I'm amazed at the speed and pace of technological advancements being made by the world's top scientists and computer engineers. In this article I will explain why artificial intelligence is moving beyond recognition and that fact is demonstrated by the latest progress with Google's voice generating AI.
What if Steem and the largest social networking site platform that sits on top of it, Steemit were able to integrate this technology before other companies jump on board?
Imagine if there was a speaker button next to this post and it could be read to you and sound just as natural as if I was reading it to you myself... There's some food for thought right?
Cryptocurrencies are still in their infancy and the Steem and SBD cryptocurrencies stand to benefit a lot from additional exposure through networking and partnerships. Doesn't this seem like a good application for increased integration with the business world?
This idea could get us some amazing press if developers got in the train quickly with this one. If anyone knows how I could get in contact with the developers of Steem and Steemit and pitch this idea to them myself, please let me know!
Anyways lets dig deeper into the Tacotron and let me try to explain why it is such a groundbreaking advancement of modern text to speech AI functionality!
If you've ever used Google products you may be familiar with Google's AI voice. There are two settings, one for male and one for female voice and when you hear it you'll know something is majorly off with unnatural pronunciation similar to that of Apple's Siri or Microsoft's Cortana.
But every year the unnatural sound that computer generated voices make is becoming less recognizable and now you'll simply be amazed as Google's voice has started to sound less robotic and more like a human and in fact, the new Tacotron 2 Google voice AI is almost undistinguishable from a human voice.
In a recently published research paper by the people at Google, the team introduces details to the impressive speech system called Tacotron 2. In the paper, Google highlights the systems ability to speak almost identically to its human creators.
The team describes the second generation speech system in the report stating, " The Tacotron 2, a neural network architecture for speech synthesis directly from text. The system is composed of a recurrent sequence-to-sequence feature prediction network that maps character embeddings to mel-scale spectrograms, followed by a modified WaveNet model acting as a vocoder to synthesize timedomain waveforms from those spectrograms."
According to the report the technology that powers this new text to speech tool consists of two deep neural networks. Basically, the way it works is that the first network translates the text into a spectrogram. After the spectrogram has been made it is sent to WaveNet, a new system specially designed by Google's Deep Mind AI program. The process results in a computer generated voice that sounds like its human counterparts.
Take a listen of Google's new Tacotron for yourself and you may just be surprised how natural it sounds!
Techblogger Wrap Up
This technology is miles ahead of Siri or Amazon's Alexa. The fact that its almost undistinguishable from a real human's voice is concerning to say the least.. Don't you think Steem / Steemit could benefit from being a first mover in the integration of this new technology into their platforms?
Do you think because of this new natural sounding AI text to speech voice capabilities, we may find ourselves become the victim of some sort of AI scam down the road? What types of positive applications and nefarious use can you imagine being conducted through the assistance of this new technology?
Please leave your thoughts and comments below.
Source:
Google's New Voice AI is Hyper Realistic - Interesting Engineering
Image Source:
Interesting Engineering
Very interesting developments indeed. AI coupled with TTS technology appear to have advanced to such a degree that the differences between the synthetic voice and actual human voice have begun to blur. However, I would like to point out that, although the example provided by Google do sound rather natural, no systems to date have matched the complexity of human language. Take for example the sentence, who are you? Depending upon the inflection of the voice, the words could imply inquisitiveness, fear, anger, or even arrogance. And for good measure, let's through in regional accents and dialects. As one can see in these examples, how people utter words carry more information than just the meaning of the words themselves. And, no systems that I am aware of can match the nuances conveyed in human speech. So there remains a lot more work to do in this area of research.
Near-term use cases. Assistive technologies software is an implementation that can, for example, help visually impaired users access the features of their electronic devices to a fuller extent. Imagine such a person accessing reading materials using voice commands and having the a device read aloud the material requested. Include the ability to bookmark, highlight passages, and add reference notes, and the world of information available to the visually impaired expands exponentially.
Other likely implementations of this technology would be in help desks, call centers , 911 services, etc. The primary drivers behind the adoption of the technology the desire to reduce personnel costs and to provide 24/7 365 availability.
To address the point about having use of this technology on the Steemit platform, I would suggest that even if Google would allow such a scenario, it would be for non altruistic reasons. Business are motivated by growth, domination and profits, Google being no exception. It is also wise to remember that Google owns the technology patents behind Tacotron and as such can dictate if and to what extent Steemit could utilize the technology. And it could pull the plug on its usage at any time. So, although the idea of integration into Steemit sounds pretty appealing, unless Google decides to open source the technology, usage on Steemit is fraught with risks.
Nice topic for discussion, many thanks.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
One of the best comments I've seen in a while. Many great points @jaalig
I wish author of this topic would reply to his followers. But he seem to ignore them :(
Would you mind sharing with me what do you think about AI being able to operate on social media platforms like steemit? Would AI be able to create content,comment posts and make it look like a human being?
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Many thanks for your kind words. Since the author did solicit feedback from his readers its quite disappointing that he has not responded in kind. Even a thanks for dropping by would have been nice.
As to your question about AI implementation on Steemit, I think that the jury is still out on that one. Although the technology has shown great promise, especially with further advancements on the neural network front, there nevertheless remains much work to be done. Evolution has had roughly a million years to work on homosapiens cognitive and communication abilities. So attempting to replicate this degree of development in a comparative amount of time is a monumental undertaking.
From Charles Babbage's Difference Engine to the most recent development in "deep learning" as being research by the likes of IBM, Google, et al., AI has only relatively recently begun to display what we consider rudimentary forms intelligence. AI has shown itself quite capable of responding to problems within well-defined parameters, it is in understanding nuisance and being able to draw conclusions when confronted with limited information and unknown variables that seems the Achilles Heel of the current state of the technology. No AI, as far as I am aware, would stand up to a Turin Test.
Some have postulated whether quantum computing will prove to be the holy grail to this challenge, but, the science of quantum computing is even newer that AI and in and of itself, a very tough nut to crack.
One last point, if the technology does manage to advance to such a degree that it begins to exhibit nearly indistinguishable characteristics when compared to sentient life, what moral challenges will society have to face?
Again, thanks for your comment and the additional topic of discussion.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
i'm very sorry for such a late reply. have been away for several days but Im back now :)
just wanted to thank you for your previous comment
I wish i could upvote it but it's a bit to late. Im big fan of AI and blockchain so i will be following you closely as you seem to be very responsive
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Great article! This technology really does excite me, but I too have my reservations. First, I think this is a huge win for those who are highly visually impaired. This will greatly increase their ability to consume written content in a pleasing manner.
I am really curious how easily the voice inflection, and overall characteristics can be changed once this hits mainstream. Immediately, my mind started to wonder if this could be integral to audio book pricing. If I could get my hands on cheaper audio books, I would definitely not complain.
My reservation is that this technology could potentially be used to blackmail people. If this is really unrecognizable from a computer, I am concerned that it could become unrecognizable from a specific person in a legal battle. Maybe that is where block chain comes in? Forcing all generated voices and speech to be registered on a blockchain could safeguard against this.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Would you mind sharing with me what do you think about AI being able to operate on social media platforms like steemit? Would AI be able to create content,comment posts and make it look like a human being?
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
This is a great initiative @techblogger
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Great job awesome addition to steemit its nice work for steemit platform from where artificial intelligence will be part of promoting decentralization your post is creativity booster welldone keep going God bless you
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
I can see an AI voice being used to frame someone for a crime.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Indeed that could easily happend.
Im huge fan of AI but I can easily imagine how it's power can be abused if used incorrectly.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Buen Post, sobre los avances tecnológicos para los programas en relación a esta plataforma.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
Im absolutely big fan of AI and blockchain and I found your post very informative.
I checked your profile and I noticed we share many common passions. Hope we can stay in touch :)
Thanks buddy. Followed and upvoted.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
One thing that scares me is the fact that if AI will one day can be used on social media to act like a real human and then recognizing real/genuine comments coming from real human being will be hardly possible.
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
The possibilities of this certainly pricked my interest when I scrolled past. Really cool man!
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit
However it will be used to decide who to kill by the US military!
https://www.google.com/amp/s/amp.theguardian.com/technology/2018/mar/07/google-ai-us-department-of-defense-military-drone-project-maven-tensorflow#ampshare=https://www.theguardian.com/technology/2018/mar/07/google-ai-us-department-of-defense-military-drone-project-maven-tensorflow
Downvoting a post can decrease pending rewards and make it less visible. Common reasons:
Submit