New way to "learn" Chinese

[fcoulter]

This just crossed my Facebook feed. While I doubt that this device will do everything claimed, it may be indicative of a future of language learning.

https://www.indiegogo.com/projects/lily ... -chinese#/

[fcoulter]

If you go to the full story, it mentions the HSK proficiency exam. It claims Level 1 only needs a vocabulary of 150 words, culminating in 5,000 words for level 6. This seems low, or am I missing something?

[zenmonkey]

This just crossed my Facebook feed. While I doubt that this device will do everything claimed, it may be indicative of a future of language learning.

https://www.indiegogo.com/projects/lily ... -chinese#/

That's an interesting page. I'm bolding your statement above because I agree with you.

If you do a search on their company (Maybe) and the key people together ... not much comes up.

They could have positioned their product as a new speaker but have instead niched it into the language learning community.
It’s interesting to also see that maybe.io only sends back to the campaign page and their company has zero web presence.

I could see how the same thing could be developed with the Google, Apple or Amazon speakers or any phone. But I'd really like to see how it analyses pronunciation before I give it a thumbs up.

Seems like the software part is still (just) a lot of promise...

[devilyoudont]

This project is a scam.

The first thing that everyone should know about crowdfunding is that if something isn't on Kickstarter, there is a reason. Now, that reason might be that Kickstarter doesn't allow certain types of campaigns, or that reason might be that Kickstarter is only available in about 20 countries. In this specific case, Kickstarter does allow tech campaigns, and Kickstarter is available in the United States...

Now, something to know about Indiegogo: It's standards are way lower than Kickstarter's. Specifically relevant: Kickstarter requires that a campaign like this have a working prototype. Additionally, 3D used to suggest a prototype (used in photos and in campaign videos in this case) is specifically banned from Kickstarter.

Conclusion: This company does not have a working prototype.

Further conclusion: Do not invest in a tech start up that does not have a working prototype.

I'm gonna break down how I can tell this is a scam in further detail, in the hopes that no one here loses their money. But the above let me know that it's a scam within three seconds of clicking on the project. I understand people might think that that is a knee-jerk reaction tho, and may not believe that it is a scam as a result. Looking into the project in further detail, I am even more confident that it is a scam.

First of all, you should know that there is a plague of robot assistant scams on Indiegogo. Here's a funny video about them. (Warning: strong language, some mature themes because I forgot what one of the robots does ) The Lily robot might seem more believable than the ones featured in this video because the claims made in this video are truly outrageous. But, we are in the same ballpark here, and the guy who made the video would have skipped more boring examples of the problem.

Here are some more red flags I see:

Does the math make sense?
This is always one of the clearest indicators that something is a scam. Scammers can make all kinds of claims about money, hoping that no one thinks about it even for a second because you are so dazzled by their presentation.
--Why is this project using this investment model? Crowdfunding is a huge pain, and most people who have run successful campaigns say that traditional investment is easier and has better outcomes in most cases. This project promises a lot of impressive things. If a functioning prototype exists, there should be no problem getting this funded thru traditional avenues.
--What exactly do they need the money for? You want to see a breakdown of this kind of thing because it indicates if the creator's are actually going to be able to manage the investment. You can have the best invention in the world, and that doesn't matter if you aren't able to actually manage money. This project doesn't say anything about what the money is for, so there is no way to know whether or not they have a realistic ability to manage the money, and actually deliver.
--Why is the goal amount so low? 10k is literally nothing when it comes to investment for hardware. It's not even an advertising budget. This leads me to believe that the creators have no realistic idea of how much this project should cost. AI, hardware that needs to be shipped out, experts on software engineering and instruction of Chinese as a second language? This project needs at least 1 extra zero on the end to be real...
--Even if all the above wasn't true, manufacturing crowdfunding campaigns often fail because they are too successful. The creator's had everything worked out, but aren't able to scale up, and so they wind up ruined due to their own success (yet another reason people generally prefer traditional modes of funding when they can get it).

Given the above, the only way that this isn't a scam, is if the product is 100% ready to go, but they decided to use indiegogo as a storefront for some reason. Generally speaking the reason would be either that the product is so shitty retailers don't want it, or a market doesn't actually exist for the product.

Who are these people?
You want to know this because you want to know if they are qualified to make this product, and you want them not to be anonymous because they take no hit to their reputation if they just vanish with the money.
--No names listed on the project
--"maybe speaker" is functionally un-google-able.
--The name they use to lend authority to the project is a speaker company that they used to work for? No proof they actually worked there without their names.
--No website
--No social media presence
--Who is their expert on Chinese?
--Who is their software engineer for the AI?
--Who developed the app?

Project is too complex
Generally speaking, a project from someone who is just starting out (thus, no name, no website, etc etc) should be as simple as possible. Every additional complexity is an additional failure point for the project. Hypothetically, a legitimate project on this scale could totally fail due to something like problems with integration to iOS. It's a bad idea to invest in something so complicated, unless the creators have a track record of projects that hypothetically could build towards something like this.

Misleading potential investors
--The clearest example here is that they are pretending to have a prototype, and they don't. If they had the prototype, they would be on Kickstarter, or not bothering with crowdfunding. The images where you see the guts inside the Lily are all 3D renders, not a prototype. The different color robots that are in the video and in photos are, very likely, just models. The closest thing that this campaign shows to a prototype is a series of still images showing gesture based music controls. What do music controls have to do with the core features of this product? Why can this campaign produce 2 very nice videos as advertisements, but this functionality is only shown in stills? Certainly video would be a much better way to show off these features. Why is this the only functionality that they give any real details on, when the core features of this are all voice activated AI?
--If they are already manufacturing, why do they need this money, and why don't they have a website?
--The focus of the project is completely wrong. The hard part of this project is the software. All the details on development are about the speaker and the fabric that covers the robot, both practically irrelevant from a language learning perspective. Who developed the software, how was it developed, who on your staff speaks Chinese?
--I assume that the campaign video is a total fraud, but in the case that there is a prototype, it is heavily doctored. The leaders in this industry cannot produce a robot that responds like this in a monolingual context. It's insanity to imagine some audio engineers on Indiegogo can do something that AI specialists at Amazon cannot.
--This technology, such that it already exists, has huge problems around accents. I can find articles as recent as this summer that Alexa cannot even correctly respond to native but non-standard English. This product would have this problem twice. First, it needs to understand your English, accented however it may be (most people don't perceive me, a speaker from a major American city, as having an accent, but google assistant has constant trouble picking up that I am trying to activate it, and understanding what I say), and then it also needs to understand your Chinese, which will absolutely be accented, and apparently it will be able to correct your non-standard Chinese? No way on earth. But they attempt to just hand wave away this problem in their FAQ

Accent recognition is a difficult problem in voice-controlled systems but solutions exist.

Someone forgot to tell Google, Apple, and Amazon I guess.

In conclusion, a scam.

On topic: Is something like this the future of language learning? It's a nice idea, but I'm not betting on it, until somebody somewhere has a working prototype.

Correction 1: They do have their names attached to the project. They are both speaker engineers who have worked for that speaker company. There is still no information presented about the other members of their team, who made the AI, etc. From my perspective, that is the most important part of the project, and if the prototype is just a speaker prototype, and not an AI prototype, it is still no prototype at all.

Correction 2: The first time I trawled thru their Indiegogo page, none of the gifs worked. Probably some issue on my end, not theirs. So the still images I referenced above about gesture controls are actually animated gifs. This is a better demonstration of a music player than what I previously said.

[zenmonkey]

Who are these people?
You want to know this because you want to know if they are qualified to make this product, and you want them not to be anonymous because they take no hit to their reputation if they just vanish with the money.
--No names listed on the project
--"maybe speaker" is functionally un-google-able.
--The name they use to lend authority to the project is a speaker company that they used to work for? No proof they actually worked there without their names.

Umm, their names are there and both are real people (and well known in the speaker community)?
I'm not sure their id's aren't being falsely used but if Timbers and Wehner are really part of this project then they are quite the reference.

Rain Factory is also the marketing company that successfully presented the Fluent Forever campaigns...

So, while I'm not investing in this company (see my comments in my first post) I'm not quite as quick as you to categorically condemn them. (Also please be careful in the language you use - don't bring a libel risk to the forum.)

But it is true that neither their company or the two people listed in that page made it into Crunchbase database for this project.

Caveat emptor!

[devilyoudont]

I missed their names. In my defense, I watched the main video twice, and triple checked the Project owner page and the FAQ, but only looked thru the rest of the project once, where their names are listed on a single image.

So I think you're right that this is probably a speaker that got re-purposed. They very likely do have a very high quality speaker prototype. But I still think it's incredibly unlikely that they have an AI enabled prototype, for reasons listed above. The important features for this product is the AI, and so I stand by what I said above.

I'll add a correction to my post above.

[zenmonkey]

I missed their names. In my defense, I watched the main video twice, and triple checked the Project owner page and the FAQ, but only looked thru the rest of the project once, where their names are listed on a single image.

So I think you're right that this is probably a speaker that got re-purposed. They very likely do have a very high quality speaker prototype. But I still think it's incredibly unlikely that they have an AI enabled prototype, for reasons listed above. The important features for this product is the AI, and so I stand by what I said above.

I'll add a correction to my post above.

You are absolutely right that the "ai" (AKA - voice recognition software, command structure analysis, response generation functions) are key. Even with Alexa or Siri or Hey Google it is hundreds of engineering years to get things this far. It might be quite easy to develop some of this using quite a few of existing libraries (I integrated a Speech to Text library in about 30 minutes on a software prototype) but getting that to generate responses is a whole other kettle of fish. And then to have a coherent learning program...

If I were to tackle this, I'd probably start from the Alexa API which already does about 30% of the work. But they specifically state that they do not work with Alexa.

The best bet is to wait a few months and see if anything ships and if so, how that works. In any case, their proposed price point at x times the cost of an Apple or Google product won't survive market pressure, in my opinion.

[taylorg]

I also felt this could be a scam. They asked me to email them (not sure why?) And here is their response they sent me (below). What do you guys think? Scam or no scam? I'm still not sure. Can we investigate their company info and the people listed below?


About us: We’re a seed-stage company headquartered between San Francisco, USA and Shenzhen, China, with a team of 15-20 people.

We’ve currently set up our website, https://www.maybe.io, to redirect to our IndieGogo page for the duration of our crowdfunding campaign because we want to capture this traffic. At the end of the campaign, our website will revert back. We’re actively working on getting our first press placement, but boy, is there competition from the politicians this week!

Our founder, Jie Meng-Gerard, was a "rocket scientist" working on algorithms for satellite launchers during his PhD in Artificial Intelligence at Pierre and Marie Curie University in Paris. He was the co-founder of voice-controlled speaker Whyd. He started working on voice-controlled speakers in 2014 before the Amazon Echo existed and has since led the development of 3 voice-controlled speakers, including one with Baidu and one with another leading Internet company. Jie is also a YCombinator alumn and speaks English, Chinese, French and German.

Our global operations manager Alexis Pons has a Master in Marketing strategy from Paris ISG Business School and studied at Fudan University in Shanghai. Alexis started his journey in Shenzhen as the CEO of MiH Innovation, a manufacturing platform that helps bridge hardware startups and Chinese manufacturers. He is also the R&D startup consultant of 3NOD group. In 2015, he started working on voice-controlled speakers with 3NOD. He was featured in the Wired video "Shenzhen, the Silicon Valley of hardware".

Our lead hardware manager, Johnny Xiao, has more than 10 years of experience with hardware as product manager and team leader. He majored in Electrical Engineering from the University of Xian. He worked on hardware development at ASUS, Huawei and Flextronics, building high-profile products, including for Google, and leading team of 100+ people. In 2016, he started developing voice-controlled speakers.

Jie and Maybe's voice AI team have a deep experience in voice technology and signal processing, developing complete "Voice Operation Systems" like Alexa. Our team has developed microphone array solutions, noise cancellation, wake word, natural language understanding and speech synthesis technology. The rest of our team comes from varied backgrounds including top tech and top consumer companies. We all share a passion for education and AI. While Lily is our first product launch together as a company, we have collectively launched more than 100 products and are excited to continue this track record.

For Lily, there are a number of product demos including touch interface and voice control embedded in the IndieGoGo content - I encourage you to take a look and explore these features. We've also worked with Greg Timbers (JBL) and Jason Wehner (Harman) for more than a year to develop the audio architecture of Lily.

Maybe as a company is funded - we do not wish to disclose the amount - and backed by Tencent, the number 1 company in Asia, through the Tencent AI Accelerator, one of the most renowned tech accelerator in China. Maybe is also developing AI chipsets to accelerate voice processing on embedded systems, including neural network instruction set architectures, IP Core design and assembly technology.

If you aren’t confident in backing our campaign at this time, we’d still love to have you support us. We’re expecting to expand into more regular online sales later next year and in the meantime can add you to our email list.

Let me know if you have any more questions and thanks for your interest.

[devilyoudont]

I remain confident in my above assessment. Maybe I will wind up eating crow.

If you are really interested in this product, zenmonkey is right that you should wait a few months before buying.

They are supposed to be delivering in March 2019. That will be here before we know it. And in April we will know: 1) if it actually shipped out and 2) how it performed based on reviews from customers.

If you're worried about the price, I would wait for like the second or third iteration of the product anyway.

[Andreas]

Well. It is two years later and it seems you were right.