Let's not forget that this is a commercial company.
Even if people will know it is a bot, these people's rational capacities will in most cased be diminished under such circumstances, at best. Many will be easily (emotionally) manipulable, even more than with all the psychological trickery in product advertisement and big data exploitation these days.
Opinion: this clearly crosses the line of criminal behaviors, exploiting people who are clearly in a vulnerable positions (at least more than usual). I think the track record of tech companies rather clearly shows that their promised of taking good care of collected data (and privacy) mean little to nothing. At the end of the day, they do this for profit. That's enough for me to make this cross the line.
Also, what's really the "invention" here? Would something like this ever deserve a patent monopoly? If you take (dead) people's personal data and build a neural net (or any other "AI") to mimic/compliment that data (needed for answering questions), have you not just done what any/every neural net already does? If we're going to call that novel and innovative, then hold my beer ... or is this just again a big tech company abusing the patent system to hijack an obvious (use of) technology, excluding anyone who doesn't belong to their cozy little cross-licensing inbreed family from using what should be free to use in the first place?
The amount of brain power and financial resources these companies have amassed for the purpose of generating targeted ads or HMIs, is disconcerting.
Maybe it's too much of an Arthur C. Clarke type fantasy but I always wonder what our world would be like if a substantial fraction of these billionaire budgets at Facebook, Google, Microsoft were spent on Natural Sciences, even if with a commercial scope narrowed to Applied Sciences.
Seems like a less advanced, but clearly more automatable version of the South Korean mother who was able to see her deceased 7-year-old in VR[0][1]. A friend of mine wrote a short story about this sort of thing a couple of years ago[2]. We actually spoke about it, and it got me to wondering what the business model of a service like this could be. Subscribing to receive videos from and communicate with a deceased relative. "Unsubscribing" would be like condemning them to the grave yourself. It's gross and scummy, but I could definitely see it being backed. Money on the table, I guess.
Someone doesn’t necessarily have to be dead for this to work either, which is a little disturbing. Or in my case I want to hook this up to my work Slack so I can take a bathroom break without someone complaining about me being AFK for a couple minutes.
It would then be another Black Mirror episode, the USS Callister one, where (spoilers:) introvert programmer Matt Damon takes DNA of his colleagues and makes virtual copies of them and imprisons them in his play universe.
Obviously the tech won't be that advanced, but having people have emotional connections to chatbots that they programmed to pretend to be their exes would be... creepy.
On the other hand I can imagine at least one Twitter user who we could put in a virtual playpen and it would benefit mankind.. Virtual Pennsylvania Ave 1600..
You can have a bot reply with "ah! Give me a second" or something along those lines while you're away. There's no need for a sophisticated solution, as people already know you're a person and are thus not very likely to suspect a bot (assuming you don't tell anyone that is).
I used to do this while botting MUDs. Immortals (administrators) would check up on you occasionally, so first my script would send a "mis-channel" communication to make it look like I was actively talking to someone else, then a quick "Ack, sorry" followed by a pause and a "one second". All while simultaneously sending me an SMS so I can take over the communication stream in person.
From looking at your application, you were not patenting a chatbot, but a vastly different thing. The word "chat" only appears once in your application, but only as an input source, not a conversation tool.
The patent app you linked doesn't mention a chatbot in any claims, which is all that matters. Saying it was for some project has zero relevance to the patent system.
Care to list the claims in your patent you think are the same as the MS claims?
The current state of chatbots is embarrasingly bad. I still haven't seen a chatbot use-case that can't be solved better with an FAQ section + documentation search and a human as second line of support.
In all fairness I used to hate them but I have seen one or two customer support bots used in a way that definitely helped me get the answers I needed. Perhaps the fact that these bots are generally set up by consultants that understand search rather than some random intern that probably set up the documentation search is the deciding factor there.
I use Alexa on a daily-basis - which can be seen as a chatbot. It provides plenty of useful functionality, especially Alexa-auto which I bought my wife for Christmas. She loves the ability to add items to her todo list whilst driving, asking it questions about the weather, playing music, performing communcation tasks, etc.
> I use Alexa on a daily-basis - which can be seen as a chatbot.
More general virtual assistants (Alexa, Google Assistant, Siri) are at a completely different power level from most chatbots. Most chatbots use cases that I come across are some company thinking that they can replace human support with a language model trained on their custom knowledge base. They expect something as good as Alexa but which also has knowledge about the company's product and the results are always disappointing.
I don't get the hate for chatbots. Sure they are not perfect yet (it's a hard problem to solve!) but I see them as early-stage technology, which will improve over time.
Chatbot mania began in 2016, it’s now 2021 what major improvements have there been? What’s one website that has a good chatbot?
Chatbots are great until
they don’t work which is all the time. You need to ask questions in a specific way to get an answer. From a user point of
view it’s a terrible experience. Most companies don’t value the time of their users that’s why chatbots have been deployed on many websites.
I've personally played some text adventure games from the 90's with better input recognition than some customer service chatbots. They almost feel like trying to figure out the commands from those games sometimes.
'I need help with payment'
Sorry can you repeat your question
'Payment information'
Sorry can you repeat your question
'Help with payment information'
You need help?
Directing you to our FAQ section where you can get help for our most common questions.
I remember reading about a startup that already did that. If I recall correctly, it developed organically out of a situation where the founder had actually lost a friend and trained a chat-bot on their email backlog.
Later she realized that there might be a business idea in this.
A work of fiction can count as prior art? It is not like anyone would not have tried to do something similar in practice, before or after that, but the episode itself?
In any case, that episode qualifies both as prior, and art.
The claims of this patent recite a chat bot system that does specific things a certain way. If it does require a particular algorithm, the algorithm itself is not protected.
New proposed rule: All patents must be accompanied by a working prototype that encompasses all claims. Anything not implemented in the prototype is not in the patent.
> New proposed rule: All patents must be accompanied by a working prototype that encompasses all claims. Anything not implemented in the prototype is not in the patent.
Interesting rule. I'm not sure that a single prototype per patent is reasonable, though, given that claims may cover mutually exclusive claims and divergent use-cases.
A set of prototypes that collectively cover all claims (though not all possible combinations of claims) might work, except that the hard part might still be reducing to practice a particular combination of claims not in any of the prototypes.
Eg. A patent that has claims for electronic storage and playback of music and claims for a handheld data storage device (each claim embodied separately in prototypes) as well as the combination of a handheld device for electronic storage and playback of music (claim not embodied in a prototype).
I'm not sure how to construct the rules in such a way that you aren't either saddling inventors with the need to produce a huge number of prototypes representing a combinatorial explosion of claims and their combinations, nor allowing patent holders to extort subsequent inventors for inventions that they could never have reduced to practice themselves.
Leaving the determination of which combinations of claims to allow as "reducing to practice is left as an exercise for the engineer" up to the patent examiner still leaves in place the existing misalignment of incentives that the current system has, except it makes applying for patents more expensive.
The claims define the protected inventions. In this example, claim 1 recites a very specific way to implement a chat bot.
"A method for creating a conversational chat bot of a specific entity, the method comprising: receiving a request associated with a specific entity; accessing social data associated with the specific entity, the social data comprising at least one of: images of the specific entity, voice data for the specific entity, conversational data associated with the specific entity, and publicly available information about the specific entity; processing the social data using at least one of machine learning techniques and one or more rule sets, wherein processing the social data comprises: identifying conversation data collected for the specific entity; identifying conversation data collected for one or more entities similar to the specific entity; and determining similarities between the one or more entities and the specific entity using at least one of expression analysis techniques, approval indicators, and characteristics comparisons; using the social data to create a personality index, wherein the personality index comprises personality information for the specific entity; and using the personality index to train a chat bot to interact conversationally using the personality information of the specific entity"
Everything except the personality index step seems like obvious boilerplate for how one would accomplish implementing a chatbot that tries to replicate a particular training corpus.
And not very clear what personality index means in this context.
Even if people will know it is a bot, these people's rational capacities will in most cased be diminished under such circumstances, at best. Many will be easily (emotionally) manipulable, even more than with all the psychological trickery in product advertisement and big data exploitation these days.
Opinion: this clearly crosses the line of criminal behaviors, exploiting people who are clearly in a vulnerable positions (at least more than usual). I think the track record of tech companies rather clearly shows that their promised of taking good care of collected data (and privacy) mean little to nothing. At the end of the day, they do this for profit. That's enough for me to make this cross the line.
Also, what's really the "invention" here? Would something like this ever deserve a patent monopoly? If you take (dead) people's personal data and build a neural net (or any other "AI") to mimic/compliment that data (needed for answering questions), have you not just done what any/every neural net already does? If we're going to call that novel and innovative, then hold my beer ... or is this just again a big tech company abusing the patent system to hijack an obvious (use of) technology, excluding anyone who doesn't belong to their cozy little cross-licensing inbreed family from using what should be free to use in the first place?
#end-of-rant