, by a long shot. It is not "AI" and seems to often resemble a complex DSP step when used in practice. I think it's overhyped, but it is far, far more useful than anything blockchain will ever be.
Speech recognition + basic NLP for automatic customer support triage. None of these are great to use as a customer, but they seem to be effective enough to continue using and save companies lots of money.
Automatic "offensive" content detection for social media. I'd bet they use ML to do a first-pass on uploaded content to make sure it doesn't contain porn, gore, etc. Probably lets these companies save money.
Automatic defect detection in factories. Instead of training humans to detect subtle issues in manufacturing issues. I think companies like Samsara are experimenting with offering this tech as a service.
Facial recognition/tracking for law enforcement/defense. Ignoring the ethics of it for a moment, it seems like governments would be willing to pay a good amount of money for this tech. Could be used to automatically search through hours of footage to find which frames (if any) contain a target.
I think we might see a winter in super big applications like self-driving cars or voice assistants, but ML in general is just a boring, non-controversial business tool with hundreds of valuable applications.
You’ll still need statistical specialists to train and operate models and ensure systems avoid pitfalls like overfitting, poor convergence, multicollinearity, confounders, etc.
So I doubt this will have much impact on ML job market. Companies that invest in ML will continue to run circles around companies that don’t. You’ll just see the unjustified over-focus on SOTA neural networks die down and become just another boring tool in the toolbox like everything else in ML.